Skip to main content
Industry security guideReviewed

Ecommerce website security: test the order path where customer data and money move

An online store passes personal details, money and stock through a chain of steps: customer accounts, carts, coupons, the handoff to a payment provider, refunds, exports, fulfilment webhooks and loyalty apps.

This guide follows that chain across platforms, shows where access breaks at each step and which guide covers each flaw, sets out the payment-page and privacy questions a store settles with others, and gives you an order-path abuse and evidence matrix to run on a staging copy of your own store.

Order-path abuse and evidence matrix

Twenty-seven rows follow an order through a store, grouped by step: before testing (B), customer accounts (A), carts and checkout (C), coupons (D), the payment handoff and its confirmation (P), payment page scripts (S), refunds (R), exports (E), fulfilment and connected apps (F), loyalty and balances (L), and change, clean-up and evidence (Z). Read each row from left to right: the data the step carries, who should and should not reach it, the test case, the expected result, the evidence to keep and what it may relate to, the limits of the row, and its references.

How to use it

Rows B1 and B2 come first. Rows B1, P2, S2 and F1 are read only and change nothing: they read settings, lists and documents. S1 loads pages, as a visitor on the live store and as cust-a on the staging copy, and changes nothing beyond one test cart on the staging copy, and B2 reads settings and then creates the test identities on the copy. D1, R1 and E1 open with a read-only step. Z1 and Z3 are planning rows, and Z2 deletes only what the run created. Every step that places an order, submits a form or changes a request runs on a staging copy of your own store, with each payment provider switched to the test or sandbox mode that provider documents, using only the run's test identities (the invented customers cust-a and cust-b, whose inboxes and phone numbers the business controls, and staff test accounts), with a marker such as ORDERTEST-1115 in every test record, never a real card, at most 20 checkout submissions across the run, and no load. If any response shows a real person's details, stop the row and record it. Replace the example names with your own, mark a step your store does not have as not applicable, and write your expected result before you send anything. Never record a password, key, signing secret or card number, and never copy into the evidence a link that opens an order, a file or a payment without a sign-in. WSTG-BUSL-10 comes from the edition the OWASP site labels Latest, read on 15 November 2026; the other WSTG ids are from v4.2, and ASVS ids from 5.0.

27 rows

Order-path abuse and evidence matrix

Row and stepData involvedWho should reach it, and who should notTest caseExpected resultEvidence to keep, and what it may relate toLimits of this rowReferences
B1. Before testing: terms and validation scopeNone. This row reads documents and asks questions.The store owner decides what is tested. The platform, the host, each payment provider and each app vendor set terms for their own systems. The acquirer or payment brand is who the PCI SSC sends merchants to about compliance validation.Read only. List every system the later rows touch: the platform, the host, each payment provider, and each fulfilment, accounting, loyalty, review or gift card app. For each, read its own terms on security testing and on test or sandbox use, note the date, and ask the vendor in writing where the terms leave doubt. Ask the acquirer or payment provider in writing how the store's PCI DSS validation is set, including whether a self-assessment questionnaire is used and which one. For your platform, read the platform guide's section on its terms first.A dated answer for every system before any row that sends a request. A system whose terms leave doubt stays out of the run until its vendor answers.Each terms page by address and date read, each written answer, and the acquirer's or provider's answer on validation. May relate to: PCI DSS validation, which the PCI SSC's Merchant Resources FAQ sends to the acquirer (merchant bank) or payment brand.Read only. No request goes to the store or to any vendor's system.PCI SSC Merchant Resources FAQ; PCI SSC post of 10 March 2025. No vendor's permission is stated here: each comes from that vendor.
B2. Before testing: the staging copy, test modes and identitiesWhatever the copy holds. A copy cloned from the live store can hold real customers' records and live connections.The person who manages the copy changes its settings. This row only reads them, apart from creating the run's test identities.Read only, then create identities. Confirm the copy belongs to the business and is not the live store. Read each payment provider's mode on the copy and on the live store, and each connection on the copy for fulfilment, shipping, accounting, email, SMS, loyalty, reviews and marketing. Then create invented customers cust-a and cust-b, each with an inbox and a phone number the business controls, a staff test account for each role the rows need, and the marker ORDERTEST-1115, or your own, in every name.Every provider on the copy in its documented test or sandbox mode, every connection off or pointing at a test account, and the live store in live mode. If any reading says otherwise, the run stops there; switching a mode or a connection belongs to the person who manages the copy, and this matrix does not do it.Each mode and connection as read, with the date and the reader. WSTG-BUSL-10 says test card details should only be usable on development or sandbox versions of payment gateways, but may be accepted on live sites if they have been misconfigured. The live mode is recorded for that reason.Read only, apart from creating the test identities on the copy. Where a mode can be read only on a screen that also shows keys, ask whoever manages the integration instead. If the copy holds real customers' records, never open them; rows that list or export records then stop at their read-only part.WSTG-BUSL-10, Test Payment Cards.
A1. Customer accounts: orders, invoices, returns and trackingNames, email addresses, phone numbers, delivery and billing addresses, items bought, order totals, the payment summary the store shows, invoice and receipt files, and return or refund requests.The customer who placed the order, and staff in roles that handle orders. Not another customer, and not a visitor with no session.As cust-a, place one paid test order and note each address that names it: the order detail, the invoice or receipt file, a return or refund request, and the tracking page. As cust-b, request each of those addresses with cust-a's order id or number in place of cust-b's own, then send each request again with no session and after signing out.Every request refused with the same response the store gives for an order that does not exist, and no response body holding any of cust-a's details.Each request and response, marked with the account and session used, with personal details and any order or tracking token redacted. A link that opens an order without a sign-in is recorded by the page it opens, never copied. May relate to: the steps APP 11.1 of the Privacy Act 1988 concerns, where the Act applies to the store.Staging copy of your own store, payment provider in test mode, cust-a and cust-b only, one checkout submission, each request once, emails only to inboxes the business controls, no load, never a real card. If any response shows a real person's details, stop, record the request and tell the owner.WSTG v4.2 WSTG-ATHZ-04 and WSTG-ATHZ-02; ASVS 5.0 v5.0.0-8.2.2.
A2. Customer accounts: guest order lookup and emailed order linksThe same details as A1, for customers who bought without an account.The person holding both the order number and the email address used for it, or the link sent to that address. Not someone holding only one of them.Where the store offers a guest order lookup or emails an order link: as a guest, place one test order with cust-a's details. Look it up with cust-b's email address. Open the emailed link, then change the order number or id in it to that of A1's order. Only where B2 found no real customers' records on the copy, try the order numbers either side of the guest order in the lookup, once each.The lookup answers only when the number and the email address belong together, the link shows only its own order, and the neighbouring numbers are refused.Each lookup request and response, redacted. The emailed link is never copied into the evidence: record only whether it carries the order number in plain view and whether the changed link was refused. May relate to: APP 11.1, where the Act applies to the store.Staging copy, payment provider in test mode, test orders from cust-a and cust-b only, one checkout submission, at most five lookups and link requests in all, the emailed link opened only for the run's test orders, emails only to inboxes the business controls, no load, never a real card. If a real order appears, stop and record it.WSTG v4.2 WSTG-ATHZ-04. The lookup and link checks are this page's own.
A3. Customer accounts: saved addresses and saved payment methodsSaved delivery and billing addresses, phone numbers, and whatever reference to a saved payment method the store keeps.The account holder. Not another customer.Give cust-a a saved address and, using the provider's documented test card, a saved payment method. As cust-b, read, edit and then delete cust-a's saved address by its id. At checkout as cust-b, submit cust-a's saved payment method reference in place of cust-b's.Each request refused, cust-a reads the address back unchanged, and the checkout refuses a payment method that belongs to another customer.Requests and responses, redacted, and cust-a's read-back. Record the payment method reference as present or absent, never its value. May relate to: APP 11.1, where the Act applies to the store.Staging copy, payment provider in test mode with its documented test card, never a real card, cust-a and cust-b only, each attempt once, at most one checkout submission, the delete only on the address created for this row, emails only to inboxes the business controls, no load. Stop and record if a real person's details appear.WSTG v4.2 WSTG-ATHZ-04; ASVS 5.0 v5.0.0-8.2.2.
C1. Carts and checkout: prices and totals in the requestProduct ids, quantities, unit prices, shipping costs, discounts and order totals.The server sets every price and total. The customer chooses products and quantities only.As cust-a, add one product to the cart and start checkout with the browser's developer tools open. If any add-to-cart or checkout request carries a price, a line total, a shipping cost or an order total, including in a hidden field, change one value at a time, submit, and complete the test payment if the store accepts it.The changed value is ignored or refused, and the order total equals the catalogue price plus the shipping the server calculates. WSTG-BUSL-10's remediation says the only things included in client-side requests should be item IDs and quantities.The original and changed requests, the order total that resulted, and the catalogue price on the day.Staging copy, payment provider in test mode, cust-a only, never a real card, at most two checkout submissions, one value changed per request, emails only to inboxes the business controls, no load. If a response shows a real person's details, stop and record it.WSTG-BUSL-10, Price Tampering on the Application; WSTG v4.2 WSTG-BUSL-03; ASVS 5.0 v5.0.0-2.2.2.
C2. Carts and checkout: quantitiesLine quantities and the totals built from them.The server decides which quantities are valid, including any pack sizes the store sells in.In the cart requests, set a quantity of 0.1, then 0, then -1, then a value outside the store's own pack sizes or drop-down choices. Remove one item repeatedly until its quantity would fall below zero. Send one request with the quantity parameter twice, such as quantity=5&quantity=4.Each value refused or corrected to a valid whole number, and the cart total never below the price of what is left in the cart.Each request, the cart as the server returns it, and the total.Staging copy, cust-a only, cart requests only. If a value is accepted, at most one checkout submission, with the payment provider in test mode, to see whether it can be paid. Never a real card, emails only to inboxes the business controls, no load. If a response shows a real person's details, stop and record it.WSTG-BUSL-10, Quantity Tampering; WSTG v4.2 WSTG-BUSL-01; ASVS 5.0 v5.0.0-2.2.1.
C3. Carts and checkout: currency, shipping and minimum basketCurrency code, delivery address, shipping cost and minimum order value.The server applies the store's rules at the final step, whatever the cart said earlier.Where the store sells in more than one currency, change the currency code in the checkout request and keep the amount. Change the delivery address after shipping has been priced. Remove items after a shipping rate or a minimum basket value has been applied.The total worked out again from the server's own rules at the last step, in the currency the server set, with the shipping for the final address and contents.Each request, and the order total that resulted.Staging copy, payment provider in test mode, cust-a only, never a real card, one checkout submission, emails only to inboxes the business controls, no load. If a response shows a real person's details, stop and record it.WSTG-BUSL-10, Currency Tampering and Breaking Payment Flows.
C4. Carts and checkout: changes after payment startsCart contents, the amount sent for payment, discounts and the paid order.The order holds what was paid for, and nothing else.As cust-a, start payment for a cart, then add an item from a second browser tab before the provider confirms. Change the cart after a coupon has been applied. After checkout completes, send a request that changes the order's contents.The order holds only what was paid for, any added item stays unpaid or the payment is refused, the discount is worked out again for the changed cart, and the customer cannot change a completed order's contents.Requests and responses, and the order as staff see it afterwards.Staging copy, payment provider in test mode, cust-a only, never a real card, one checkout submission, emails only to inboxes the business controls, no load. If a response shows a real person's details, stop and record it.WSTG-BUSL-10, Adding Items After Payment Initiation and Breaking Payment Flows; ASVS 5.0 v5.0.0-2.3.1.
D1. Coupons: the rules for each codeCoupon codes, discount values, usage limits, allowed customers and products, and expiry dates.Eligible customers apply codes. Staff in the roles the store names create them. Nobody else should be able to find one.On the live store, read the coupon list without saving and note any codes that follow a pattern a customer could guess, such as a word plus a number. Search the copy's checkout page source for any code. Then, on the copy, create a test coupon limited to one use per customer and set not to combine, and as cust-a apply it twice, apply it together with a second test coupon, apply it to a product it excludes, and apply it after its expiry date.Each attempt refused with the store's message, no code in the page source or a hidden field, and no live code that follows a guessable pattern.The coupon settings as read, with each live code recorded by its pattern rather than in full, each message word for word, and the order total.The list and page source are read only. Requests only on the staging copy, with test coupons made for this row; no code is tried against the live store. cust-a only, at most one checkout submission, payment provider in test mode, never a real card, emails only to inboxes the business controls, no load. If a response shows a real person's details, stop and record it.WSTG-BUSL-10, Discount Codes; WSTG v4.2 WSTG-BUSL-05.
D2. Coupons and stock: a one-time code and the last unit at the same momentA coupon's usage count and a product's stock level.One redemption per single-use code, and no more units sold than exist.Create a single-use test coupon and a test product with one unit in stock. As cust-a, prepare three checkouts of that product with the coupon applied, and submit them at nearly the same moment, once.One order succeeds, the coupon's usage count reads one, and stock reads zero, never below. WSTG-BUSL-10 lists multiple success responses for a single-use action and resources allocated beyond stated capacity among the indicators of vulnerability.All three responses, the orders created, and the usage count and stock level afterwards.Staging copy, payment provider in test mode, the test coupon and test product only, cust-a only, never a real card, three checkout submissions sent once, with no repeats, emails only to inboxes the business controls. This is a timing check, not load testing. If a response shows a real person's details, stop and record it.WSTG-BUSL-10, Race Conditions; ASVS 5.0 v5.0.0-2.3.4. The race conditions guide covers the fixes.
P1. Payment handoff: what the store sends to the providerOrder id, amount, currency and sometimes item lines, with card details typed at the provider's page or fields.The store sets the amount. The customer types card details only into the provider's page or fields.Note which of the methods WSTG-BUSL-10 lists the checkout uses: a redirect to the provider, the provider loaded in an iframe, an HTML form posting to the provider, or card details taken by the store and posted to the provider's API. Where an amount, currency or item total travels in a form field or an iframe address, change it once and complete the test payment.The provider charges the amount the store set, or the store does not mark the order paid because the amount paid differs from the order total.The method, the field that carried the amount, the amount the provider's test dashboard shows, and the order's status. Keep the method with B1's questions for the acquirer; this page draws no conclusion from it.This row sends a request to the payment provider's test environment, so it runs only once B1 holds the provider's terms. Staging copy, payment provider in test mode with its documented test card, never a real card, cust-a only, one checkout submission, emails only to inboxes the business controls, no load. If a response shows a real person's details, stop and record it. WSTG-BUSL-10 notes that payment gateways are usually run by third parties and may not be included in the scope of testing, and that while price tampering may be acceptable, other types of attacks, such as SQL injection, should not be performed without explicit written approval.WSTG-BUSL-10, Payment Gateway Integration Methods and Price Tampering on the Payment Gateway.
P2. Payment handoff: signing or encryption settingsThe integration's settings, and the names of the keys or secrets it uses.The person who manages the payment integration.Read only. From the provider's own documentation, list each option it offers to sign, hash or encrypt the transaction details. Ask the person who manages the integration whether each one is switched on, and the name of the setting or variable that holds each secret, rather than opening a settings screen that displays keys.Each option the provider documents is on, or the reason it is off is written down with who decided. WSTG-BUSL-10's remediation says to review payment gateway documentation and use all available security features, such as encryption and secure hashes.The option names, on or off, the date and the reader.Read only. Never open, reveal or copy a key or secret.WSTG-BUSL-10, Remediation.
P3. Payment confirmation: the return pageThe order's paid status, its fulfilment and its confirmation emails.Only the provider's confirmation should mark an order paid. Not the customer's browser.As cust-a, place one test order and leave it unpaid at the provider, then request the store's success or return address directly. For a second test order that was paid, load the success address a second time.The unpaid order stays unpaid and is not sent for fulfilment, and the second load creates no second fulfilment, email or loyalty credit.The requests, the order statuses and notes, and each email that arrived at the controlled inbox. Keep both orders for P4, R1 and R2.Staging copy, payment provider in test mode, cust-a only, never a real card, two checkout submissions, each address loaded at most twice, emails only to inboxes the business controls, no load. If a response shows a real person's details, stop and record it.WSTG-BUSL-10, Breaking Payment Flows; ASVS 5.0 v5.0.0-2.3.1.
P4. Payment confirmation: notifications from the providerPayment events carrying an order reference, an amount and a status.The provider sends them. Nobody else should be able to move an order with one.To the staging copy's notification address only, send one notification with no signature, and one copy of a genuine notification with its body altered. Then resend P3's genuine paid notification once, from the provider's test tools where it offers that.The unsigned and altered notifications refused, and the resent one answered without a second paid status, fulfilment, email or loyalty credit. WSTG-BUSL-10 says that if the back end lacks proper idempotency checks, a gateway callback replayed rapidly in parallel can, among other things, trigger multiple order fulfilment events or mark the same order as paid multiple times.Each delivery and the store's response, and the order notes afterwards.The staging copy's address only, never the live store's. Payment provider in test mode, P3's test orders only, each delivery once, emails only to inboxes the business controls, no load. Resending from the provider's tools follows its terms, recorded in B1. If a response shows a real person's details, stop and record it.WSTG-BUSL-10, Race Conditions. The Stripe webhook security guide covers signatures, timestamps and duplicate events in depth.
S1. Payment page scripts: what loads, and who approved itCard and personal details typed into the checkout, and anything else a script on that page can read.Scripts the business has a reason for and has approved. Nothing else.Read only, apart from one test cart on the copy. On the staging copy, as cust-a, put one product in a cart and open the checkout and payment pages a single time each while the developer tools record network requests, leaving the payment fields empty. On the live store, open one product page and the empty cart page a single time each, without adding anything. For each script host, record its business reason, who approved it, and how its integrity is checked, such as Subresource Integrity, which MDN says works by providing a cryptographic hash that a fetched resource must match. Note any script host that the live pages load and the staging checkout does not, or the other way round.Every host has an owner, a reason, an approval and an integrity method, or a written reason for none. A host nobody can explain goes to the owner as a question.The dated script list. May relate to: PCI DSS v4.x Requirements 6.4.3 and 11.6.1, which the PCI SSC post describes as focused on payment page scripts being properly authorised, checked for integrity and monitored for tampering. Whether they apply is settled with the acquirer, as B1 records.Page loads, plus one test cart for cust-a on the staging copy, which goes on Z2's list. Nothing typed into payment fields, nothing added to a cart or submitted on the live store, each page loaded once, emails only to inboxes the business controls, no load, and no script removed; removing one is a change for the owner to make. If a response shows a real person's details, stop and record it.PCI SSC post of 10 March 2025; OWASP Third Party JavaScript Management Cheat Sheet; MDN Subresource Integrity.
S2. Payment page scripts: headers, change alerts and who can change the pageThe payment page's response headers, and the accounts that can change what it loads.Named people who maintain the theme, apps and tags, and one named person who is told of a change.Read only. From S1's staging page loads, record the security-related response headers by name, such as Content-Security-Policy or Content-Security-Policy-Report-Only. List the accounts that can change what the payment page loads: theme editors, app installers and tag manager users. Then find out which named person hears about it, and through what route, when a script or a header on the live payment page changes.A named person receives change alerts, and the list of accounts that can change the page is up to date. The OWASP Content Security Policy Cheat Sheet says the report-only header delivers a policy that is not enforced, while violation reports are still printed to the console, and delivered to a violation endpoint if the report-to and report-uri directives are used. Record which of the two headers the page sends.Header names and values marked as read on the copy, the alert route, and the account list. May relate to: the post's statement that the requirements also address the need to manage both scripts and security-impacting HTTP headers carefully, to prevent unauthorised changes to webpages. Whether that applies is settled with the acquirer.Read only. No header, tag or account is changed during the review.PCI SSC post of 10 March 2025; OWASP Content Security Policy Cheat Sheet; OWASP Third Party JavaScript Management Cheat Sheet, on tag managers.
R1. Refunds: who can send money backThe order value, its payment reference and the customer's contact details.The staff roles the business names for refunds. Not customers, and not other staff roles.First, read only: list the roles that hold refund permission on the live store. Then on the copy, have a staff test account allowed to refund make a small partial refund on P3's paid order, record that request, and send it again from a staff test account without the permission, from cust-b and with no session. Note whether refunds above an amount the business sets need a second person.Only the named roles succeed, and the provider's test dashboard shows no refund from any other request. WSTG v4.2's role definitions test says sensitive admin functionality should leverage a maker-checker principle, or use MFA.The role list, each request and response, and the provider's test dashboard.Staging copy, payment provider in test mode, P3's paid test order only, staff test accounts and cust-b only, each request once, refund emails only to inboxes the business controls, no load. If a response shows a real person's details, stop and record it.WSTG v4.2 WSTG-ATHZ-02 and WSTG-IDNT-01; ASVS 5.0 v5.0.0-8.2.1.
R2. Refunds: how much, and on which orderAmounts captured and refunded, and each order's state.Refunds that add up to no more than was paid, on orders that were paid.On P3's paid test order, refund more than is left to refund, then refund in two parts that together exceed the amount captured. Try to refund P3's unpaid order. Once the paid order is fully refunded, cancel it and check whether the cancellation sends money back again.The total refunded never exceeds the amount captured, the unpaid order cannot be refunded, and the cancellation moves no money. The business logic guide writes the first rule as an invariant and plays a cancellation after a full refund as one of its abuse cases.The provider's test dashboard, and the order's records before and after each step.Staging copy, payment provider in test mode, P3's two test orders only, staff test accounts only, each attempt once, refund emails only to inboxes the business controls, no load. If a response shows a real person's details, stop and record it.ASVS 5.0 v5.0.0-2.3.1; the business logic vulnerabilities guide.
E1. Exports: order and customer filesEvery customer's names, addresses, email addresses, phone numbers and order contents, in one file.The staff roles the business names for exports. Not customers, and not anyone without a session.First, read only: list the roles that can export orders and customers, where the files are saved and shared, and how long they are kept. Then on the copy, send the export request from a staff test account without export permission and from cust-b. Have a staff test account with export permission generate one export holding only the run's test records, list its columns, and request that file's address with no session and after signing out.Other roles refused, the file address refused without a session, and the columns limited to what the business uses the file for.The role list, the file location and retention, the column names only, and each request and response, with the file's address redacted rather than copied. May relate to: APP 11.1 and APP 11.2 of the Privacy Act 1988, where the Act applies to the store.Staging copy, staff test accounts and cust-b only, each request once, no load. If the copy holds real customers' records, do not run an export there; record the roles and settings only. Never open a live export, and record column names, never contents. If a response or file shows a real person's details, stop, record it and tell the owner.WSTG v4.2 WSTG-ATHZ-02; ASVS 5.0 v5.0.0-8.2.1 and v5.0.0-8.2.3. The sensitive data exposure guide covers what an export should hold.
F1. Fulfilment and connected apps: what leaves the storeNames, addresses, phone numbers, items and order values sent to fulfilment, shipping, accounting, loyalty, review and gift card apps.Each receiving system the business names, holding only the fields it needs.Read only. List each outbound connection and webhook: the receiving system, its owner, the fields sent, whether deliveries are signed, where deliveries are logged, and, from each app's own documentation and agreement, what it keeps and for how long. Ask the developer who set each connection up whether it is signed, rather than opening a form that shows its secret. Confirm from B2 that each one on the copy points at a test account.Each receiver has an owner and a reason, its fields match what it needs, and deliveries are signed wherever the platform offers signing.The dated list, with each secret recorded only by the name of the setting that holds it. May relate to: APP 11.1, where the Act applies to the store.Read only. No delivery is sent and no app's own platform is tested; testing an app's own system waits for its vendor's answer in B1.This page's own list. The WooCommerce and Shopify guides record their platforms' webhook settings.
F2. Fulfilment: status updates coming backShipped, delivered and cancelled statuses, tracking numbers and stock movements.The fulfilment provider, for orders it holds. Nobody else.To the staging copy's inbound status address only, send one update with no signature, and one for P3's paid test order with the order number altered to A1's order. Have the fulfilment provider's test account send one genuine update, and resend it once.The unsigned and altered updates refused, and the resent update causing no second shipped email, stock movement or loyalty credit.Each delivery, the store's response and the order notes afterwards.The staging copy's address only, the fulfilment provider's test account within the terms recorded in B1, the run's test orders only, each delivery once, emails only to inboxes the business controls, no load. If a response shows a real person's details, stop and record it.This page's own checks, modelled on the signature and duplicate checks the Stripe webhook security guide sets out for payment events.
L1. Loyalty: points on orders that are refunded or cancelledPoints balances, and the order history the loyalty app receives.Points follow what the customer finally paid for.As cust-a, place one paid test order that earns points. Refund one item and read the balance, then refund or cancel the rest and read it again.Points taken back in line with the store's rule at each step. WSTG v4.2's work-flow test says, for a customer who cancels or removes items after points are added, that points should not be applied until the transaction is tendered, or should be rolled back if the increment does not match the final tender. This row applies the same rule to refunds after the sale.The balance after each step, and the order notes.Staging copy, payment provider in test mode, the loyalty app in its test or sandbox mode within the terms recorded in B1, cust-a only, never a real card, one checkout submission, each step once, emails only to inboxes the business controls, no load. If a response shows a real person's details, stop and record it.WSTG v4.2 WSTG-BUSL-06.
L2. Loyalty and balances: spending a balance or reward twiceGift card, store credit or points balances, and one-time referral or reward codes.One spend per unit of balance, and one use per single-use code.Give cust-a a test gift card or points balance that covers one order. Prepare three checkouts that each spend the whole balance, and submit them at nearly the same moment, once. Separately, redeem a single-use test reward code in one checkout, then try to apply it again to a new cart.The balance never below zero, one checkout succeeding, and the code working once. WSTG-BUSL-10 lists negative balances after concurrent debits among the indicators of vulnerability.The responses, the orders created, and the balance afterwards.Staging copy, payment provider in test mode, test balances and codes only, cust-a only, never a real card, three checkout submissions in the burst plus one for the code, each sent once, emails only to inboxes the business controls, no load. If a response shows a real person's details, stop and record it.WSTG-BUSL-10, Race Conditions; the race conditions guide.
Z1. Significant change: when to run rows againWhatever the change touches.The owner decides; the developer or agency names each change before it ships.Planning. After each change below, rerun its rows on the staging copy before the change reaches the live store where you can, or as soon after as the release allows. A new payment provider or integration method: P1 to P4, S1, S2 and C1. A change to cart, checkout or account templates: A1 to A3, C1 to C4, S1 and S2. An app or plugin added, removed or updated that touches carts, coupons, orders, customers or checkout: the rows for what it touches, plus S1. A new script, tag or tag manager change on payment pages: S1 and S2. Changed coupon, refund, loyalty or gift card rules: D1, D2, R1, R2, L1 and L2. A new fulfilment, accounting or marketing connection: F1, F2 and E1. A new platform version or theme, or a change of agency, developer or host: every row.Each change in the store's change log has a dated rerun of its rows, with results.The change, the rows rerun, the date and the build. May relate to: any testing or scanning schedule in the store's PCI DSS validation, which this page does not set; ask the acquirer or payment brand.Planning only. Each rerun row keeps its own limits.This page's own list.
Z2. Clean-upEverything the run created.Each system's owner removes records in that system.List every test order, cart, customer, coupon, product, balance, saved address, export file and refund the run created, each carrying the marker or belonging to a test identity, and each connected test account that received any of them: the payment provider, fulfilment, loyalty and email. Remove them on the copy, and ask each system's owner to remove them there.Nothing the run created remains, or each item that does is listed with who removes it and when.The clean-up list with dates. WSTG-BUSL-10's logistics notes include keeping a record of any orders made so that they can be cancelled and refunded.Deletes only what the run created: records carrying the run's marker or belonging to its test identities.WSTG-BUSL-10, Logistics.
Z3. Evidence pack, and what it may relate toThe run's records, redacted.The owner holds the pack and decides who sees it.Planning. Keep together: B1's dated terms and answers, B2's readings, each row's redacted requests and responses with the build and date, the lists from S1, S2, E1 and F1, Z1's change log and Z2's clean-up list.Each item dated, with who holds it and who may see it.May relate to: PCI DSS validation, which the PCI SSC's Merchant Resources FAQ sends to the acquirer or payment brand; PCI DSS v4.x Requirements 6.4.3 and 11.6.1 as the PCI SSC post describes them (S1, S2); APP 11.1 to 11.3 of the Privacy Act 1988, where the Act applies to the store; the Notifiable Data Breaches scheme, as the OAIC explains it, if a breach occurs; and customer questionnaires and insurer forms, covered by the security questionnaires and cyber insurance checklist guides. Applicability is for the owner to establish with the acquirer and their own adviser.No password, key, secret, card number or real customer record goes in the pack.PCI SSC Merchant Resources FAQ; PCI SSC post of 10 March 2025; Privacy Act 1988, Compilation No. 104; OAIC, About the Notifiable Data Breaches scheme.

Row and step

Data involved

  1. B1. Before testing: terms and validation scope
    Data involved
    None. This row reads documents and asks questions.
    Who should reach it, and who should not
    The store owner decides what is tested. The platform, the host, each payment provider and each app vendor set terms for their own systems. The acquirer or payment brand is who the PCI SSC sends merchants to about compliance validation.
    Test case
    Read only. List every system the later rows touch: the platform, the host, each payment provider, and each fulfilment, accounting, loyalty, review or gift card app. For each, read its own terms on security testing and on test or sandbox use, note the date, and ask the vendor in writing where the terms leave doubt. Ask the acquirer or payment provider in writing how the store's PCI DSS validation is set, including whether a self-assessment questionnaire is used and which one. For your platform, read the platform guide's section on its terms first.
    Expected result
    A dated answer for every system before any row that sends a request. A system whose terms leave doubt stays out of the run until its vendor answers.
    Evidence to keep, and what it may relate to
    Each terms page by address and date read, each written answer, and the acquirer's or provider's answer on validation. May relate to: PCI DSS validation, which the PCI SSC's Merchant Resources FAQ sends to the acquirer (merchant bank) or payment brand.
    Limits of this row
    Read only. No request goes to the store or to any vendor's system.
    References
    PCI SSC Merchant Resources FAQ; PCI SSC post of 10 March 2025. No vendor's permission is stated here: each comes from that vendor.
  2. B2. Before testing: the staging copy, test modes and identities
    Data involved
    Whatever the copy holds. A copy cloned from the live store can hold real customers' records and live connections.
    Who should reach it, and who should not
    The person who manages the copy changes its settings. This row only reads them, apart from creating the run's test identities.
    Test case
    Read only, then create identities. Confirm the copy belongs to the business and is not the live store. Read each payment provider's mode on the copy and on the live store, and each connection on the copy for fulfilment, shipping, accounting, email, SMS, loyalty, reviews and marketing. Then create invented customers cust-a and cust-b, each with an inbox and a phone number the business controls, a staff test account for each role the rows need, and the marker ORDERTEST-1115, or your own, in every name.
    Expected result
    Every provider on the copy in its documented test or sandbox mode, every connection off or pointing at a test account, and the live store in live mode. If any reading says otherwise, the run stops there; switching a mode or a connection belongs to the person who manages the copy, and this matrix does not do it.
    Evidence to keep, and what it may relate to
    Each mode and connection as read, with the date and the reader. WSTG-BUSL-10 says test card details should only be usable on development or sandbox versions of payment gateways, but may be accepted on live sites if they have been misconfigured. The live mode is recorded for that reason.
    Limits of this row
    Read only, apart from creating the test identities on the copy. Where a mode can be read only on a screen that also shows keys, ask whoever manages the integration instead. If the copy holds real customers' records, never open them; rows that list or export records then stop at their read-only part.
    References
    WSTG-BUSL-10, Test Payment Cards.
  3. A1. Customer accounts: orders, invoices, returns and tracking
    Data involved
    Names, email addresses, phone numbers, delivery and billing addresses, items bought, order totals, the payment summary the store shows, invoice and receipt files, and return or refund requests.
    Who should reach it, and who should not
    The customer who placed the order, and staff in roles that handle orders. Not another customer, and not a visitor with no session.
    Test case
    As cust-a, place one paid test order and note each address that names it: the order detail, the invoice or receipt file, a return or refund request, and the tracking page. As cust-b, request each of those addresses with cust-a's order id or number in place of cust-b's own, then send each request again with no session and after signing out.
    Expected result
    Every request refused with the same response the store gives for an order that does not exist, and no response body holding any of cust-a's details.
    Evidence to keep, and what it may relate to
    Each request and response, marked with the account and session used, with personal details and any order or tracking token redacted. A link that opens an order without a sign-in is recorded by the page it opens, never copied. May relate to: the steps APP 11.1 of the Privacy Act 1988 concerns, where the Act applies to the store.
    Limits of this row
    Staging copy of your own store, payment provider in test mode, cust-a and cust-b only, one checkout submission, each request once, emails only to inboxes the business controls, no load, never a real card. If any response shows a real person's details, stop, record the request and tell the owner.
    References
    WSTG v4.2 WSTG-ATHZ-04 and WSTG-ATHZ-02; ASVS 5.0 v5.0.0-8.2.2.
  4. A2. Customer accounts: guest order lookup and emailed order links
    Data involved
    The same details as A1, for customers who bought without an account.
    Who should reach it, and who should not
    The person holding both the order number and the email address used for it, or the link sent to that address. Not someone holding only one of them.
    Test case
    Where the store offers a guest order lookup or emails an order link: as a guest, place one test order with cust-a's details. Look it up with cust-b's email address. Open the emailed link, then change the order number or id in it to that of A1's order. Only where B2 found no real customers' records on the copy, try the order numbers either side of the guest order in the lookup, once each.
    Expected result
    The lookup answers only when the number and the email address belong together, the link shows only its own order, and the neighbouring numbers are refused.
    Evidence to keep, and what it may relate to
    Each lookup request and response, redacted. The emailed link is never copied into the evidence: record only whether it carries the order number in plain view and whether the changed link was refused. May relate to: APP 11.1, where the Act applies to the store.
    Limits of this row
    Staging copy, payment provider in test mode, test orders from cust-a and cust-b only, one checkout submission, at most five lookups and link requests in all, the emailed link opened only for the run's test orders, emails only to inboxes the business controls, no load, never a real card. If a real order appears, stop and record it.
    References
    WSTG v4.2 WSTG-ATHZ-04. The lookup and link checks are this page's own.
  5. A3. Customer accounts: saved addresses and saved payment methods
    Data involved
    Saved delivery and billing addresses, phone numbers, and whatever reference to a saved payment method the store keeps.
    Who should reach it, and who should not
    The account holder. Not another customer.
    Test case
    Give cust-a a saved address and, using the provider's documented test card, a saved payment method. As cust-b, read, edit and then delete cust-a's saved address by its id. At checkout as cust-b, submit cust-a's saved payment method reference in place of cust-b's.
    Expected result
    Each request refused, cust-a reads the address back unchanged, and the checkout refuses a payment method that belongs to another customer.
    Evidence to keep, and what it may relate to
    Requests and responses, redacted, and cust-a's read-back. Record the payment method reference as present or absent, never its value. May relate to: APP 11.1, where the Act applies to the store.
    Limits of this row
    Staging copy, payment provider in test mode with its documented test card, never a real card, cust-a and cust-b only, each attempt once, at most one checkout submission, the delete only on the address created for this row, emails only to inboxes the business controls, no load. Stop and record if a real person's details appear.
    References
    WSTG v4.2 WSTG-ATHZ-04; ASVS 5.0 v5.0.0-8.2.2.
  6. C1. Carts and checkout: prices and totals in the request
    Data involved
    Product ids, quantities, unit prices, shipping costs, discounts and order totals.
    Who should reach it, and who should not
    The server sets every price and total. The customer chooses products and quantities only.
    Test case
    As cust-a, add one product to the cart and start checkout with the browser's developer tools open. If any add-to-cart or checkout request carries a price, a line total, a shipping cost or an order total, including in a hidden field, change one value at a time, submit, and complete the test payment if the store accepts it.
    Expected result
    The changed value is ignored or refused, and the order total equals the catalogue price plus the shipping the server calculates. WSTG-BUSL-10's remediation says the only things included in client-side requests should be item IDs and quantities.
    Evidence to keep, and what it may relate to
    The original and changed requests, the order total that resulted, and the catalogue price on the day.
    Limits of this row
    Staging copy, payment provider in test mode, cust-a only, never a real card, at most two checkout submissions, one value changed per request, emails only to inboxes the business controls, no load. If a response shows a real person's details, stop and record it.
    References
    WSTG-BUSL-10, Price Tampering on the Application; WSTG v4.2 WSTG-BUSL-03; ASVS 5.0 v5.0.0-2.2.2.
  7. C2. Carts and checkout: quantities
    Data involved
    Line quantities and the totals built from them.
    Who should reach it, and who should not
    The server decides which quantities are valid, including any pack sizes the store sells in.
    Test case
    In the cart requests, set a quantity of 0.1, then 0, then -1, then a value outside the store's own pack sizes or drop-down choices. Remove one item repeatedly until its quantity would fall below zero. Send one request with the quantity parameter twice, such as quantity=5&quantity=4.
    Expected result
    Each value refused or corrected to a valid whole number, and the cart total never below the price of what is left in the cart.
    Evidence to keep, and what it may relate to
    Each request, the cart as the server returns it, and the total.
    Limits of this row
    Staging copy, cust-a only, cart requests only. If a value is accepted, at most one checkout submission, with the payment provider in test mode, to see whether it can be paid. Never a real card, emails only to inboxes the business controls, no load. If a response shows a real person's details, stop and record it.
    References
    WSTG-BUSL-10, Quantity Tampering; WSTG v4.2 WSTG-BUSL-01; ASVS 5.0 v5.0.0-2.2.1.
  8. C3. Carts and checkout: currency, shipping and minimum basket
    Data involved
    Currency code, delivery address, shipping cost and minimum order value.
    Who should reach it, and who should not
    The server applies the store's rules at the final step, whatever the cart said earlier.
    Test case
    Where the store sells in more than one currency, change the currency code in the checkout request and keep the amount. Change the delivery address after shipping has been priced. Remove items after a shipping rate or a minimum basket value has been applied.
    Expected result
    The total worked out again from the server's own rules at the last step, in the currency the server set, with the shipping for the final address and contents.
    Evidence to keep, and what it may relate to
    Each request, and the order total that resulted.
    Limits of this row
    Staging copy, payment provider in test mode, cust-a only, never a real card, one checkout submission, emails only to inboxes the business controls, no load. If a response shows a real person's details, stop and record it.
    References
    WSTG-BUSL-10, Currency Tampering and Breaking Payment Flows.
  9. C4. Carts and checkout: changes after payment starts
    Data involved
    Cart contents, the amount sent for payment, discounts and the paid order.
    Who should reach it, and who should not
    The order holds what was paid for, and nothing else.
    Test case
    As cust-a, start payment for a cart, then add an item from a second browser tab before the provider confirms. Change the cart after a coupon has been applied. After checkout completes, send a request that changes the order's contents.
    Expected result
    The order holds only what was paid for, any added item stays unpaid or the payment is refused, the discount is worked out again for the changed cart, and the customer cannot change a completed order's contents.
    Evidence to keep, and what it may relate to
    Requests and responses, and the order as staff see it afterwards.
    Limits of this row
    Staging copy, payment provider in test mode, cust-a only, never a real card, one checkout submission, emails only to inboxes the business controls, no load. If a response shows a real person's details, stop and record it.
    References
    WSTG-BUSL-10, Adding Items After Payment Initiation and Breaking Payment Flows; ASVS 5.0 v5.0.0-2.3.1.
  10. D1. Coupons: the rules for each code
    Data involved
    Coupon codes, discount values, usage limits, allowed customers and products, and expiry dates.
    Who should reach it, and who should not
    Eligible customers apply codes. Staff in the roles the store names create them. Nobody else should be able to find one.
    Test case
    On the live store, read the coupon list without saving and note any codes that follow a pattern a customer could guess, such as a word plus a number. Search the copy's checkout page source for any code. Then, on the copy, create a test coupon limited to one use per customer and set not to combine, and as cust-a apply it twice, apply it together with a second test coupon, apply it to a product it excludes, and apply it after its expiry date.
    Expected result
    Each attempt refused with the store's message, no code in the page source or a hidden field, and no live code that follows a guessable pattern.
    Evidence to keep, and what it may relate to
    The coupon settings as read, with each live code recorded by its pattern rather than in full, each message word for word, and the order total.
    Limits of this row
    The list and page source are read only. Requests only on the staging copy, with test coupons made for this row; no code is tried against the live store. cust-a only, at most one checkout submission, payment provider in test mode, never a real card, emails only to inboxes the business controls, no load. If a response shows a real person's details, stop and record it.
    References
    WSTG-BUSL-10, Discount Codes; WSTG v4.2 WSTG-BUSL-05.
  11. D2. Coupons and stock: a one-time code and the last unit at the same moment
    Data involved
    A coupon's usage count and a product's stock level.
    Who should reach it, and who should not
    One redemption per single-use code, and no more units sold than exist.
    Test case
    Create a single-use test coupon and a test product with one unit in stock. As cust-a, prepare three checkouts of that product with the coupon applied, and submit them at nearly the same moment, once.
    Expected result
    One order succeeds, the coupon's usage count reads one, and stock reads zero, never below. WSTG-BUSL-10 lists multiple success responses for a single-use action and resources allocated beyond stated capacity among the indicators of vulnerability.
    Evidence to keep, and what it may relate to
    All three responses, the orders created, and the usage count and stock level afterwards.
    Limits of this row
    Staging copy, payment provider in test mode, the test coupon and test product only, cust-a only, never a real card, three checkout submissions sent once, with no repeats, emails only to inboxes the business controls. This is a timing check, not load testing. If a response shows a real person's details, stop and record it.
    References
    WSTG-BUSL-10, Race Conditions; ASVS 5.0 v5.0.0-2.3.4. The race conditions guide covers the fixes.
  12. P1. Payment handoff: what the store sends to the provider
    Data involved
    Order id, amount, currency and sometimes item lines, with card details typed at the provider's page or fields.
    Who should reach it, and who should not
    The store sets the amount. The customer types card details only into the provider's page or fields.
    Test case
    Note which of the methods WSTG-BUSL-10 lists the checkout uses: a redirect to the provider, the provider loaded in an iframe, an HTML form posting to the provider, or card details taken by the store and posted to the provider's API. Where an amount, currency or item total travels in a form field or an iframe address, change it once and complete the test payment.
    Expected result
    The provider charges the amount the store set, or the store does not mark the order paid because the amount paid differs from the order total.
    Evidence to keep, and what it may relate to
    The method, the field that carried the amount, the amount the provider's test dashboard shows, and the order's status. Keep the method with B1's questions for the acquirer; this page draws no conclusion from it.
    Limits of this row
    This row sends a request to the payment provider's test environment, so it runs only once B1 holds the provider's terms. Staging copy, payment provider in test mode with its documented test card, never a real card, cust-a only, one checkout submission, emails only to inboxes the business controls, no load. If a response shows a real person's details, stop and record it. WSTG-BUSL-10 notes that payment gateways are usually run by third parties and may not be included in the scope of testing, and that while price tampering may be acceptable, other types of attacks, such as SQL injection, should not be performed without explicit written approval.
    References
    WSTG-BUSL-10, Payment Gateway Integration Methods and Price Tampering on the Payment Gateway.
  13. P2. Payment handoff: signing or encryption settings
    Data involved
    The integration's settings, and the names of the keys or secrets it uses.
    Who should reach it, and who should not
    The person who manages the payment integration.
    Test case
    Read only. From the provider's own documentation, list each option it offers to sign, hash or encrypt the transaction details. Ask the person who manages the integration whether each one is switched on, and the name of the setting or variable that holds each secret, rather than opening a settings screen that displays keys.
    Expected result
    Each option the provider documents is on, or the reason it is off is written down with who decided. WSTG-BUSL-10's remediation says to review payment gateway documentation and use all available security features, such as encryption and secure hashes.
    Evidence to keep, and what it may relate to
    The option names, on or off, the date and the reader.
    Limits of this row
    Read only. Never open, reveal or copy a key or secret.
    References
    WSTG-BUSL-10, Remediation.
  14. P3. Payment confirmation: the return page
    Data involved
    The order's paid status, its fulfilment and its confirmation emails.
    Who should reach it, and who should not
    Only the provider's confirmation should mark an order paid. Not the customer's browser.
    Test case
    As cust-a, place one test order and leave it unpaid at the provider, then request the store's success or return address directly. For a second test order that was paid, load the success address a second time.
    Expected result
    The unpaid order stays unpaid and is not sent for fulfilment, and the second load creates no second fulfilment, email or loyalty credit.
    Evidence to keep, and what it may relate to
    The requests, the order statuses and notes, and each email that arrived at the controlled inbox. Keep both orders for P4, R1 and R2.
    Limits of this row
    Staging copy, payment provider in test mode, cust-a only, never a real card, two checkout submissions, each address loaded at most twice, emails only to inboxes the business controls, no load. If a response shows a real person's details, stop and record it.
    References
    WSTG-BUSL-10, Breaking Payment Flows; ASVS 5.0 v5.0.0-2.3.1.
  15. P4. Payment confirmation: notifications from the provider
    Data involved
    Payment events carrying an order reference, an amount and a status.
    Who should reach it, and who should not
    The provider sends them. Nobody else should be able to move an order with one.
    Test case
    To the staging copy's notification address only, send one notification with no signature, and one copy of a genuine notification with its body altered. Then resend P3's genuine paid notification once, from the provider's test tools where it offers that.
    Expected result
    The unsigned and altered notifications refused, and the resent one answered without a second paid status, fulfilment, email or loyalty credit. WSTG-BUSL-10 says that if the back end lacks proper idempotency checks, a gateway callback replayed rapidly in parallel can, among other things, trigger multiple order fulfilment events or mark the same order as paid multiple times.
    Evidence to keep, and what it may relate to
    Each delivery and the store's response, and the order notes afterwards.
    Limits of this row
    The staging copy's address only, never the live store's. Payment provider in test mode, P3's test orders only, each delivery once, emails only to inboxes the business controls, no load. Resending from the provider's tools follows its terms, recorded in B1. If a response shows a real person's details, stop and record it.
    References
    WSTG-BUSL-10, Race Conditions. The Stripe webhook security guide covers signatures, timestamps and duplicate events in depth.
  16. S1. Payment page scripts: what loads, and who approved it
    Data involved
    Card and personal details typed into the checkout, and anything else a script on that page can read.
    Who should reach it, and who should not
    Scripts the business has a reason for and has approved. Nothing else.
    Test case
    Read only, apart from one test cart on the copy. On the staging copy, as cust-a, put one product in a cart and open the checkout and payment pages a single time each while the developer tools record network requests, leaving the payment fields empty. On the live store, open one product page and the empty cart page a single time each, without adding anything. For each script host, record its business reason, who approved it, and how its integrity is checked, such as Subresource Integrity, which MDN says works by providing a cryptographic hash that a fetched resource must match. Note any script host that the live pages load and the staging checkout does not, or the other way round.
    Expected result
    Every host has an owner, a reason, an approval and an integrity method, or a written reason for none. A host nobody can explain goes to the owner as a question.
    Evidence to keep, and what it may relate to
    The dated script list. May relate to: PCI DSS v4.x Requirements 6.4.3 and 11.6.1, which the PCI SSC post describes as focused on payment page scripts being properly authorised, checked for integrity and monitored for tampering. Whether they apply is settled with the acquirer, as B1 records.
    Limits of this row
    Page loads, plus one test cart for cust-a on the staging copy, which goes on Z2's list. Nothing typed into payment fields, nothing added to a cart or submitted on the live store, each page loaded once, emails only to inboxes the business controls, no load, and no script removed; removing one is a change for the owner to make. If a response shows a real person's details, stop and record it.
    References
    PCI SSC post of 10 March 2025; OWASP Third Party JavaScript Management Cheat Sheet; MDN Subresource Integrity.
  17. S2. Payment page scripts: headers, change alerts and who can change the page
    Data involved
    The payment page's response headers, and the accounts that can change what it loads.
    Who should reach it, and who should not
    Named people who maintain the theme, apps and tags, and one named person who is told of a change.
    Test case
    Read only. From S1's staging page loads, record the security-related response headers by name, such as Content-Security-Policy or Content-Security-Policy-Report-Only. List the accounts that can change what the payment page loads: theme editors, app installers and tag manager users. Then find out which named person hears about it, and through what route, when a script or a header on the live payment page changes.
    Expected result
    A named person receives change alerts, and the list of accounts that can change the page is up to date. The OWASP Content Security Policy Cheat Sheet says the report-only header delivers a policy that is not enforced, while violation reports are still printed to the console, and delivered to a violation endpoint if the report-to and report-uri directives are used. Record which of the two headers the page sends.
    Evidence to keep, and what it may relate to
    Header names and values marked as read on the copy, the alert route, and the account list. May relate to: the post's statement that the requirements also address the need to manage both scripts and security-impacting HTTP headers carefully, to prevent unauthorised changes to webpages. Whether that applies is settled with the acquirer.
    Limits of this row
    Read only. No header, tag or account is changed during the review.
    References
    PCI SSC post of 10 March 2025; OWASP Content Security Policy Cheat Sheet; OWASP Third Party JavaScript Management Cheat Sheet, on tag managers.
  18. R1. Refunds: who can send money back
    Data involved
    The order value, its payment reference and the customer's contact details.
    Who should reach it, and who should not
    The staff roles the business names for refunds. Not customers, and not other staff roles.
    Test case
    First, read only: list the roles that hold refund permission on the live store. Then on the copy, have a staff test account allowed to refund make a small partial refund on P3's paid order, record that request, and send it again from a staff test account without the permission, from cust-b and with no session. Note whether refunds above an amount the business sets need a second person.
    Expected result
    Only the named roles succeed, and the provider's test dashboard shows no refund from any other request. WSTG v4.2's role definitions test says sensitive admin functionality should leverage a maker-checker principle, or use MFA.
    Evidence to keep, and what it may relate to
    The role list, each request and response, and the provider's test dashboard.
    Limits of this row
    Staging copy, payment provider in test mode, P3's paid test order only, staff test accounts and cust-b only, each request once, refund emails only to inboxes the business controls, no load. If a response shows a real person's details, stop and record it.
    References
    WSTG v4.2 WSTG-ATHZ-02 and WSTG-IDNT-01; ASVS 5.0 v5.0.0-8.2.1.
  19. R2. Refunds: how much, and on which order
    Data involved
    Amounts captured and refunded, and each order's state.
    Who should reach it, and who should not
    Refunds that add up to no more than was paid, on orders that were paid.
    Test case
    On P3's paid test order, refund more than is left to refund, then refund in two parts that together exceed the amount captured. Try to refund P3's unpaid order. Once the paid order is fully refunded, cancel it and check whether the cancellation sends money back again.
    Expected result
    The total refunded never exceeds the amount captured, the unpaid order cannot be refunded, and the cancellation moves no money. The business logic guide writes the first rule as an invariant and plays a cancellation after a full refund as one of its abuse cases.
    Evidence to keep, and what it may relate to
    The provider's test dashboard, and the order's records before and after each step.
    Limits of this row
    Staging copy, payment provider in test mode, P3's two test orders only, staff test accounts only, each attempt once, refund emails only to inboxes the business controls, no load. If a response shows a real person's details, stop and record it.
    References
    ASVS 5.0 v5.0.0-2.3.1; the business logic vulnerabilities guide.
  20. E1. Exports: order and customer files
    Data involved
    Every customer's names, addresses, email addresses, phone numbers and order contents, in one file.
    Who should reach it, and who should not
    The staff roles the business names for exports. Not customers, and not anyone without a session.
    Test case
    First, read only: list the roles that can export orders and customers, where the files are saved and shared, and how long they are kept. Then on the copy, send the export request from a staff test account without export permission and from cust-b. Have a staff test account with export permission generate one export holding only the run's test records, list its columns, and request that file's address with no session and after signing out.
    Expected result
    Other roles refused, the file address refused without a session, and the columns limited to what the business uses the file for.
    Evidence to keep, and what it may relate to
    The role list, the file location and retention, the column names only, and each request and response, with the file's address redacted rather than copied. May relate to: APP 11.1 and APP 11.2 of the Privacy Act 1988, where the Act applies to the store.
    Limits of this row
    Staging copy, staff test accounts and cust-b only, each request once, no load. If the copy holds real customers' records, do not run an export there; record the roles and settings only. Never open a live export, and record column names, never contents. If a response or file shows a real person's details, stop, record it and tell the owner.
    References
    WSTG v4.2 WSTG-ATHZ-02; ASVS 5.0 v5.0.0-8.2.1 and v5.0.0-8.2.3. The sensitive data exposure guide covers what an export should hold.
  21. F1. Fulfilment and connected apps: what leaves the store
    Data involved
    Names, addresses, phone numbers, items and order values sent to fulfilment, shipping, accounting, loyalty, review and gift card apps.
    Who should reach it, and who should not
    Each receiving system the business names, holding only the fields it needs.
    Test case
    Read only. List each outbound connection and webhook: the receiving system, its owner, the fields sent, whether deliveries are signed, where deliveries are logged, and, from each app's own documentation and agreement, what it keeps and for how long. Ask the developer who set each connection up whether it is signed, rather than opening a form that shows its secret. Confirm from B2 that each one on the copy points at a test account.
    Expected result
    Each receiver has an owner and a reason, its fields match what it needs, and deliveries are signed wherever the platform offers signing.
    Evidence to keep, and what it may relate to
    The dated list, with each secret recorded only by the name of the setting that holds it. May relate to: APP 11.1, where the Act applies to the store.
    Limits of this row
    Read only. No delivery is sent and no app's own platform is tested; testing an app's own system waits for its vendor's answer in B1.
    References
    This page's own list. The WooCommerce and Shopify guides record their platforms' webhook settings.
  22. F2. Fulfilment: status updates coming back
    Data involved
    Shipped, delivered and cancelled statuses, tracking numbers and stock movements.
    Who should reach it, and who should not
    The fulfilment provider, for orders it holds. Nobody else.
    Test case
    To the staging copy's inbound status address only, send one update with no signature, and one for P3's paid test order with the order number altered to A1's order. Have the fulfilment provider's test account send one genuine update, and resend it once.
    Expected result
    The unsigned and altered updates refused, and the resent update causing no second shipped email, stock movement or loyalty credit.
    Evidence to keep, and what it may relate to
    Each delivery, the store's response and the order notes afterwards.
    Limits of this row
    The staging copy's address only, the fulfilment provider's test account within the terms recorded in B1, the run's test orders only, each delivery once, emails only to inboxes the business controls, no load. If a response shows a real person's details, stop and record it.
    References
    This page's own checks, modelled on the signature and duplicate checks the Stripe webhook security guide sets out for payment events.
  23. L1. Loyalty: points on orders that are refunded or cancelled
    Data involved
    Points balances, and the order history the loyalty app receives.
    Who should reach it, and who should not
    Points follow what the customer finally paid for.
    Test case
    As cust-a, place one paid test order that earns points. Refund one item and read the balance, then refund or cancel the rest and read it again.
    Expected result
    Points taken back in line with the store's rule at each step. WSTG v4.2's work-flow test says, for a customer who cancels or removes items after points are added, that points should not be applied until the transaction is tendered, or should be rolled back if the increment does not match the final tender. This row applies the same rule to refunds after the sale.
    Evidence to keep, and what it may relate to
    The balance after each step, and the order notes.
    Limits of this row
    Staging copy, payment provider in test mode, the loyalty app in its test or sandbox mode within the terms recorded in B1, cust-a only, never a real card, one checkout submission, each step once, emails only to inboxes the business controls, no load. If a response shows a real person's details, stop and record it.
    References
    WSTG v4.2 WSTG-BUSL-06.
  24. L2. Loyalty and balances: spending a balance or reward twice
    Data involved
    Gift card, store credit or points balances, and one-time referral or reward codes.
    Who should reach it, and who should not
    One spend per unit of balance, and one use per single-use code.
    Test case
    Give cust-a a test gift card or points balance that covers one order. Prepare three checkouts that each spend the whole balance, and submit them at nearly the same moment, once. Separately, redeem a single-use test reward code in one checkout, then try to apply it again to a new cart.
    Expected result
    The balance never below zero, one checkout succeeding, and the code working once. WSTG-BUSL-10 lists negative balances after concurrent debits among the indicators of vulnerability.
    Evidence to keep, and what it may relate to
    The responses, the orders created, and the balance afterwards.
    Limits of this row
    Staging copy, payment provider in test mode, test balances and codes only, cust-a only, never a real card, three checkout submissions in the burst plus one for the code, each sent once, emails only to inboxes the business controls, no load. If a response shows a real person's details, stop and record it.
    References
    WSTG-BUSL-10, Race Conditions; the race conditions guide.
  25. Z1. Significant change: when to run rows again
    Data involved
    Whatever the change touches.
    Who should reach it, and who should not
    The owner decides; the developer or agency names each change before it ships.
    Test case
    Planning. After each change below, rerun its rows on the staging copy before the change reaches the live store where you can, or as soon after as the release allows. A new payment provider or integration method: P1 to P4, S1, S2 and C1. A change to cart, checkout or account templates: A1 to A3, C1 to C4, S1 and S2. An app or plugin added, removed or updated that touches carts, coupons, orders, customers or checkout: the rows for what it touches, plus S1. A new script, tag or tag manager change on payment pages: S1 and S2. Changed coupon, refund, loyalty or gift card rules: D1, D2, R1, R2, L1 and L2. A new fulfilment, accounting or marketing connection: F1, F2 and E1. A new platform version or theme, or a change of agency, developer or host: every row.
    Expected result
    Each change in the store's change log has a dated rerun of its rows, with results.
    Evidence to keep, and what it may relate to
    The change, the rows rerun, the date and the build. May relate to: any testing or scanning schedule in the store's PCI DSS validation, which this page does not set; ask the acquirer or payment brand.
    Limits of this row
    Planning only. Each rerun row keeps its own limits.
    References
    This page's own list.
  26. Z2. Clean-up
    Data involved
    Everything the run created.
    Who should reach it, and who should not
    Each system's owner removes records in that system.
    Test case
    List every test order, cart, customer, coupon, product, balance, saved address, export file and refund the run created, each carrying the marker or belonging to a test identity, and each connected test account that received any of them: the payment provider, fulfilment, loyalty and email. Remove them on the copy, and ask each system's owner to remove them there.
    Expected result
    Nothing the run created remains, or each item that does is listed with who removes it and when.
    Evidence to keep, and what it may relate to
    The clean-up list with dates. WSTG-BUSL-10's logistics notes include keeping a record of any orders made so that they can be cancelled and refunded.
    Limits of this row
    Deletes only what the run created: records carrying the run's marker or belonging to its test identities.
    References
    WSTG-BUSL-10, Logistics.
  27. Z3. Evidence pack, and what it may relate to
    Data involved
    The run's records, redacted.
    Who should reach it, and who should not
    The owner holds the pack and decides who sees it.
    Test case
    Planning. Keep together: B1's dated terms and answers, B2's readings, each row's redacted requests and responses with the build and date, the lists from S1, S2, E1 and F1, Z1's change log and Z2's clean-up list.
    Expected result
    Each item dated, with who holds it and who may see it.
    Evidence to keep, and what it may relate to
    May relate to: PCI DSS validation, which the PCI SSC's Merchant Resources FAQ sends to the acquirer or payment brand; PCI DSS v4.x Requirements 6.4.3 and 11.6.1 as the PCI SSC post describes them (S1, S2); APP 11.1 to 11.3 of the Privacy Act 1988, where the Act applies to the store; the Notifiable Data Breaches scheme, as the OAIC explains it, if a breach occurs; and customer questionnaires and insurer forms, covered by the security questionnaires and cyber insurance checklist guides. Applicability is for the owner to establish with the acquirer and their own adviser.
    Limits of this row
    No password, key, secret, card number or real customer record goes in the pack.
    References
    PCI SSC Merchant Resources FAQ; PCI SSC post of 10 March 2025; Privacy Act 1988, Compilation No. 104; OAIC, About the Notifiable Data Breaches scheme.

What ecommerce website security covers on a working store

Ecommerce website security splits between the platform with its extensions, which is code someone else wrote and fixes through updates, and the store itself: the prices, coupons, refunds and loyalty rules this business set, the staff and apps it gave access to, the scripts on its checkout and the systems it sends orders to.

An update can fix a flaw in the platform's code. It cannot tell you whether this store's refund permission sits with the right role, whether one of its coupons can be used twice, or whether its fulfilment connection accepts a status update from anyone who knows the address. Those are set store by store, and they are what this guide's matrix checks.

OWASP gives payment functionality a test of its own, WSTG-BUSL-10, in the edition of its Web Security Testing Guide that the OWASP site labels Latest, which the site says may frequently change; this page uses it as read on 15 November 2026.

The test's summary says many applications implement payment functionality, including e-commerce sites, and that vulnerabilities in it could allow attackers to steal from the organisation, make fraudulent purchases, or even steal payment card details from other users.

This page calls the sequence an order travels the order path: the customer account, the cart, coupons, the handoff to the payment provider and the confirmation that comes back, the order record and its emails, refunds, exports, fulfilment and loyalty. Each step reads or changes personal information, money or stock, and each has people and systems that should reach it and others that should not.

The matrix on this page gives every step a test case for an ecommerce security audit of your own store, to run yourself or hand to your developer, on a staging copy with the payment provider in test mode.

It covers the web side of ecommerce cyber security, meaning what customers and their data reach through the store. Office networks, staff devices and email accounts sit outside it.

Where customer and payment data travel on the order path

Before testing anything, it helps to know what each step carries and where it goes.

The three groups below follow an order from the first visit to the last system that receives a copy of it.

Accounts and carts

A customer account can hold names, email addresses, phone numbers, delivery and billing addresses and order history, and, where the store offers saved payment methods, whatever reference to them the store keeps. 1 min read

The payment handoff and the confirmation that comes back

WSTG-BUSL-10 lists the most common ways an application integrates payment functionality as redirecting the user to a third-party payment gateway, loading the gateway in an iframe on the application, an HTML form that makes a cross-domain POST request to the gateway, and accepting the card details directly and then posting from the application's back end to the gateway's API. 1 min read

After the sale: refunds, exports, fulfilment and loyalty

Once an order exists, its data keeps moving. 1 min read

Where access breaks at each step, and the guide that covers it

Each step fails in its own way, and the flaws below have guides of their own in this library.

This section names the step, the flaw and the guide, and leaves the test case to the matrix.

  • Accounts. One customer opening another customer's order, invoice, address or return request by changing an id in a request is broken access control, which the broken access control and IDOR guide explains with a two-account test. Taking over the account itself, through password reset, reused passwords or session handling, is the ground of the broken authentication, credential stuffing and session management guides.
  • Carts and checkout. A price, total, quantity or currency taken from the request is a business logic flaw, and turning an application's rules into invariants that a test can check is the subject of the business logic vulnerabilities guide. WSTG v4.2's integrity test, WSTG-BUSL-03, says an application must not depend on non-editable controls, drop-down menus or hidden fields for business logic processing, because these fields remain non-editable only in the context of the browser.
  • Coupons. WSTG v4.2's limits test, WSTG-BUSL-05, gives the example of an eCommerce site that may only allow a user to apply a discount once per transaction. WSTG-BUSL-10 adds checks for discount codes, among them whether codes are easily guessable, whether multiple codes can be applied at once or one code applied multiple times, and whether codes are exposed in the HTML source or hidden input fields. A single-use code redeemed by several requests at nearly the same moment is what the race conditions guide deals with.
  • Payment confirmation. An order marked paid because the browser reached a success page, or processed again when a confirmation repeats, is a broken payment flow in WSTG-BUSL-10's terms. Its race conditions section lists replaying the gateway's callback rapidly in parallel among its common test scenarios, and says that if the back end lacks proper idempotency checks, this can mark the same order as paid multiple times, among other effects. Signature and duplicate handling for provider notifications sits in the Stripe webhook security guide.
  • Payment page scripts. OWASP's Third Party JavaScript Management Cheat Sheet says the single greatest risk is a compromise of the third-party JavaScript server and the injection of malicious JavaScript into the original tag JavaScript, and that the content a tag manager returns can be dynamically changed by host site employees, such as the marketing part of the business. The supply chain attacks guide covers third-party code in general, and the payment page section below covers what the matrix records.
  • Refunds. A refund sent by a role that should not hold that permission is a vertical access failure, which the broken access control guide covers. A refund larger than the amount paid, or a cancellation that refunds an order that was already refunded in full, is the kind of case the business logic guide plays in its lab.
  • Exports. What an export should contain is the subject of the sensitive data exposure guide. Who may start an export, and whether its file answers a request with no session or after sign-out, are access questions: for each role and function, WSTG v4.2's authorisation test, WSTG-ATHZ-02, checks whether a resource can be accessed even if the user is not authenticated, and after log-out.
  • Fulfilment and loyalty. An inbound status update accepted without a signature, or acted on twice, has the same shape as an unverified or repeated payment event, so this page models its checks on the Stripe webhook security guide. WSTG v4.2's work-flow test, WSTG-BUSL-06, uses a loyalty scheme as its first example: where a customer cancels or removes items after points have been added, it says the system either should not apply points until the transaction is tendered, or should roll them back if the increment does not match the final tender.

What the WooCommerce, Shopify and Magento guides carry

Platform settings live in the platform guides, and this page does not repeat their worksheets.

The WooCommerce security guide maps what each gateway extension loads, answers and sends, and records on staging how test orders, a coupon, a refund and a cancellation move through that chain. The Shopify security guide puts Shopify's terms and documentation beside the settings each store owner controls, and its test-order checklist is run on a dev store only.

The Magento security guide covers Adobe's patch releases, the routes scripts take onto a Magento checkout, and the Admin and web API settings, with a register for finding a store's patch gap.

Each of the three has a section on the terms to read before anyone tests a store on that platform.

Read the guide for your platform first. This page then takes the order path across platforms. Its rows name a step and a test rather than a screen, so the developer who knows your store maps each row to the right page, request or setting, and a step your store does not have, such as guest checkout or a loyalty app, is marked not applicable rather than left blank.

Payment page scripts and PCI DSS v4.x Requirements 6.4.3 and 11.6.1

For a store, the question about payment page scripts is which ones run there, who approved each one, and who would notice a change.

On 10 March 2025 the PCI Security Standards Council published a post about its new information supplement, "Payment Page Security and Preventing E-Skimming". In the post's words, PCI DSS v4.x "includes Requirements 6.4.3 and 11.6.1, which aim to reduce the risk during e-commerce transactions".

According to the post, the requirements focus on payment page scripts being properly authorised, checked for integrity and monitored for tampering, and they also address the need to manage both scripts and security-impacting HTTP headers carefully, to prevent unauthorised changes to webpages whether in the server environment or as rendered in the consumer's browser. It names v4.0.1 as the version of PCI DSS current at the time of publication.

Other sources describe the controls a store can read for. OWASP's Third Party JavaScript Management Cheat Sheet says there is usually no guarantee that the code hosted at the third party will remain the same as seen by the developers and testers, and lists in-house script mirroring, Subresource Integrity and secure transmission of the third-party code among typical defences.

MDN says Subresource Integrity lets browsers verify that resources they fetch are delivered without unexpected manipulation, by providing a cryptographic hash that a fetched resource must match, and that the browser refuses to load a resource whose hash matches none of the values in its integrity attribute.

MDN also says that an attacker who gains control of a third-party host can inject arbitrary malicious content into its files or replace the files completely.

OWASP's Content Security Policy Cheat Sheet says the Content-Security-Policy-Report-Only header delivers a policy that is not enforced, while violation reports are still printed to the console, and delivered to a violation endpoint if the report-to and report-uri directives are used.

Rows S1 and S2 of the matrix collect the facts those sources point at: each script on the payment page, its business reason, who approved it, how its integrity is checked, which headers the page sends and who hears about a change. On enforcement, the PCI SSC post is explicit: the Council does not enforce compliance or determine whether specific implementations are compliant, and entities should collaborate with the organisations managing their compliance programmes, such as acquirers or payment brands, to understand any specific compliance validation and reporting responsibilities.

Whether 6.4.3 or 11.6.1 bears on a particular store is left to those organisations here.

A filled-in row is a record to take to them, not a compliance statement, and this library's payment script controls guide covers the requirements themselves.

Card payment and privacy questions a store settles with others

Each part below reports a source's own text and the people or businesses it names.

None of it decides anything for your store: card payment questions go to your acquirer or payment provider, and privacy questions to your own adviser.

PCI DSS: validation questions go to your acquirer or the payment brand

The Council's Merchant Resources page carries an FAQ for merchants. 1 min read

The Privacy Act 1988: start with whether it covers the store

Coverage under the Privacy Act 1988 turns on its definitions. 3 min read

Notifiable data breaches

The OAIC explains how the Notifiable Data Breaches scheme applies to entities covered by the Privacy Act 1988. 1 min read

Platform, provider and app terms

Before any row that sends a request, read the terms of each business whose system the row touches: the platform, the host, each payment provider and each fulfilment, loyalty or gift card app. 1 min read

Questionnaires and insurer forms

A business customer, a marketplace or an insurer may send a store its own security questions. 1 min read

How to run the order-path matrix on your own store

The matrix is written for the owner or ecommerce manager, working alongside whoever builds and maintains the store.

Read down its first column and it works as an online store security checklist; read across a row and it is a test case with its own limits. Keep one copy per store, in a spreadsheet, and work through it in this order.

  • Rows B1 and B2 first, and nothing else until both are answered. B1 collects the terms and the acquirer's answer, and B2 confirms the staging copy, the test modes and the test identities.
  • Every step that places an order, submits a form or changes a request runs on a staging copy of your own store, with each payment provider switched to the test or sandbox mode that provider documents, using only the run's test identities: the invented customers cust-a and cust-b, whose email addresses and phone numbers are ones the business controls, and staff test accounts. Never use a real card, and never a real customer's account. Steps that read the live store, such as its payment modes, coupon list or staff roles, only read, and save nothing.
  • Keep to each row's cap and to the run's total of 20 checkout submissions. Rows D2 and L2 send three requests at nearly the same moment, once each. Nothing in the matrix measures how much traffic the store can take.
  • Write the expected result for your store before you send anything, from your own rules: the price list, the coupon terms, the refund policy and the loyalty rules. A cell you cannot fill in is a rule nobody has written down.
  • Change one thing per request, keep each request and response with personal details and tokens redacted, and read the stored records afterwards, not only the response.
  • If any response shows a real person's details, stop that row, record the request and tell the owner before going further.
  • Rows B1, P2, S2 and F1 are read only and change nothing: they read settings, lists and documents. S1 loads pages, as a visitor on the live store and as cust-a on the staging copy, and changes nothing beyond one test cart on the staging copy. Leave each screen without saving, and never record a password, key, signing secret or card number: record whether one exists and the name of the setting that holds it. A link that opens an order, a file or a payment without a sign-in is a secret too: open it only for a test record, and never copy it into the evidence.
  • Finish with Z2, the clean-up, and keep the results together as Z3 describes.

Significant changes: when the matrix runs again

A clean run describes the store on the build and the day it ran against.

The order path changes whenever someone alters the checkout, adds an app or a tag, edits a coupon or refund rule, or connects a new system. On this page's reading, a change to one step can reach others: a new payment app can bring a script for the payment page, a callback address and a refund path at once.

Row Z1 is this page's own list of significant changes, each with the rows it reruns. Run them on the staging copy before the change reaches the live store where you can, or as soon after as the release allows, and log the change, the rows, the build and the date.

This list sets no testing or scanning schedule for PCI DSS purposes and stands in for none. For validation and reporting responsibilities, the PCI SSC post sends entities to the organisations managing their compliance programmes, such as acquirers or payment brands.

The same logic applies to an outside test. A Clever Ops test includes one consolidated retest, which covers the reported findings within the agreed scope; new features and unrelated changes are scoped as separate work, because they were not part of the original test.

What a run on your own store cannot show

A run records how one staging copy behaved on one day, what the live store showed to a visitor, and the settings and answers read on that day.

Several questions sit beyond it:

  • How the live store compares with the copy. A copy can carry different apps, scripts, settings and connections, and a clean result covers only what the copy carried.
  • What a payment provider, fulfilment provider or app does with the data on its own systems. The matrix records what the store sends and who owns each receiver.
  • What ran on the payment page last month, or whether a third-party script changes after you read it. OWASP's cheat sheet says there is usually no guarantee that the code hosted at the third party stays the same as the developers and testers saw it.
  • That a timing gap is closed. Three requests that behave as expected in D2 or L2 are a result for that burst, not proof that no gap exists.
  • Flaws on paths no row exercised, such as staff-only plugins, trade price lists, subscriptions or a marketplace with outside sellers. Those need a scope of their own.
  • Whether someone has already broken into the store. Finding that out belongs to an incident response specialist, and the hacked website guide covers the first hour.
  • Whether any duty applies to your store. The evidence column says what a record may relate to; applicability is for you, your acquirer and your adviser to settle.

Ecommerce penetration testing and audits with Clever Ops

Clever Ops tests what customers reach through a browser or an API, and for a store the right service depends on how the store is built.

A WooCommerce store falls under the WordPress security audit. It works through every role with a login, covers the login and password-reset flows and the plugin and theme versions installed, and on a WooCommerce store takes in customer accounts, orders and checkout pages as well. The WooCommerce security guide in this library sets out what those additions mean on a store.

A custom-built store that customers sign in to is a job for the web application and API penetration test. That test signs in and works through sign-in and session handling, what each role is able to see and change, whether one customer can get at another's records, direct calls to the API as well as use through the screens, payment and billing flows, and the business rules the store is meant to enforce.

Its payment testing covers the store's side of the payment, not the payment provider's platform, and runs against a staging copy linked to a non-live payment account wherever one is available, so no real card is charged.

Very logic-heavy products, such as marketplaces, are scoped individually or referred to a larger firm. Live Shopify storefronts are outside Clever Ops services, so this page describes no Clever Ops work on one.

Some of the work goes elsewhere. Clever Ops is not a PCI Approved Scanning Vendor or a Qualified Security Assessor, so where a store needs a scan from an Approved Scanning Vendor or an assessment by a Qualified Security Assessor, we refer it to a provider on the Council's list for that role.

Clever Ops is not CREST accredited, so where a tender or contract requires a CREST-accredited provider, we refer that work to a provider that is. The payment provider's hosted pages and platform, your host's platform, and services other businesses run, such as a fulfilment or loyalty provider's own platform, stay out of scope unless their owner agrees.

If the store may already be compromised, call an incident response specialist first: Clever Ops does not provide incident response or malware cleanup, and an audit can follow the cleanup.

If Clever Ops built or hosts the store, the work is a security review rather than an independent test, and the report names who built it and who hosts it. A report is evidence for a decision someone else makes, such as an auditor or a customer, and it is not a certificate.

Bring the filled-in matrix to the scoping call: it shows which steps your store has, which systems it connects to and which terms are already answered.

Frequently asked questions

This page does not say, and the PCI Security Standards Council points merchants elsewhere for the answer. Its Merchant Resources FAQ says self-assessment questionnaires are validation tools for SAQ-eligible merchants and service providers, with several developed for specific types of environments as defined in each SAQ's eligibility criteria. It says whether a small merchant is required to validate compliance is determined by the individual payment brands, and that merchants should contact their acquirer (merchant bank) or payment brand about compliance validation and reporting requirements. Ask in writing, and keep the answer with row B1 of the matrix.

The Council's Merchant Resources FAQ says PCI DSS is intended for all entities that store, process or transmit cardholder data and/or sensitive authentication data, or could impact the security of that data. Its post on payment page security describes PCI DSS v4.x Requirements 6.4.3 and 11.6.1 as focused on payment page scripts being properly authorised, checked for integrity and monitored for tampering. How either applies to a store that hands payment to a provider is a question for your acquirer or payment provider. Rows S1 and S2 give you the script and header record to take to that conversation.

Not with this matrix, which never uses a real card and runs every order on a staging copy with the payment provider in its documented test mode. WSTG-BUSL-10 discusses the logistics of testing payment functionality, including keeping a record of any orders made so they can be cancelled and refunded, and not placing orders that cannot be cancelled or that will cause other actions, such as goods being immediately dispatched from a warehouse. It also says test card details should only be usable on development or sandbox versions of payment gateways, but may be accepted on live sites if they have been misconfigured. Row B2 reads the live store's payment mode for that reason.

The Act's definitions decide it, and this page cannot apply them to your business. A small business operator is not an organisation under section 6C(1). The turnover line in section 6D(1) is $3,000,000 or less for the previous financial year, section 6D(2) tests a business not carried on in that year differently, and under section 6D(3) a small business operator carries on one or more small businesses and no business that is not a small business. Section 6D(4) then carves some entities out regardless, among them one whose business has had an annual turnover of more than $3,000,000 for a financial year that has ended after the later of the time it started to carry on the business and the section's commencement, and one that discloses personal information about another individual to anyone else for a benefit, service or advantage. Later subsections qualify some of those exceptions. Take the question to your own privacy adviser.

Whatever the work is called, ask what it signs in as and which requests it changes. In Clever Ops terms, a website security check stays outside the login. A WooCommerce store gets the WordPress security audit, which takes in every role with a login and, on a store, customer accounts, orders and checkout pages. A custom-built store that customers sign in to gets the web application and API penetration test, which looks at what each role can see and change, payment and billing flows, and the rules the store is meant to enforce. Live Shopify storefronts are outside Clever Ops services.

Row Z1 lists the changes this page treats as significant, with the rows each one reruns: a new payment provider or integration method, a change to cart, checkout or account templates, an app or plugin added, removed or updated that touches carts, coupons, orders, customers or checkout, a new script, tag or tag manager change on payment pages, changed coupon, refund, loyalty or gift card rules, a new fulfilment, accounting or marketing connection, a new platform version or theme, and a change of agency, developer or host. The list is this page's own and sets no PCI DSS testing schedule; questions about PCI DSS validation go to your acquirer or payment brand, as the PCI SSC's Merchant Resources FAQ says.

Scope the right test for your application

Tell us what you have built, who uses it and when it needs to launch. You will get a plain-English recommendation: a website audit, a web application test or a referral to a specialist.