Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Performance Marketing › Advanced

Checkpoint: Conversion Signals Across Ad Platforms

Scenario questions set in an invented pet insurer test the judgements of the first advanced module: why platforms optimising to different events drift apart, what server-side and hashed-data matching can and cannot recover, how deduplication fails, when bidding should move to a later outcome, how consent and minimisation bound the data sent, and how to read match diagnostics without mistaking them for results.
CheckpointAdvanced levelModule 1: Conversion signals across platforms4 minUpdated

What this checkpoint covers

Marrowfield Pet Insurance is an invented insurer whose visitors start quotes online and buy online or by phone; its policy system records every sale and cooling-off cancellation. The questions draw on the enhanced conversions guide, the Conversions API guide and the guide to connecting a CRM to ads and analytics, and they assume the boiler installer signals exercise.

Task

Write an answer and its reason for each question before reading the explanations; it counts only if the reasoning matches. Then list the signal each of your own platforms learns from and whether it is the same everywhere.

Questions

  1. Marrowfield's Google campaigns bid towards quotes started and its Meta campaigns towards policies bought online. After a few months, Google brings far more quotes and Meta far fewer, better customers. The board asks which platform is better. What do you say?
  2. After enhanced conversions and the Conversions API go live, the developer says every lost sale has now been recovered and both platforms finally report the true number of policies. What is wrong with that claim?
  3. The week the server integration launches, Meta's reported purchases nearly double while the policy system is flat. The browser sends the quote reference as its event ID and the server sends the policy number. What happened, and what do you do?
  4. Online buyers usually purchase within a day of their quote; phone buyers take up to a fortnight, and a share of policies are cancelled in the cooling-off period. The team wants to move both platforms' bidding from quotes to policies tomorrow. What do you advise?
  5. The CRM team proposes sending each customer's full record, including the pet's breed and medical history, to both platforms "for better matching", for every customer, because the data is hashed anyway. How do you respond?
  6. Meta's match rating for the server purchase event rises after phone numbers are added, and reported cost per policy falls the same month. The agency calls it a performance win. Is it?
  7. Google and Meta together report more policies than the policy system recorded for the same month. The finance director asks which platform is lying. How do you answer?

Answers explained

  1. Neither comparison is fair yet. Each platform optimises towards the event it is given, so Google found people who start quotes and Meta found people who buy. The difference describes the events, not the channels. Give both the same outcome from the same source, then compare on cost per policy in the policy system (why the stage you send matters).
  2. They measure more of what can be matched, not what was caused. Hashed data and server events recover sales lost to blockers, expired identifiers and cross-device journeys, but only for people the platform can match and who consented. Unmatched buyers stay invisible, the two platforms still overlap, and neither tool says whether the ad made the sale; that needs a test (when the gain is small).
  3. Deduplication failed. Meta keeps one copy only when event name and event ID match, so it counted both. Generate one identifier when the action happens, pass it to pixel and server, check the deduplication share in Events Manager, and annotate the launch so the doubled weeks are not read as a win (deduplication with event IDs).
  4. Not tomorrow, and not policies first. A policy is a late outcome for phone buyers, and cooling-off cancellations would teach bidding from sales that later vanish. Send a qualified quote or the policy as it is bought, adjust for cancellations when the platform supports it, check that uploads reconcile for a full month, then change the goal before the strategy, holding the cost target (CPA) near what history shows (sending stages back).
  5. No. Hashing reduces risk; it does not create consent. Customers who declined advertising must be left out on the server as well, and matching needs only the documented fields, normalised and hashed. Breed and medical history improve nothing and should never leave the system (consent and data minimisation).
  6. It is a measurement change. A better match rating means more of the same sales are attributed, so reported cost per policy falls with no extra policies. Check cost per policy in the policy system before and after the date, and keep the rating as a diagnostic for missing parameters (event match quality).
  7. Neither is lying. Each platform credits the policies it can connect to its own ads under its own Attribution rules, and many buyers saw both. The policy system is the count of record; the platforms' totals inform optimisation within each channel, and the split between channels needs an incrementality test. Even a perfect Conversions API set-up leaves the overlap in place (closed-loop reporting).

Where to go next

If you missed the deduplication or consent questions, reread the Conversions API guide; if you missed the switching question, redo that step of the boiler installer exercise. Then continue to the incrementality module with the flower delivery holdout exercise, or revisit target setting on the intermediate performance marketing level.

Pages this lesson applies

Terms used on this page

Also related

Related