Paid ads conversion tracking setup checklist

Nothing else in paid media works until this does. Bidding algorithms train on your conversion feed, budget decisions come out of it, and every optimisation argument eventually resolves to whether the number is real. Build this before you spend.

The order below is a build sequence. Steps depend on the ones above them, so skipping ahead creates rework.

Step 1: Define the conversion event taxonomy before touching a tag

Five events, five written definitions, five named owners. Not a Google Doc nobody reads: a row in the CRM data dictionary with a person’s name on it.

The failure mode here is definitional drift. Six months in, marketing counts qualified leads at form submit and sales counts them at accept, and the CPQL argument becomes unwinnable. Write the definitions down, date them, and require a change request to alter one.

Step 2: Install and verify the three browser tags

The most common real-world break: tags live on the marketing site but not on the app subdomain where the trial signup completes. Then Google Ads shows demo requests and no signups, and someone concludes search does not drive product-led volume. It does. You just cannot see it.

Step 3: Add server-side tagging and the conversions APIs

Browser-only tracking loses 10 to 30 percent of B2B conversions to ad blockers, Safari ITP cookie truncation and consent rejection. Server-side recovers most of that.

Step 4: Capture click identifiers and persist them through the CRM

This is where most builds quietly fail. Capturing GCLID into a form field is easy. Keeping it attached to the deal 94 days later when it closes is the part that requires design.

If you are running LinkedIn lead gen forms rather than landing pages, the hidden field trick does not apply, so map li_fat_id through the native CRM integration instead. The tradeoff is real: lead gen forms convert better and track worse.

This chain is what makes offline conversion tracking possible at all. Without persisted identifiers you can never feed closed-won back to the platforms, which means smart bidding optimises towards form fills forever.

Step 5: Enhanced conversions, values and consent

Enhanced conversions for leads sends a SHA-256 hashed email with the conversion so Google can match it to a signed-in click even when the cookie is gone. Turn it on. It is free, takes an afternoon, and typically recovers a chunk of otherwise unattributed conversions.

Then set values per stage so bidding has a gradient to climb:

Configure Google Consent Mode v2 and document what you expect to lose. In the EU with a well-designed banner, expect 20 to 40 percent consent rejection and plan for modelled conversions to fill part of the gap. Write that number down so nobody panics at it later.

Step 6: The QA suite. Eight tests, eight expected results

Treat this as a release gate. Nothing launches until all eight pass.

Test	What you do	Expected result
End-to-end test lead	Click a real ad with a test GCLID appended, submit every form, progress the record to closed-won in a sandbox	Conversion appears in all three platforms, identifier present on every CRM object
CRM field mapping audit	Inspect the test record's custom properties field by field	All six identifier fields populated, none truncated or null
Row count reconciliation	Compare platform conversion count to CRM row count for a 30 day window, aligned on click date	Within 5 to 15 percent, gap explained by modelled conversions
Duplicate conversion check	Query CRM for contacts with two or more conversion events within 60 seconds	Zero, or a documented reason such as a double-submit form
Lag window check	Compare your conversion window setting to actual median days from click to closed-won	Window exceeds the 80th percentile of real sales cycle length
Consent mode test	Reject cookies, submit a form, check both browser and server paths	Server event fires with hashed email, browser event suppressed
Deduplication test	Fire one conversion, inspect both browser and server logs	One conversion recorded in-platform, two events sent with matching IDs
Value integrity test	Export closed-won conversions for a month and sum the values	Total matches finance bookings for the same cohort within 2 percent

What this costs, honestly

A full build takes two to four weeks of combined marketing ops and engineering time, plus 50 to 300 dollars a month of hosting. On a 20,000 dollar a month paid programme that is easy to justify. On a 3,000 dollar a month programme, do steps 1, 2, 4 and the QA suite, and skip server-side until spend grows. Half-built server-side tracking is worse than none, because it produces numbers that look authoritative and are not.

The other honest cost: this build makes your reported CPA go up. Deduplication removes phantom conversions, tighter definitions remove unqualified ones, and the number that comes out is larger and true. Prepare leadership for that before you ship, not after.

What to do next

Work the six steps in order, gate the launch on the eight tests, and put tests 1, 3 and 4 on a monthly calendar invite with a named owner. Then go read how this feeds lead quality scoring and pick your stack from the PPC tooling guide. If you are still deciding where the spend goes, settle Google Ads versus LinkedIn Ads first, because the tracking build differs slightly between them, and the rest of the SaaS PPC programme depends on getting this feed right.