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

Performance Marketing › Advanced

Aligning Conversion Signals Across Platforms for a Boiler Installer

You audit an invented boiler installer whose Google and Meta campaigns each learn from a different, inflated form event and neither knows which quotes became installed boilers. The exercise practises specifying enhanced conversions, a deduplicated Conversions API set-up and CRM outcome imports, then choosing one shared outcome both platforms are judged on, the event each platform bids on and the conditions for switching.
ExerciseAdvanced levelModule 1: Conversion signals across platforms7 minUpdated

Why this matters

Two platforms optimising towards different events find different people, so comparing them on their own reported events says more about the events than the channels. Define one outcome in the CRM and judge both platforms on it, even where one has to bid on an earlier stage for a while, and budget moves on outcomes. The enhanced conversions guide and the Conversions API guide cover each side's mechanics, and the guide to connecting a CRM to ads and analytics the pattern that joins them. The cost targets assume the intermediate performance marketing level.

The installer and its conversion events

Quorncliffe Heating Services is an invented boiler installer. Search and social ads drive visitors to a quote form, a coordinator books a survey and the CRM records each stage. Meta reports the cheapest leads, and the owner doubts it is the cheapest channel.

Illustrative example — not a real result

Quorncliffe Heating Services: events and one quarter's results

PlatformEvent bidding learns fromWhen it firesHow it is sentCounting
Google AdsQuote formThank-you page viewGoogle tag, no enhanced conversionsEvery conversion, so reloads count again
MetaLeadClick on the submit button, before the form validatesPixel onlyEvery click, including failed submissions
  • The form collects name, email, phone, postcode and boiler age. Hidden fields store the campaign name only; no click or browser identifiers are kept.
  • Surveys are usually booked within three days of the enquiry; installation follows two to four weeks after it.
  • The consent banner records advertising consent on 64% of form submissions.
  • About 15% of quote requests arrive by phone and are logged in the CRM by the call handler.
  • Margin per installed boiler after parts, labour and finance fees: £1,050. The owner will spend up to 40% of it, £420, to win an installation.

One quarter by platform:

Row labelGoogle AdsMeta
Spend£27,000£16,500
Conversions the platform reports690610
Quote requests in the CRM with that campaign source450330
Surveys booked21666
Boilers installed8115

One outcome, two platforms

A shared signal is one outcome, defined once in the CRM and sent to both platforms from the same trigger, under the same consent, timing and identifier rules, with the CRM as the count of record. Each platform still matches the outcome to its own ads under its own Attribution rules, so their totals overlap; the guide to why platforms and analytics disagree explains why that is normal and why neither total decides the budget split. Pick the latest stage that still arrives quickly and often enough to learn from on the thinner platform, not just the busier one, and test that against the platform's learning threshold using the volume left after consent. If no shared stage passes, the thinner platform bids on an earlier stage while the shared outcome stays the stage both are judged on.

Task

  1. Audit each platform's event: what it counts, what it misses, how far it sits from the CRM.
  2. Calculate each platform's cost per reported conversion, CRM quote, survey and installation, and say what its bidding has been learning.
  3. Specify the Google side: counting, enhanced conversions and the import route for CRM stages.
  4. Specify the Meta side: server events, deduplication, CRM stage events, identifiers and consent.
  5. Check each stage's weekly volume on Meta after consent against its learning threshold, then choose and define the shared outcome, name the event each platform bids on and list the stages that stay secondary.
  6. Write the switching sequence with a condition per step, and the reconciliation check.

Worked answer

Illustrative example — not a real result

Quorncliffe's costs by stage

Row labelGoogle AdsMeta
Reported conversions per CRM quote690 ÷ 450 ≈ 1.53610 ÷ 330 ≈ 1.85
Cost per reported conversion≈ £39≈ £27
Cost per CRM quote£60£50
Cost per survey booked£125£250
Cost per installation≈ £333£1,100
Survey to installation81 ÷ 216 = 37.5%15 ÷ 66 ≈ 22.7%

Surveys a month: about 72 from Google and 22 from Meta. Installations a month: 27 and 5. Account-wide survey to installation: 96 ÷ 282 ≈ 34%, so the most the owner should pay per survey is about £420 × 0.34 ≈ £143.

Meta volume check, per week over a 13-week quarter. Phone enquiries, about 15% of quote requests, cannot be matched to an ad, and only form submissions carry the 64% consent rate, so the check assumes surveys follow the same 15% and 64% as quote requests. Surveys: 66 ÷ 13 ≈ 5 in the CRM, and 66 × 0.85 × 0.64 ÷ 13 ≈ 2.8 after phone and consent. Quote requests: 330 ÷ 13 ≈ 25 in the CRM, and 330 × 0.85 × 0.64 ÷ 13 ≈ 13.8 after phone and consent. Meta's documented exit threshold is roughly 50 optimisation events in 7 days (check Meta's current documentation), so survey booked reaches under a tenth of it and the quote request a little over a quarter.

What bidding learnt. Google learnt from a page view that counts reloads; Meta from button clicks, failed submissions included, so it found people who press buttons. Meta looks cheapest per reported conversion and costs more per installation than the margin itself.

Google specification. Count one conversion per click, fired on a confirmed submission pushed to the data layer with the quote reference. Turn on enhanced conversions through Tag Manager, hashing normalised email and phone, gated by consent. Import survey booked and installation, the latter with the margin as value, through enhanced conversions for leads, and only that route: a parallel click identifier import would double count.

Meta specification. Fire the lead on confirmed submission from pixel and server, with the quote reference as the shared event ID and the same event name. Store the browser identifiers in hidden fields so the server event carries them with hashed email, phone and postcode. Send survey booked and installation from the CRM through the Conversions API, keyed to the identifiers captured at the form. Consent travels with the lead record, and the server sends nothing for anyone who declined. Verify test events, deduplication share and match rating first.

Shared outcome. Survey booked: a confirmed surveyor appointment, once per household per enquiry, sent to both platforms by the same CRM trigger and used as the stage both are compared on each month. Google bids on it. Meta cannot learn from it at this volume, about 3 events a week against a threshold of roughly 50, and the quote request is also short at about 14, so Meta's bidding event is the confirmed quote request, the corrected Lead from the Meta specification and the more frequent stage, as a temporary measure to review, with survey booked sent alongside as a secondary signal. Installation stays secondary: it is too rare on Meta and too late on both. Phone enquiries cannot be matched; they stay a documented gap in the CRM report.

Switching sequence. First fix counting and deduplication only, which turns Meta's Lead into the confirmed quote request, date-note both accounts and expect reported conversions to fall with no change in the business. Once a full month of survey events reconciles with the CRM, switch Google's goal to survey booked with a cost target near its observed cost per survey, changing the goal and not the strategy. A fortnight later, tighten Meta's set-up while its bidding event stays the confirmed quote request: send survey booked and installation as secondary signals, consolidate ad sets and set a cost target no higher than the account-wide ceiling restated per quote, using the ceiling worked out in the costs example above (the platform sees only consented quotes, so its own figure reads higher), on a budget cut to test level while it relearns. Move Meta's goal to survey booked only when its consented survey events approach the threshold within 7 days. Judge both monthly on cost per survey booked in the CRM, and check the budget split against cost per installation once the installations from those surveys have arrived.

Reconciliation. Monthly, compare each platform's surveys with the CRM's, track the unmatched share and show both claims beside the CRM total.

Common mistakes

  • Comparing platforms on the events they each report. The cheaper event wins, not the better channel. Compare them on the shared outcome in the CRM.
  • Running the click identifier import and enhanced conversions for leads together. One lead, two conversions.
  • Choosing installation as the bidding goal for both platforms. Too few and too late for the thinner platform.
  • Picking a stage without checking its weekly volume after consent. Survey booked looked sensible until the count showed about 3 a week on Meta.
  • Judging the platforms on a target cost per action (CPA) after the switch. The CRM's cost per survey and cost per installation decide the split.

Next, test your judgement in the conversion signals checkpoint.

Pages this lesson applies

Terms used on this page

Also related

Related