UNDERSTAND PATTERNS
Follow recorded journeys and compare changes.
Follow an interaction, save a funnel and compare two periods with clear sample counts.
In this guide 7 sections
Review a clue in context
#A clue is a lead, not proof of a broken button. Legitimate interactions can meet these conditions. The review status applies to this observation, not an entire website.
- Open Clues and choose an observation. Watch opens shortly before its recorded moment.
- Inspect what happened before and after. Three clicks within one second can create a repeated-click clue; an eligible button with no observed response for about 1.5 seconds can create a possible unresponsive-click clue.
- Mark the observation New, Reviewing, Resolved or Ignored. Note the evidence and the next thing you will check.
Follow an ordered path
#Journeys shows page paths within a retained recording. Save up to twenty funnels, each with two to eight ordered steps matching page paths, clues or form observations. A path prefix includes its child paths.
Use a recorded layout ID when structural field positions matter. Journeys do not identify a person across separate visits or devices.
Compare two periods fairly
#The instant Compare screen analyzes batches of up to 200 retained recordings. Use Whole-range reports for all eligible recordings in the period. Consent, sampling, retention and bounded observations affect coverage. Changes in rates do not establish that a website edit caused the difference. Older recordings do not receive invented new observations.
- Choose two separate UTC date ranges.
- Use matching page, device and layout filters where appropriate.
- Read the sample counts and denominators before interpreting rates.
- Inspect representative replays from both periods before drawing a conclusion.
Build a whole-period report
#Use Whole-range reports for all eligible retained recordings in the chosen UTC period. Progress and the observation cutoff remain visible while bounded jobs process the indexes. WP-Cron continues work; manual advancement also works on low-traffic sites. Refresh to include later arrivals.
Counts exclude deleted and expired recordings. A report cannot reconstruct observations that were never captured. Legacy instant Discovery views still show their sample limits; heatmap aggregation follows its separate existing path.
Temporary jobs expire after one hour and are limited to two per reviewer and eight per site. Each job can hold up to 100,000 recording memberships and 500,000 derived facts; reaching a bound is labelled partial. Narrow the period to complete a larger investigation. Each recording retains up to 300 indexed observations.
Invalid ranges and failed requests keep the last successful report visible. Its scope remains the scope shown with that report; it is not a result for the rejected dates. New results replace it after a successful build.
Follow up a recurring clue
#Whole-range reports groups repeated observations by kind, page, layout, device and structural target. Review affected visits, frequency, trend and replay moments. Set an owner and review state, and record an edit date when relevant.
Before/after comparisons describe observations. They cannot establish that your edit caused a change. Layout changes, sampling, consent and different traffic can all affect the result.
Configure meaningful goals
#Use Goals to create fixed named success observations and select supported form adapters. Manual integration must call the goal only from your actual successful-completion callback. Never pass form values, email addresses, customer IDs or other identifying content.
An integration-reported completion is separate from a submit attempt and does not prove delivery, payment or a business result. Configuring enabled goals changes the consent scope. Use timed funnel steps and the product, enquiry and sign-up templates to inspect the journey before completion.
Write a comparison that another reviewer can check
#Record the question, two UTC windows, page/device/layout filters, retained sample counts and the supporting replay references. Note capture interruptions, changed consent or sampling, and any layout change.
Illustrative example: a route appears in 6 of 18 reviewed recordings before a menu edit and 9 of 24 afterward. Report those denominators with the counts. The observations alone do not show that the menu edit improved usability or caused the difference.
If the earlier recordings have expired, or capture conditions cannot be compared, report the comparison as inconclusive. A new visit cannot recreate a missing baseline.
Use the menu-rename review sheet or explore journeys and funnels.