What this checkpoint covers
Cragmoor Fitness Studios is an invented chain of gyms that sells memberships online and through free trial passes claimed from Meta lead ads. The questions draw on the low event match quality troubleshooting page, the guide to connecting a CRM to ads and analytics and the attribution settings guide. They assume the intermediate measurement work and the jewellery attribution exercise.
Task
Write a short answer and its reason for each question before reading on; an answer counts only if the reasoning matches. Reread the linked page for every miss, then note the first signal fix you would make on an account you manage and how you would verify it.
Questions
- Cragmoor's membership purchase is sent from the server with only an IP address and user agent, although the sign-up form collects email, phone, name and postcode. The operations manager suggests also sending events for people who declined advertising cookies, "because it all happens on the server anyway". What do you advise?
- After phone numbers are added, coverage shows phone on nearly every server event but the match rating barely moves. The developer has also hashed the IP address and user agent "to be safe". What is likely wrong?
- Free trial leads are cheap, but few trialists become members. The team wants to send something back from the CRM. Which stages do you send, and which one should the lead campaign optimise for?
- A new marketing lead wants to switch every ad set to a one-day click window "so the numbers are honest". What happens to reporting, and what happens to delivery?
- Studio A's campaign uses the default click-and-view window, Studio B's uses a week-long click-only window. The board asks why Studio A's cost per membership is so much lower. How do you answer?
- On Monday, last week's results look far stronger than yesterday's, and the area manager wants yesterday's ads paused. What do you say?
- Meta reports noticeably more memberships than the analytics tool credits to paid social, and some results are flagged as partially modelled. Many visitors decline cookies. The finance director concludes Meta is inventing members. How do you explain the gap?
Answers explained
- Add the identifiers Cragmoor holds and may share, and send nothing for people who declined. Email and phone are the strongest fields. Sending from the server does not change the consent decision: the stored state travels with the order, and declined visitors produce no event at all. The lower match ceiling is correct, not a fault (consent limits).
- Formatting. Phone numbers stored with spaces or without the country code produce hashes that match nobody, so coverage overstates usable data. Normalise to the documented format before hashing. The IP address and user agent are expected unhashed, so hashing them removed useful signal. A unit test that hashes a known value keeps the fix in place (hashing and format errors).
- Send trial attended and membership started, with its value. Form fills duplicate what the lead ad already reports. Trial attendance arrives soon and often enough to steer delivery, so the campaign optimises for it once volume allows; memberships are the reporting outcome. The CRM must keep the identifiers captured at the form, unchanged, for the match to work (which stage to send).
- Reported results fall, and delivery changes too. The short window credits fewer conversions, and the conversions it credits are the examples delivery learns from, so the system narrows towards people who join almost immediately. Honest reporting comes from the compare columns, not from retraining delivery (optimisation versus reporting).
- The comparison is invalid until both sit under one window. A wider window always reports more results and a lower CPACost of each conversion.. Put both studios in the same compare column, then judge; any target must state the window it was set under (comparing windows).
- Yesterday is not finished. Meta reports by the date of the ad interaction, and recent days fill in as delayed and modelled data arrives. Judge complete periods (modelled and aggregated conversions).
- The two systems count by different rules. Meta credits views and clicks within its window, matches people across devices and models what it could not observe; analytics follows a last-source rule within one browser and loses visitors who decline. Compare a click-only view with paid social in analytics, accept the identity gap, and keep a separate source of truth for budgets. Better matching through the Conversions APIServer-side event sharing with Meta. narrows the part caused by thin signals (reconciling with analytics).
Where to go next
If you missed the signal questions, redo the event-matching and CRM steps of the jewellery exercise and write a parameter contract for Cragmoor's membership event. If you missed the setting questions, reread how the window trains delivery and how AttributionAssigning conversion credit to touchpoints. rules differ between systems. Then move to the incrementality module, starting with the baby clothing holdout exercise, or review earlier material on the intermediate Meta Ads level.
