Google Ads says your campaign returned 5x. Salesforce says 2x. The instinct is to trust the CRM because it's "your" data, mark the campaign as mediocre, and trim the budget.
The platform's 5x is inflated. But the CRM's 2x isn't truth either. The gap between them isn't fraud. It's two systems answering different questions badly.
Why Both Numbers Point the Wrong Way
Ad platforms count generously by design: view-through conversions, modeled conversions from users who never granted consent, a conversion window that credits a click from three weeks ago. Platforms grade their own homework, then set your budget based on the result.
Your CRM does the opposite. Most backend revenue reporting is last-click or close to it. The sale gets stamped to whatever the buyer touched last, usually a brand search or direct visit. The paid click that started the journey three weeks earlier gets zero credit.
Same ambiguity, opposite defaults. The platform books the assist as a win. The CRM books the assist as nothing. The 5x-to-2x gap is mostly the size of that disagreement. Averaging them doesn't give you the fever. It gives you a number that describes nobody.
The Channels Your CRM Can't See
The under-crediting scales with distance from the click, which means impression-based channels get quietly robbed. Someone scrolls past a Meta ad, doesn't click, searches your brand three days later, and buys. Your CRM credits brand search. The impression that created the demand gets nothing because there was no click for a last-click system to record. To your CRM, that Meta campaign looks like it did nothing at all.
iOS ATT made it worse. Privacy changes reduced deterministic attribution signals by an estimated 35–50% on iOS, with global ATT opt-in stabilizing around 29%. Even the social clicks that do happen lost their match back to the purchase.
Search sits closer to the purchase click, so last-click backends capture more of what search actually did. The channel getting destroyed by last-click logic is often the one generating demand upstream, not the one capturing it at the bottom. "Cut what doesn't convert" becomes "cut what you can't measure."
Reconciliation Doesn't Fix This
The standard fix is to reconcile: blend sources, or move to data-driven attribution and let the model split credit. But DDA still comes from the same platform whose top-line number you already don't trust. It reconciles. It doesn't measure.
Run Google and Meta together, pull each platform's reported conversion revenue for the same period, and add them up. For most accounts at scale, the sum exceeds what your CRM says you actually made. The same sale gets booked twice: Google logs the click conversion, Meta logs it on a view-through basis. Neither can see the other.
Multi-touch attribution answers "which touch gets credit" with more granularity and better-looking charts. It never answers "would this have happened anyway." The most sophisticated version of the wrong question is still the wrong question.
The One Measurement That Answers "Would This Have Happened Anyway"
You get that number by turning the channel off somewhere and watching what backend revenue does without it. That's incrementality: the revenue that exists because the ads ran and wouldn't exist otherwise.
A practical starting point: geo holdout. Hold the channel out of comparable markets for a full purchase cycle (four weeks minimum for most B2B SaaS). Measure the delta in backend revenue between held-out and live regions.
The hypothesis, stated so it's falsifiable: if we pause Meta spend in Region B while keeping it live in Region A, branded search volume and pipeline in Region B will decline measurably within 30 days, confirming Meta impressions were generating demand the CRM attributed elsewhere. Success = statistically significant lift in pipeline between test and control regions. Guardrails = pipeline velocity doesn't collapse beyond the expected dip. Stop-loss = if total pipeline drops more than 20% in the holdout, restart spend.
What to Measure (and What Not to Over-Interpret)
Platform metrics like CTR and CPC are diagnostic signals for creative and landing-page performance. They're weak predictors of pipeline quality. Cost per SQL and revenue-linked attribution tell you more. In long-cycle B2B, evaluate paid performance with an economic lens: LTV:CAC and CAC payback period, not first-touch ROAS alone. Segment by channel, persona, and region. Blended averages hide where conversion is actually strong or weak.
The uncomfortable truth for demand gen operators in September 2026: the measurement environment keeps getting less reliable as privacy constraints reduce observable data and AI optimizes from partial signals. The fix isn't a better dashboard. It's a first-party, consented data foundation, rigorous data quality checks, and the willingness to turn something off to find out what it actually did.
Every other number is a more expensive guess.
