B2B Buying Decisions Are Moving Where Ad Tech Can't See

A campaign that attributes 50 conversions may have driven only five, according to an AdExchanger "Data-Driven Thinking" column by Margarita Savytska — because incrementality testing shows most tracked buyers would have converted anyway. That gap between what attribution credits and what actually influences decisions is becoming ad tech's most consequential blind spot.

The industry built precise measurement: every impression counted, every click tracked, every conversion attributed. But that infrastructure doesn't measure influence. As B2B buying shifts into channels ad tech can't observe — AI-assisted research, private peer conversations, dark social sharing — the share of decisions happening in invisible channels grows while the measurable share shrinks.

What does the measurement blind spot look like?

Consider the buyer who asks an AI assistant for a vendor recommendation, gets an answer that skips your brand, forms a shortlist without visiting any website and fills out a competitor's form. That buyer is invisible to every measurement system the industry has built.

Then there's the buyer who receives a forwarded email from a colleague — "This is the company I mentioned" — clicks through and converts. The dashboard logs a direct visit with no campaign attribution, while the most influential touch, the peer recommendation, happened where no platform can see it. Attribution distributes full credit across a partial journey: visible touch points get overcredited, invisible ones get zero.

The takeaway: impressions are measured, decisions aren't.

Why does incrementality beat attribution?

Incrementality measures actual impact — the difference between what happened with the marketing activity and what would have happened without it. Savytska's example: a campaign reaching 1,000 already-in-market buyers that attributes 50 conversions. An incrementality test might reveal 45 of those 50 would have converted regardless, putting the real contribution at five.

The tools exist — holdout groups, geographic splits, matched market analysis. The industrywide commitment to use them as the primary measure doesn't, and the reason is blunt: incrementality almost always produces lower numbers than attribution, and lower numbers are harder to report.

"The incentive structure rewards the metric that makes performance look best, rather than the metric that reflects reality."

In plain terms, the bias compounds over time and quietly reshapes the budget toward measurable channels, whether or not they're the influential ones.

How does AI widen the gap?

When a buyer researches a purchase through an AI assistant, the research happens inside a conversation no ad tech platform can observe. No impression was served, no click happened, no cookie was set — yet the AI's recommendation shaped the shortlist based on content the buyer never visited.

As AI purchasing agents begin evaluating B2B vendors, the top of the funnel moves into a channel current measurement can't capture. The industry that built its value proposition on measurability now faces its most consequential buyer interactions being unmeasurable by existing infrastructure.

What should measurement look like instead?

Savytska proposes a paradigm built around decisions rather than impressions:

• Self-reported attribution — a free-text "How did you hear about us?" field at every high-intent conversion point
• Incrementality as the primary metric for paid activity
• Pipeline and revenue as anchor metrics — did the activity produce pipeline, and did it close?
• Buyer research as input — post-deal interviews and win/loss analysis, which she calls the highest-fidelity and most underused measurement available

Solving measurement for decisions, not impressions, is the harder problem — and, as the buyer's journey moves beyond the reach of today's infrastructure, the only one that matters.