Search the guides

Filtering & analytics

When a filter blocks a buyer: a recovery playbook

Find the rule that stopped a legitimate visitor, restore the narrowest safe path and measure whether the fix worked.

Begin with one reproducible failure

A buyer says the form refuses to send, but it works on your laptop. Start with that buyer’s failed action. A false positive is a legitimate request that a protection tool mistakenly blocks or challenges. Your goal is to make that action work again and identify why it failed.

Ask when the error appeared, which page was open and what the message said. A request reference is an identifier that support can use to find that specific attempt in the logs. If one is visible, collect it along with the browser name. You do not need passwords, card details or the person’s browsing history.

Check whether the complaints began after a rule or site change. If many customers are affected, reverting that particular change may restore service while you investigate. Record the previous setting and the time of the rollback so you can tell which change helped.

Find the component that made the decision

Event flowWhere did the request stop?
  1. BrowserFind the error time and request reference.
  2. Protection serviceLook for a block or challenge.
  3. Site serverCheck whether the request arrived.
  4. Receiving systemConfirm the enquiry or order exists.

A request missing from the application may have stopped at an earlier layer.

Start at the first service that receives the browser’s request. A CDN is a delivery service in front of your website; it may also run security checks. A firewall applies rules that allow, challenge or reject requests. Either can stop a visit before it reaches the server running your site. An empty application log therefore does not prove that the buyer never tried.

Use the same time and request reference to follow the attempt through those services. If the site received it, inspect the response and then the separate request made when the buyer pressed Send. Check the documentation for the exact protection product and mode before adding an exception: an exception supported in one mode may not work in another.

Record the rule name or number, the action taken and the reason shown in the log. A risk score alone does not tell you which browser setting caused the problem. If the service does not expose a reason, mark the cause as unresolved and gather a controlled test instead of guessing.

Restore the smallest necessary path

Match the correction to the request that failed. If one form address is affected, review the rule for that address before changing protection across the whole website. An exception should say what it allows, who owns it and when it will be reviewed. Keep the old setting so the change can be reversed.

Consider a hypothetical office where several employees send enquiries through one shared internet address. A rate limit—a cap on requests within a time window—may count them as one unusually active visitor. Check which requests the limit counts and whether it groups unrelated people together. Raising the limit for everyone can hide that mistake; the correction should address the counting rule while still limiting repeated abuse.

  • Reproduce the original failure before changing multiple settings.
  • Change one relevant rule or exception at a time.
  • Retest the exact failed journey and a known working journey.
  • Check that the original abusive pattern is still addressed.
  • Record a rollback path and remove expired temporary exceptions.
Two approachesA recovery fix versus a blind reset

Controlled correction

  • One linked rule
  • The same failed journey retested
  • A recorded rollback

Hard to interpret

  • Several layers disabled together
  • A different device used for acceptance
  • No record of the change

A narrow change makes the result easier to verify and reverse.

A worked example: the page opens, the form fails

Suppose a customer opens a mobile landing page, completes the form and sees “Access denied” after pressing Send. Loading the page and sending the form are separate requests. Find the second request using the submission time, rather than checking only whether the page opened.

Diagnostic routeFollow the failed action
  1. Symptom

    The form shows an access error.

  2. Locate

    Find the submission request by time.

  3. Narrow

    Identify its rule and action.

  4. Retest

    The receiving system gets one enquiry.

Keep the page load and the form submission as two separate checks.

In an authorised test, repeat the same mobile journey and confirm which rule rejects the submission. Change that rule only as far as the evidence supports, then repeat the test with the original browser and connection conditions. A successful submission from a different desktop browser would not establish that the mobile problem is fixed.

Check the system that stores enquiries, not just the success message in the browser. The test should produce one enquiry with the expected details. If the customer tried repeatedly during the incident, investigate whether those attempts created duplicate records before treating each record as a separate lead.

Measure real recovery

Define success before testing: the page opens, the form sends and the receiving system records the enquiry. For checkout, use the site’s approved test purchase process and confirm the expected order state. Compare the result under the conditions that originally failed; a fall in blocked requests is useful evidence, but it is not proof of recovery.

Also inspect the requests made after the first page load. The page can appear normal while a script that controls the Send button is still blocked. If the form now submits but the enquiry never appears, continue tracing delivery to the receiving system. Do not close the incident at the first successful HTTP response.

Turn the incident into a regression check

Save a short repeatable check from the incident. For example: open the landing page in the affected mobile browser, complete the form over the relevant connection and confirm one enquiry arrives. This is a regression check: it catches the same fault if a later change brings it back.

Repeat that check after changes to forms, security rules, caching or tracking. Caching means reusing a saved response instead of generating it again, so it can also affect what a visitor sees after a fix. Note release times and compare complaints and completed enquiries for the affected group before and after the change.

Assign someone to remove temporary exceptions when they are no longer needed. Keep the failed request reference, the rule that changed and the successful retest together. That record lets the next person investigate the same symptom without disabling several protections and starting again.

Questions answered

Should I turn off all protection to restore access?

First locate the rule linked to the failed request. Change that rule or roll back the relevant recent change, then retest. If a widespread incident requires a broader temporary rollback, record its scope and assign someone to restore protection.

Why is the failed request missing from the site’s log?

A delivery or security service may have stopped it before it reached the site. Search that service’s log using the same time and request reference. Also check that you are looking for the form submission, not only the earlier page load.

ADSTRAFFIC / ENAll guides →