Here’s a wild one. Last month, I split a portfolio company's funnel by traffic source. Direct traffic that reached the first gate abandoned at 71%. The blended funnel abandoned at 38%. Nobody had looked at the two numbers separately before.

One company, one split. I'm not claiming it's a benchmark. I'm claiming it's the first thing to check as I'm running answer ownership audits across dozens of companies right now, seed-stage to public, and the same mechanism is at work in every one: as AEO improves, we see changes in direct traffic, referrers disappear, and the funnel data gets less reliable. Whether your own gap is 30 points or 10, the direction is the same. The person typing your name into a clean browser is coming in so much more informed than they were even two years ago.

Below is what's causing it, and the fix we're deploying with companies like Consensus: a self-guided demo on the main path, and a follow-up system that runs off the engagement data it produces. The build, the data breakdown, the routing rules, and a 30-day rollout are all here. The last section covers the part almost nobody does: feeding the demo data back into the AI answers that sent the buyer to you in the first place.

Consensus sponsors this piece. I turn down most sponsors. The ones I take are tools we run in our own client work or that I'd recommend anyway, and Consensus is both.

5 minutes in ChatGPT. A Reddit thread from someone who churned off your competitor. A YouTube review from a guy with 900 subscribers. A Perplexity comparison table your marketing team has never seen and can't edit. By the time a buyer lands on your site, they've done more diligence than your SDRs do before a discovery call.

They arrive wanting one thing: to check that what the machines told them is real.

What they get is a hero headline explaining a category they already understand, a tour of a problem they already have, and a gate at the bottom: a calendar link, four days out, behind a form with nine fields, one of which is "How did you hear about us," which no buyer answers honestly. The highest-intent visitors you have hit that gate and leave. The blended funnel averages it away. The conversion rate optimization team ships a headline test above the calendar link, gets a small lift, and posts the win in Slack.

For buyers who arrive already decided, the first gate is the funnel. Everything above it is for people who haven't decided yet, and fewer of your visitors are those people every quarter.

If you're self-serve, the answer is short: working product above the fold, one signup field, tour deleted, and time to first real action on the board. Most PLG companies already know this and have drifted away from it.

If you're enterprise, you can't open the product. Multi-tenant data, security posture, implementation complexity, all real. You can open a demo of it, on the buyer's screen, right now. This is demo automation, it's a Consensus deployment you can run, and here's the build:

Security review, procurement, legal, the MSA, all of it still happens. It happens after they've touched the product, not as a condition of reaching it.

That fixes the abandonment problem. It's also the smaller half of the value. The larger half is the data the demo hands back.

One note on tooling first as it decides everything below: we standardized on Consensus because of what happens after the demo gets shared. Plenty of tools can show the product. Buying group discovery, the Level 2 data described next, only exists on a platform built to track who the demo travels to once your champion starts forwarding it.

Every self-guided demo view produces a record of what the buyer did, and it comes back at three levels. Most teams only ever look at the first one.

Level 1: the person. For every individual viewer: which chapters they selected, which they skipped, completion depth per chapter, what they rewatched and how many times, total time in the demo, when they watched (an 11pm session is a signal on its own), and whether they came back. Consensus's data across 6M+ buyer interactions shows how selective viewers already are: the average demo runs about 15 minutes, and buyers watch about 5. They were always going to choose. Chaptering means you can see what they chose. Read together, that's a list of the buyer's priorities, produced before your first human interaction. A buyer who selects integrations, watches security twice, and skips the overview has told you the shape of the deal before anyone says hello.

Level 2: the deal. This level opens up the moment the demo gets shared, and it gets shared, because it's the easiest artifact in your company to forward. Now you see: who your champion sent it to, each new viewer's role and domain, the order viewers arrive in, how many people from the account have touched it, which chapters the group has collectively covered, and which they haven't. Consensus calls this Demolytics. In practice it's the org chart of the buying committee, assembled as the deal moves.

Every forward is discovery you didn't have to run. No forms, no "who else should be involved?" and the partial answer it gets. Your champion maps the committee for you, one send at a time, and the demo reports back with names, roles, and what each person cared enough to watch. And these forwarded viewers are more engaged, not less: they engage at 61%, more than double the 28% rate of the original recipient, and about four days sooner. Enterprise deals usually stall with people the vendor never met. This is how you meet them.

Level 3: the market. Aggregate chapter behavior across every deal and you have market research your buyers ran for you. Which chapters get selected most is demand. Which get skipped most is material you're still paying to produce. Which get rewatched most is your actual differentiator, whatever the positioning doc says. Where pricing sits in the selection order tells you how your category is being bought. And the difference in selection patterns between closed-won and closed-lost is a win-loss study that doesn't depend on anyone agreeing to an exit interview.

Every buyer-data source you pay for is useful for something and fails at something specific:

First-party. Person-level. Behavioral. Voluntary. Pre-first-call. Each of the other sources clears one or two of those. Demo engagement clears all five, and the only cost of collecting it is dropping a gate your best traffic was already refusing.

Many teams buy the tool, watch the dashboards, and change nothing. The plays below follow the three levels: what you do with a person, with a deal, with the market.

With the person

Route on behavior, not titles. Write tiers into your speed-to-lead rules and put SLAs on them. Here's our starting version; tune it to your motion, but write it down:

Tier 1: watched 60%+, or shared the demo internally, or two-plus viewers from the same domain. A human reaches out within four business hours. Alert fires to a Slack channel the moment the threshold trips.

Tier 2: partial view. Follow-up within one business day, built around the one chapter they finished, and only that chapter.

Tier 3: bounced inside two minutes. Standard nurture. Don't spend AE time here.

The outreach references the behavior. "Saw you spent most of your time in the integrations chapter. Want 15 minutes with the engineer who built it?" outperforms "just checking in" because it's specific. Most intent data is inferred. Demo engagement is what the buyer actually did, and it deserves different rules than a badge scan.

Open the first call where they left off. Pull the watch report before every demo-sourced meeting and make it a required prep step, the same way reps check LinkedIn. Paste it into the opp record so everyone who touches the deal reads the same behavior. The AE opens on the most-watched chapter, skips the twenty minutes of "so tell me about your stack," and retires the generic discovery deck for these meetings. Then carry the data downstream: follow-up subject lines named after the chapter they finished, assets ordered in their watch order, not your pitch order.

With the deal

Bring in the right people the moment a new viewer shows up. Map role-matched assets in advance so the response is same-day: a security viewer gets the documentation pack and a direct line to your architect. An exec who skimmed ROI means your champion gets the business case one-pager before they're asked for it. An end user deep in one workflow gets the tactical guide for that workflow. Every new viewer is a thread, and the watch data tells you which asset each thread needs.

Treat coverage gaps as the to-do list. The absence of a viewer is data too. Stage 3 with no one from security in the watch log means security review is going to surface in month two. Send the security chapter to your champion now, with a note about who it's for. The data tells you what's coming, not just what happened.

Arm the champion, then watch the room. The hardest selling in enterprise happens in meetings you're not in. Ask your champion who's going to be in that room, then build them a share link with chapters ordered for exactly those people. Who opened it, what they watched, what they skipped: that's the readout of an internal meeting no vendor gets invited to.

Forecast off engagement, not optimism. Add three fields to the CRM: viewer count, chapter coverage, days since last view. Make them stage evidence. A deal marked "buying committee engaged" with one viewer isn't. A rewatch spike across multiple viewers usually means an internal meeting is coming, so that's when your champion gets fresh material. Silence for ten days after an internal share usually means the deal stalled in a room you're not in, so that's the trigger for a check-in.

With the market

This is the part almost nobody does, and it's where the demo data connects back to why buyers showed up already decided in the first place.

Feed it back into AEO. The chapters a buyer selects are the questions they were asking the AI before they arrived. That makes monthly chapter selection data the best question-mining source you have: it's pulled from people who got far enough to want proof, it's ranked by what they chose rather than what they typed, and it's yours alone. Export it monthly, rank it, and treat the top ten as next quarter's answer and content roadmap.

Then close the loop in the other direction. For each of those ten topics, check what the answer engines currently say about you. If buyers keep selecting the integrations chapter and the AI answers about your integrations are thin, wrong, or cite a competitor, that's the highest-value answer gap on your list, because you already know buyers care enough to verify it. If a chapter is heavily selected and the AI answers are accurate and cite you, that topic is working; protect it and move on. Run the check against every top-ten chapter and you have an AEO priority list built from demo behavior instead of keyword volume.

The two systems feed each other. Better answers upstream mean more buyers arriving decided, which means more demo views, which means more chapter data, which tells you which answers to fix next. The demo tells you what the market wants to verify. Your AEO work makes sure you're the one they verify it with.

Let the rewatch data rewrite your positioning. Your most-rewatched chapter is what the market finds hardest to believe and most wants to be true. Lead with it, on the homepage, in the first call, in the AI answers. Your most-skipped chapter is either a message problem or a feature nobody buys for. Either way, stop opening with it.

Mine it for win-loss. Compare selection patterns across closed-won and closed-lost. If lost deals disproportionately entered through pricing and never touched product chapters, that's a packaging problem, not a sales-execution problem. This analysis takes one afternoon a quarter and replaces the win-loss interviews you rarely get.

Marketing owns the messaging. Product owns the tour. Sales owns the gate. Nobody owns the distance between a buyer wanting the product and touching it, which is where your best traffic goes. It stays unowned until someone with a revenue number claims it. I'd claim it. It's cheaper than any other pipeline you'll build this year.

Filter to direct traffic. Measure drop-off at your first gate. Compare it to blended. If your direct number looks like mine, your buyers already made up their minds somewhere else and your homepage is still trying to talk them into it.

Send me what you find, the ugly version. Nobody's number is going to be flattering, including mine, and that's the point. I'm anonymizing the patterns and publishing them, so the one split at the top of this piece stops being an anecdote and starts being a benchmark.

One more thing on Consensus. I've spent a lot of hours with that team over the last few months, and the thing I keep noticing is that they get more excited about the watch data than the demo itself. Most demo vendors are proud of the video. These folks are proud of the log. The Level 2 and 3 breakdowns above came out of those conversations, and I'm stoked on how they're continuing to push this market forward.