Measuring a Funnel You Can No Longer See Into

The visibility gap is now most of the funnel

This isn't the familiar dark social problem in a new outfit — it's larger, because a single interaction can replace an entire research phase that previously generated dozens of trackable sessions across multiple vendors. The buyer arrives already informed, already narrowed, and their analytics footprint is one direct visit. Everything upstream is invisible by construction.

What you can still see

Outcomes, and the shape of arriving demand. Are more people arriving? Are they arriving better informed? Is the shortlist you're on getting shorter or longer? Branded search still tells you whether awareness is growing. First-call quality tells you how well-prepared arrivals are. Win rate against named competitors tells you how you're being framed. None of these traces a path and together they answer the question the path was a proxy for.

• Branded search volume — still the cleanest awareness signal available
• First-call preparedness, scored consistently by sales after every call
• Ask directly: 'did you use an assistant to research this, and what did it say?'
• Periodic manual checks of what assistants say about you in your category

Add the question to your sales process

The cheapest instrument here, exactly as it was for dark social: ask. Add a standard question to discovery — how did you come across us, and did anything you read shape your shortlist? Buyers answer this readily and often mention an assistant summary explicitly, sometimes quoting what it said. That feedback is more actionable than any dashboard, and it costs one question in a call you were already having.

Accept a permanently lower measurement ceiling

The trackable share of your influence has been shrinking for a decade and this accelerates it. The productive response isn't better attribution technology — vendors will sell you that and it won't work. It's building a decision-making culture that can act on partial evidence: triangulating between outcome metrics, direct questions and deliberate holdout tests, and being honest with the board about what's inference rather than measurement.