The demand capture ceiling

A man is on his hands and knees under a streetlight, searching the pavement. A passer-by stops to help and asks where exactly he dropped it.

“Over there,” he says, pointing into the dark. “But the light is much better here.”

It’s an old joke - it goes back at least to a Boston newspaper in 1924 - and social scientists have used it for decades to describe looking for answers only where it’s easiest to see.

So what has a man under a streetlight got to do with your pipeline report?

Quite a lot. Most marketing measurement lights up one small patch of the market: the buyers who are ready now, clicking, searching and filling in forms. It’s well lit, easy to count and easy to optimise. And for most companies it’s a small fraction of the people who will ever buy.

When the numbers look better than they are

I’ve been under that streetlight more than once.

Paid search is the classic case. You run paid ads for a popular whitepaper, the downloads go up and the total looks great. But some of those people would have found you through organic search anyway, so part of what you’re paying for is demand you already had. Working out what’s truly incremental takes complicated maths and, honestly, smarter people than me.

Free trials are another. As a total they generated a good amount of revenue, and on that view they looked like a winner. But on a revenue-per-lead basis they didn’t stack up as a profitable offer. Same programme, two very different answers, depending on which number you looked at.

Neither of those is a disaster. They’re just what happens when you optimise the patch you can see. So I’ve broken this into three parts: why capture plateaus, how to look beyond the streetlight, and how to build tomorrow’s demand on purpose.

1. Why capture plateaus

Here’s the maths that sets the ceiling.

John Dawes of the Ehrenberg-Bass Institute, working with LinkedIn’s B2B Institute, showed that only around 5% of business buyers are in market for many products and services at any one time. (He’s clear the percentages aren’t meant literally. The point is the gap.)

Demand capture works on the 5%. Optimise it hard and it gets more and more efficient, because you’re getting better at harvesting the same small group. But the group doesn’t get any bigger. Eventually you’re paying more to win the same buyers, or fighting your own organic results for them.

The cod fishermen of the Grand Banks learned a brutal version of this. For years the catch looked healthy while the stock beneath it shrank - by one account, the final years’ catches took around 60% of the population rather than the 16% the plan allowed. In July 1992 Canada closed the fishery after almost five centuries.

Marketing rarely fails that dramatically. But the pattern is the same - a harvest that looks fine right up until there’s nothing left to harvest.

2. Look beyond the streetlight

The dark is getting bigger.

More of the buying journey now happens where your tracking can’t see. G2 found that half of software buyers now start their research in an AI chatbot. 6sense found buyers make first contact about 61% of the way through their journey, and the vendor they favoured before that first conversation wins around 80% of the time. By the time someone fills in your form, much of the decision has already been made out of view.

So the measurement has to stretch. Intent signals at the account level, so you can see interest before anyone raises a hand. Holdout tests, so you know what a programme really adds rather than what it touches. And some humility about attribution, because last-touch credit will always reward whatever sits closest to the streetlight.

None of this is precise. Gartner notes that marketing mix models have less credibility in B2B because of long, variable deal cycles. But a rough view of the whole street beats a perfect view of one paving slab.

3. Build tomorrow’s demand on purpose

If capture harvests today’s demand, something has to plant tomorrow’s.

I studied cognitive science, and my final-year dissertation was on attention and how it shapes decisions. It’s stayed with me, because brand building is mostly about attention and memory. Most buyers aren’t ready today. The job is to be the name they remember when they are. Les Binet and Peter Field’s work for the B2B Institute found B2B businesses grow best with roughly 46% of budget on brand and 54% on activation - because both ends of the funnel need feeding.

Generative engine optimisation is the newest version of this, and it’s how I made the case for it recently.

We’d seen growth slow in one part of the business. Rather than jump to a fix, I laid the foundation first: a detailed list of possible reasons, from market dynamics through to how we were showing up, and I worked with global teams on tests that had already been run in other regions. One of the leading hypotheses that came out was that more of those buyers were doing their research in LLMs, and we needed to improve how we showed up there.

I pulled the data into a compelling narrative and we moved forward with a pilot. It’s too early to share results. But the way the case was made - diagnose, test the hypothesis against what others had already learned, then pilot - is the point. Future demand is easier to fund when it’s framed as a test, not an act of faith.

What good looks like

A few golden rules I use to check whether an engine is balanced:

• Incrementality, not just totals. For your biggest capture channels, you know roughly what would have happened anyway.
• Unit economics as well as volume. Programmes are judged per lead and per dollar, not only in total.
• Signals from outside the form. Account-level intent and AI visibility sit alongside form fills.
• A planned share for future demand. Brand and reach have a budget line, not whatever’s left over.
• Hypotheses before fixes. When a segment slows, the reasons are listed and tested before the money moves.
• Pilots to make the case. Future-demand bets start small, with a clear way to judge them.

For me, yes and no

There’s a fair challenge to all of this, and it usually comes from finance. Capture shows up this quarter. Brand and reach might not show up for a year, and when they do they’re hard to prove. In a year where every dollar is scrutinised, why fund what you can’t measure?

I understand that completely. It’s a reasonable question.

But the answer isn’t to stop measuring. It’s to measure differently - holdouts, pilots, and a narrative that connects the data to a clear hypothesis. And to remember what the alternative looks like. An engine that only harvests will look efficient, right up until there’s nothing left under the streetlight to find.

Where to start on Monday

1. Ask what would have happened anyway. Pick your biggest paid channel and estimate how much of its result was truly incremental.
2. Look at one programme per unit. Take a programme that looks good in total and check its revenue per lead.
3. List the reasons before the fix. Next time a segment slows, write down every possible cause before anyone proposes a campaign.

Where’s your streetlight? I’d love to hear how other teams have found the demand they couldn’t see.

Next: staircase or flywheel? I’ve described the engine as three layers built in order. In practice, they feed each other - and that’s worth arguing about.