A 2019 Marketing Science study covering 15 randomized Facebook experiments and 1.6 billion impressions found that observational attribution methods overstated advertising's effect compared to randomized lift tests. That gap between what platform dashboards report and what ads actually cause is the measurement problem most B2B teams are living inside right now, whether they've named it or not.

The problem gets worse when your highest-value accounts don't click. Buying groups at mid-market and enterprise accounts often include six to ten decision makers, according to research summarized by Gartner. The person who sees your LinkedIn Sponsored Content is rarely the person who fills out the demo form. Click-based attribution treats that impression as waste. Account-level pipeline data often tells a different story.

Why View-Through Attribution Isn't the Answer (by Itself)

The reflexive fix is to turn on view-through attribution and count it as pipeline credit. Don't. VTA records that an impression preceded a conversion. It doesn't prove the impression caused it. People already likely to convert are also more likely to be served ads (selection bias), which inflates perceived ROAS. Enalitica's analysis puts it bluntly: treating VTA as causal can mislead stakeholders and distort budget decisions.

That doesn't mean VTA is useless. Labeled clearly, with a defined lookback window and separated from click-through numbers in every dashboard, it's a directional exposure signal. But if your ops team doesn't document the window, the deduplication logic, and the label conventions, leadership will read VTA numbers as revenue claims. They're not.

The Measurement Stack That Actually Holds Up

If clicks undercount influence and VTA overcounts it, you need a stack that triangulates.

Layer 1: Incrementality testing. Holdout or geo-lift experiments are the defensible way to answer whether your ads created additional pipeline. Success = statistically significant lift in opportunity creation. Guardrails = monitor branded search and direct traffic in both groups for contamination. Stop-loss = if the holdout group outperforms the exposed group by week three, kill the test and investigate. The trade-off: you need enough volume to reach significance, and B2B cycles are long. A geo-lift can work with smaller budgets but introduces regional confounders. Don't force a holdout if you only have 200 target accounts.

Layer 2: Account-level pipeline progression. Match ad exposure data (by company, not by individual) to CRM stage movement. You're looking for whether exposed accounts progress faster or at higher rates than unexposed accounts. This is still correlational, not causal, so label it that way. When it aligns with incrementality results, it builds a credible narrative.

Layer 3: Buyer-reported source. Add a free-text or dropdown "How did you hear about us?" field to your demo request form. It's imperfect, biased toward recent touchpoints, and it will still surface channels your attribution model can't see: podcasts, Slack communities, peer recommendations. Much of B2B preference formation happens in places no analytics tool can observe, and buyer-reported data is one of the few ways to recover that signal.

Layer 4: Branded search and direct traffic trends. A sustained increase in branded search volume among your target segments, correlated with campaign flights, is a leading indicator that awareness programs are working. Not proof of causality, but when the trend moves with your spend and flattens when you pause, the signal is directional.

The Platform Reporting Trap

Platform metric definitions change, and those changes break trendlines. LinkedIn recently updated in-feed video impression and click reporting to align with MRC standards, counting impressions and clicks only when playback starts. That shift can reduce reported impressions, clicks, and CTR while increasing CPM and CPC, with zero change in actual campaign performance. If your QBR compares pre-change and post-change periods without a like-for-like adjustment, you'll misread the data. Maintain a change log. Annotate your dashboards when a platform updates its methodology. Boring, unglamorous ops work, and it's the difference between a credible performance narrative and a panicked budget conversation.

What to Run This Week

Label VTA as "exposure-correlated," not as pipeline credit. Then pick the incrementality test design that fits your volume (account holdout if you have 500+ target accounts; geo-lift if you don't). Set a four-week window, instrument your CRM for account-level exposure matching, and define your success metric before you launch.

Clicks still matter for diagnosing creative resonance and immediate response. The problem was never clicks themselves. It was treating them as the only evidence that your marketing worked, and then watching the budget follow the measurement instead of the outcome. The accounts that never clicked might already be on their way to you. The question is whether your reporting can see them before someone decides to cut the program that brought them there.