You Don't Have a Leads Problem. You Have a Demand Problem.

Why performance marketing plateaus, what the effectiveness research actually says, and the two jobs your budget has to do.

There's a moment most growing businesses hit. The ads are working. ROAS looks healthy. The dashboard is green. And yet sales have flattened, cost per lead keeps climbing, and putting more money into the machine returns less than it did last year.

You didn't do anything wrong. You optimised. You cut what didn't convert and doubled down on what did. Finance approved. And the whole time, the strategy had a ceiling built into it.

Here's the ceiling.

Only a sliver of your market is buying right now

At any point in time, your market splits in two.

The buyers who are in-market now. In most categories that's a small minority. The Ehrenberg-Bass Institute's work with LinkedIn's B2B Institute put a number on it — the 95-5 rule: in a typical quarter, around 5% of business buyers are in the market. The other 95% will buy eventually, on their own schedule, when the contract ends or the platform gets outgrown or the truck dies. Consumer categories run the same shape — most category buyers, most of the time, aren't buying.

Everyone else. The much larger group who'll enter the market later, and who are filtering out your ads right now because they don't need you yet.

That's the whole problem with a performance-only budget. Performance marketing doesn't create demand. It captures demand that already exists and steers it toward you instead of a competitor. Econometric work by Dr Grace Kite found that returns on ad spend from online media track the level of demand in the category — when the category's busy, ROAS looks great; when it's not, no amount of optimisation saves it.

So if all your spend goes to capture, you're fishing one small pool and never restocking it. Eventually you run out of in-market buyers to convert. Then you bid harder for the same people, CPA climbs, and the dashboard that was green goes amber. That's not a targeting problem. It's a demand problem.

Buyers pick you before they meet you

The second thing the research is blunt about: most purchase decisions are largely made before the buying process starts.

In 2025, WPP Media and Oxford's Saïd Business School analysed 1.2 million purchase journeys and found 84% of purchases went to a brand the buyer was already leaning toward before they entered the market. Shoppers arrive with a shortlist. Usually a short one. The work that put you on it happened months or years earlier.

The mechanism is boring and powerful: familiarity. People buy what they recognise. And familiarity is worth hard money at the point of conversion — Tracksuit and TikTok's published study of brand awareness against ad performance found brands with 40% awareness converted 48% better than brands with 20% awareness. At 60%, conversion rates nearly tripled.

Sit with that from a media-buying angle. The same performance budget works roughly twice as hard for a brand people already know. Being known isn't the soft alternative to performance marketing. Being known is what sets your performance marketing's exchange rate.

Which is why the numbers keep landing in the same place. Google and Ipsos studied e-commerce brands across 11 markets in 2025 and found the highest total ROI went to advertisers spending roughly half on brand, half on performance. Binet and Field's IPA analysis put it near 60/40 back in 2013. Analytic Partners' ROI Genome shows the knife cutting both ways: brands that slashed brand investment to feed performance saw their performance ROI drop the following year; brands that lifted it saw performance ROI rise. Starve the top of the market and the bottom gets more expensive. Every dataset, same shape.

"Every client we open the books on has the same split: 90% of budget chasing the 5% who are buying this quarter. Nobody's talking to the 95% who'll buy next year, and that's the cheapest audience they'll ever reach."

— Nate, Head of AI-Enabled RevOps

The two jobs

So marketing has two jobs, and they want opposite tools.

Demand creation. Reach the buyers who aren't buying yet — broadly, consistently, and in a way they'll actually remember. This work is emotional and creative because it has to be: people not in the market have zero use for your feature list, and they'll only remember what made them feel something. You're not chasing a click. You're planting the memory that puts you on the shortlist later.

Demand capture. When buyers go looking, be there with the rational stuff — the offer, the proof, the price, the button. Tightly targeted, ruthlessly measured on CPA and ROAS.

Trying to make one ad do both jobs is how you do neither. The IPA data shows campaigns that split the two jobs into separate executions outperform single do-everything ads. Same brand, same look, two different tools — a hammer and a saw, not a hammer-saw.

Two more findings worth stealing while you're at it:

Spend is competitive. Share of voice predicts share of market, and it has since Peckham's Nielsen work in the 1970s: brands that out-shout their market share tend to grow, brands that under-spend it tend to shrink. Your budget isn't just a cost line. It's a position in a contest your competitors are also entering.

Consistency compounds. Ad wear-out is mostly myth — when Analytic Partners tested 50,000 ads for it, they found fourteen. Campaigns get more efficient the longer they run, and System1's analysis of brand consistency found the least consistent brands need to spend nearly double on media to buy the same growth. The urge to relaunch every quarter is the expensive habit dressed up as agility.

Now add the machines

Here's where this stops being a brand-versus-performance debate and becomes the next five years of your marketing.

Your future buyers are increasingly asking AI before they ask Google. Around 45% of consumers already lean on AI tools during buying journeys, and the traffic those tools send converts at multiples of social. The metric quietly replacing click-through rate is reference rate: how often the machines cite you when someone asks the question you answer.

And AI assistants choose the way nervous humans do. They're trained on the web's accumulated judgement, and they weight authority, consistency and third-party validation when deciding which brands to name. Known, credible, consistently-presented brands get recommended. Invisible ones don't — however good the product. We watched this movie with review sites: aggregators were supposed to kill brand by surfacing the objectively best option, and instead the biggest hotel brands doubled their market share right through the TripAdvisor era. People flood toward what they trust. So do the models trained on people.

That means demand creation now has two audiences. Humans, who need to feel something about you before they'll shortlist you. And machines, which need to find you structured, consistent and corroborated before they'll cite you. The second one is answer engine optimisation — AEO — and it's mostly discipline rather than media dollars, which makes it the cheapest brand-building lever most businesses haven't pulled.

Keep it in proportion, though. Even McKinsey's aggressive projections put agent-driven commerce at 10–14% of retail spend by 2030. The rest is still humans, buying the brand they remember. You need both audiences. Which is rather the theme.

What to do about it

Audit your split. Most businesses we open the bonnet on are running 80–90% capture. The research says the ceiling on that arrives fast. Younger businesses genuinely should weight toward capture — but the drift toward it should be a decision, not a default.

Score each job on its own metrics. Capture gets judged on CPA, conversion rate, ROAS. Creation gets judged on awareness, consideration, preference — the numbers that tell you how much demand you're building for next year. Judge everything on last-click and you'll systematically starve next year's pipeline, because last-click overstates the role of clickable activity several times over.

Pick one idea and hold it. The compounding effects — cheaper growth, softer price sensitivity, fatter margins — only arrive for brands that stay on one platform for years. Chopping and changing resets the clock every time.

Get cited, not just clicked. Structure your content, tighten your consistency, build the third-party proof that makes machines confident naming you. Your competitors mostly haven't started. That window won't stay open.

The businesses that grow do both jobs. They capture the 5% who are buying today and they build a memory in the 95% who'll buy tomorrow — for the humans who choose with their gut and the machines that choose like humans. Demand creation and demand capture. It was never one or the other.