It is the most predictable meeting in marketing. Google Ads says the month produced 87 conversions. GA4 says 64. The CRM says 71 new leads — and sales insists half of those were junk anyway. Somebody asks which number is “right,” somebody else suggests the tracking is broken, and an hour disappears into a reconciliation exercise that will repeat, unresolved, next month. Here is the liberating truth behind that meeting: the three numbers are not supposed to match. Google Ads, GA4, and your CRM are three different instruments measuring three different things with three different rulebooks — different attribution logic, different counting units, different time conventions, and different exposure to tracking loss. Expecting them to agree is like expecting a thermometer, a barometer, and a rain gauge to display the same value.
That doesn’t mean discrepancies are ignorable. There is a normal, explainable gap — typically Ads reporting 15–40% higher than GA4 for the same campaigns, and the CRM sitting somewhere between or below depending on lead-path coverage — and there are pathological gaps that indicate genuinely broken tracking. The skill is telling them apart, and knowing which system to trust for which decision.
This guide is the reconciliation manual: the seven structural reasons the numbers diverge, a diagnostic table mapping gap patterns to causes, the thresholds separating “normal physics” from “broken pipes,” which system is authoritative for which question, and the reporting practice that ends the monthly which-number-is-right meeting permanently.
The three systems disagree by design, not by defect. Google Ads counts conversions and credits them to the click date using its own (ads-favoring, modeled, view-through-capable) attribution — its job is feeding Smart Bidding. GA4 counts events across all channels and credits the conversion date with its own attribution and consent-driven modeling — its job is cross-channel comparison. The CRM counts actual people, deduplicated, from every source including ones no tag sees — its job is the truth about pipeline. A 15–40% Ads-over-GA4 gap is normal physics: attribution differences, date conventions, cookie loss, consent modeling, and counting units explain it. Investigate when gaps exceed ~50%, appear suddenly, or the CRM shows leads no system attributes. Decision rule: optimize campaigns on Ads data, compare channels in GA4, judge the business in the CRM — and reconcile monthly with a three-column report instead of arguing about one true number.
Three Instruments, Three Rulebooks
| Google Ads | GA4 | CRM | |
|---|---|---|---|
| Counts | Conversions (per action settings) | Key events (once per event, per session logic) | People / records, deduplicated |
| Attribution | Own model, ads-only view of the journey; includes view-through & engaged-view for some formats; modeled conversions fill consent/cookie gaps | Cross-channel data-driven (or last-non-direct); no view-through for paid search; own modeling | Usually first-known-source or self-reported; often none |
| Date convention | Click date — a conversion 12 days after the click reports on the click’s day | Event date — the day the conversion happened | Record creation / stage-change date |
| Sees | Only journeys touching its ads | Tagged website/app activity that consent & browsers allow | Everything that becomes a record — including calls, walk-ins, referrals no tag saw |
| Built to answer | “What should bidding buy more of?” | “How do channels compare?” | “What actually entered the pipeline?” |
The Seven Structural Reasons the Numbers Diverge
- Attribution scope. Ads happily claims a conversion that GA4’s cross-channel model credits to the organic visit that closed the journey. Neither is lying; they answer different questions about the same journey. This alone commonly accounts for the largest slice of the gap.
- Date conventions. Ads back-dates conversions to the click; GA4 stamps the conversion day. Any month boundary slices the same conversions into different months in the two systems — and makes week-over-week comparisons structurally unstable in Ads for long-lag services (recent days always look worse until lagged conversions arrive).
- Counting units. An Ads action set to “every” vs GA4 counting per-session, vs a CRM deduplicating the same human who submitted twice — three defensible counts of one person’s enthusiasm. Counting settings are half of every “Ads shows double” mystery, covered in our conversion actions guide.
- Modeled conversions. With Consent Mode active, both Google systems statistically model conversions for unconsented traffic — but they model independently, with different methods, and the CRM contains only real records. Modeling is a feature, not a bug; it just guarantees non-matching totals.
- Tracking loss asymmetry. Ad blockers and browser tracking prevention hit GA4’s collection harder than Ads’ conversion measurement (especially with enhanced conversions recovering matches). The CRM loses nothing to browsers — people who call still exist — but gains records no web analytics saw, in both directions.
- Lead-path coverage. The CRM ingests phone calls, emails, chat, referrals, and walk-ins. Unless call tracking and every intake path write source data into records, the CRM will always contain leads that no digital system attributes — the gap that makes paid channels look worse than they are in CRM-source reports.
- Junk and duplicates. Spam form fills inflate Ads and GA4 equally but get deleted or merged in the CRM. Sales’ “half of these were junk” is a fourth number — qualified leads — and it’s usually the one the business should actually care about.
Because Ads back-dates conversions to click dates, the most recent 7–14 days are always incomplete in Ads — conversions from those clicks are still arriving. Comparing systems (or judging performance) on a window that hasn’t closed guarantees phantom ‘drops’ and false alarms. The professional habit: reconcile and evaluate on a window lagged by at least your typical click-to-lead time (for most service businesses, compare the month once it’s ~2 weeks old), and annotate dashboards so stakeholders stop panicking at the perpetually-soft recent edge.
Normal Physics vs Broken Pipes: The Diagnostic Table
| Pattern | Likely cause | Verdict |
|---|---|---|
| Ads 15–40% above GA4, stable over months | Attribution scope + date convention + modeling + counting | Normal — document it, stop relitigating it |
| Ads ≈ 2× GA4 (or worse), stable | Duplicate conversion actions (tag + GA4 import both Primary), or “every” counting on lead forms | Broken — deduplicate the conversion architecture |
| Sudden gap change with no spend change | A tag broke, a consent banner update blocked firing, a form redesign killed a trigger, or an action was reclassified | Broken — check change history & tag firing first |
| GA4 far above Ads for paid campaigns | Auto-tagging off / broken GCLID passthrough, redirects stripping parameters, cross-domain misconfig | Broken — fix tagging plumbing |
| CRM well below both | Junk/duplicate inflation upstream, or leads dying between form and CRM (integration failures, unrouted notifications) | Investigate both — the second cause loses real revenue |
| CRM well above digital systems | Untracked lead paths (calls, chat, email) — measurement coverage gap, not tracking failure | Fix coverage: call tracking & source capture per path |
Which System to Trust for Which Decision
- Bid strategy, budget by campaign, keyword decisions: Google Ads data — it’s the only view Smart Bidding shares, and the only one with click-date alignment to spend. Optimizing Ads against GA4 numbers is steering one machine with another machine’s speedometer.
- Channel mix, landing page behavior, funnel drop-off: GA4 — the only instrument that sees all channels under one attribution rulebook. Its per-channel absolute numbers are understated by tracking loss, but understated consistently, which is what comparison requires.
- Is marketing working, cost per qualified lead, cost per customer, revenue: the CRM — the only system counting real deduplicated humans and their outcomes. This is the layer where closed-loop tracking pays off: import CRM outcomes back into Ads and the bidding instrument starts learning from the truth instrument.
Most gaps are measurement philosophy. One is operational bleeding: leads that convert on the website but never become CRM records — integration failures, forms emailing an unmonitored inbox, chat leads nobody transcribes, webhook errors silently dropping submissions. These aren’t attribution differences; they’re paid-for prospects evaporating before sales ever sees them. Audit it directly once a quarter: pull a day’s raw form submissions from the form tool itself and trace every single one to a CRM record within the expected SLA. Any leakage found here outweighs every philosophical Ads-vs-GA4 debate on this page — fix it first.
The Reporting Practice That Ends the Argument
Replace the one-true-number hunt with a standing three-column reconciliation, refreshed monthly on a lagged window: Ads conversions (with the conversion actions listed), GA4 key events, CRM new leads and qualified leads — plus the computed ratios between them. The first month, document why each gap exists using the seven reasons above; thereafter, the ratios themselves become the monitoring instrument: stable ratios mean healthy tracking regardless of absolute disagreement, and a moving ratio is the alarm that something changed. Add two derived metrics that only the combined view can produce — cost per CRM-qualified lead by campaign (Ads spend joined to CRM outcomes) and lead-path coverage (share of CRM records with known source) — and the meeting that used to argue about numbers starts allocating budget instead. Guard the practice against the classic trap of celebrating whichever system flatters this month’s narrative: pick the decision-to-system mapping above once, write it down, and let qualified outcomes, not the biggest available number, define success.
5 Common Reconciliation Mistakes
- Judging this week’s Ads performance. Click-dating makes the recent edge perpetually incomplete; the “drop” refills itself within days.
- “Fixing” the gap by importing GA4 conversions into Ads as Primary. Now the same leads count twice and the gap you closed reopens as double-counting.
- Treating GA4 as the referee. It’s a third instrument with its own losses and models — useful for comparison, not arbitration.
- Letting the CRM’s source field rot. A truth-layer where half the records say “unknown” can’t settle anything; source capture per lead path is what makes the CRM authoritative.
- Re-deriving the explanation every month. Document the structural gap once; monitor the ratio thereafter. The recurring debate is a process failure, not a data one.
Frequently Asked Questions
What's a 'normal' discrepancy between Google Ads and GA4 conversions?
For the same campaigns over a closed (lagged) window, Ads reporting 15–40% more conversions than GA4 shows as key events is the common, explainable range — driven by attribution scope (Ads credits journeys GA4 assigns elsewhere), click-date vs event-date conventions, independent conversion modeling under consent mode, counting-unit differences, and GA4’s higher exposure to blockers. The exact figure varies with your sales-cycle length, audience consent rates, and traffic mix, so benchmark your own stable ratio rather than chasing a universal number. The actionable rule: a stable ratio — whatever it is — means healthy instrumentation; a ratio outside roughly 1.1–1.6, or one that moves sharply without a corresponding configuration change, earns an investigation starting with duplicate conversion actions and counting settings on the Ads side, and tagging/consent plumbing on the GA4 side.
Our CRM shows fewer leads than Google Ads reports. Is Google inflating numbers?
Usually it’s three mundane causes stacked, none of them inflation. First, counting units: Ads may count the same person’s repeat submissions (check ‘every’ vs ‘one’ counting) while the CRM deduplicates humans. Second, junk: spam and bot form fills convert in Ads’ eyes and get deleted at the CRM door — a reCAPTCHA/honeypot gap, not a reporting one. Third — and the one worth losing sleep over — leakage: real submissions that never became CRM records due to integration failures or unmonitored intake paths. Distinguish them by tracing a sample day of raw form-tool submissions record-by-record into the CRM: dedupes and junk will be visible and benign; missing legitimate leads are an operational emergency. Genuine Ads-side inflation does exist in one specific configuration — duplicate conversion actions both set to Primary — and takes two minutes to check in the conversions settings.
Should I just pick one system and ignore the others?
No — each system is irreplaceable for its decision and misleading outside it. Run Ads optimization on Ads data (it’s the only view aligned with what Smart Bidding sees and spends against); compare channels and diagnose on-site behavior in GA4 (the only cross-channel rulebook you have); and judge business results — qualified leads, customers, revenue by campaign — in the CRM, ideally with that truth fed back into Ads via offline/enhanced conversion imports so the optimization layer learns from it. What you should pick once and enforce is the mapping of decisions to systems, written into your reporting so nobody cherry-picks the flattering number. The teams that ‘solve’ discrepancies by standardizing on a single system invariably standardize on the one that answers their least important question.
Why did our Google Ads conversions drop 30% this week when nothing changed?
First suspect the calendar, not the campaigns: Ads reports conversions on the click date, so the most recent days are always incomplete — a click from Tuesday that converts next Monday will retroactively appear on Tuesday. For businesses with multi-day consideration, the trailing 7–14 days perpetually look 20–40% soft and then ‘recover’ as lagged conversions land. Check whether the drop persists on a window older than your typical click-to-lead lag; most weekly panics dissolve there. If a genuinely closed window shows the drop, then investigate in order: change history (did someone edit conversion actions, goals, or consent settings?), tag firing on the conversion page (form or thank-you page redesigns are the classic silent killer), consent banner updates blocking tags, and only then market factors. Sudden measurement changes almost always have configuration causes; gradual ones have market causes.
Is there a way to make all three systems match exactly?
No — and vendors or agencies promising it are proposing to break something. The systems differ by architecture: attribution scope, date conventions, modeling, deduplication, and coverage are design choices, not bugs, and ‘fixing’ them means either double-counting (importing one system’s conversions into another as additional Primaries) or discarding capability (turning off modeling and enhanced measurement, which blinds bidding to recover a cosmetic match). The achievable and genuinely valuable states instead: documented gaps (each ratio explained once, in writing), stable ratios monitored as tracking-health alarms, full lead-path coverage so the CRM sees a source for everything, zero form-to-CRM leakage, and a closed loop where CRM outcomes flow back into Ads for bidding. That stack gives you numbers that disagree predictably and inform correctly — which is what measurement is actually for.
Tired of the monthly which-number-is-right meeting?
We’ll audit all three layers — conversion actions, GA4 configuration, and CRM source capture — document your normal ratios, fix the genuinely broken pipes, and build the reconciliation report that turns three disagreeing numbers into one coherent decision system.
Get a Measurement Audit Explore Lead Generation