Creating demand and capturing it are two different jobs. The first is measured in years and produces no click. The second is measured in days and produces a spreadsheet. That asymmetry, not the relative merit of the two activities, decides most budget arguments about them.
The vocabulary is unhelpful. Demand generation, lead generation, brand, performance, top and bottom of funnel: these labels overlap, come from practitioners rather than research, and are frequently used to mean the opposite of each other by two people in the same meeting.
This page sets out what the two activities actually do, what the experimental evidence says about each, and why the honest answer to “which one” starts with a question about your own market.
Two jobs, not two stages
The framing as a sequence is the source of most of the confusion. They are concurrent activities with different mechanisms.
Capturing demand. Somebody has a problem, has named it, and is looking. Your job is to be present, credible and easy to choose at that moment. Search advertising, comparison content, pricing pages, review sites and referral all do this.
Creating demand. Somebody has a problem and has not named it, or does not know solutions like yours exist. Your job is to make the problem legible and associate your name with it, so that when they eventually look, you are already familiar.
Why the “stages” framing misleads. It implies people move from one to the other on your schedule, in a queue, and that you can therefore build the first and wait for the second. In practice most of your market is doing neither today, and the people who are looking are frequently people you never influenced.
The failure modes are different and both are common. All capture and no creation plateaus: you eventually harvest everything available and growth stops, with no obvious cause on any dashboard. All creation and no capture produces awareness that competitors convert, because you were interesting and somebody else was findable.
Which failure you have is diagnosable. If your search impression share is high and your pipeline is flat, you are out of demand to capture. If your branded search volume is growing while enquiries are not, you are creating demand somebody else is capturing.
What the experiments say about capture
The strongest evidence on demand capture is uncomfortable, and its limit is as important as its finding.
The experiment. A large online marketplace switched off its brand keyword advertising on two search engines while keeping it on a third as a control.
The result. “Almost all (99.5 percent) of the forgone click traffic from turning off brand keyword paid search was immediately captured by natural search traffic.”
What that establishes about capture in general. Capturing demand you already dominate is close to worthless. The demand was going to arrive; the advertising was renting a position you already held organically.
The wider figure from the same study. On non-brand terms, across 68 test markets against 142 controls over 60 days, “the entire regime of paid search adds only 0.66 percent to sales”, with an experimental return of minus 63% where standard regression reported over 4,100%.
The limit the authors state, which is the whole story for a smaller company. “Our results show that for a well-known brand like eBay, the efficacy of SEM is limited at best… This may not be true for small and new entities that have no brand recognition.”
Read that limit carefully, because it inverts the lesson. The reason capture added so little is that the brand had already been created. A company nobody has heard of is in the opposite position: capture is the only thing that works, and it works precisely because there is no organic position to fall back on.
What the experiments say about creation
The evidence here is thinner, because creating demand is far harder to test. What exists is instructive.
The experiment. A national retailer matched its customer database against a large web platform: 1,577,256 individuals, randomised 81% treatment and 19% control, a 14-day display campaign delivering 32.3 million impressions.
The result. “The advertising profitably increases purchases by 5%”, with incremental revenue “more than seven times the cost of the ads”.
The two findings that matter more than the 5%. “93% of the increase occurs in brick-and-mortar stores” and “78% of the increase derives from consumers who never click the ads.”
What that establishes. Advertising that creates or reactivates demand can produce large, real, profitable effects that are essentially invisible to click-based measurement. Not weakly visible. Invisible.
The limit. The population was the retailer’s existing customers, so this is evidence about reactivating a known audience rather than about creating demand among strangers. It is the closest large randomised evidence available, and it is not exactly the case most B2B companies are in.
Why better evidence does not exist. Measuring this properly requires enormous samples. A review of 25 online field experiments concluded that “the required sample size for an experiment to generate sufficiently informative confidence intervals is typically in excess of ten million person-weeks”. The absence of clean evidence about demand creation is a measurement problem, not a verdict.
Where the vocabulary comes from, and why it fights itself
Part of the confusion in this debate is genuine disagreement. Part of it is that the words were never defined by anybody with authority to define them.
These are practitioner terms. Demand generation, demand capture, lead generation: none was established by research, and none has an accredited definition. They spread through conference talks, agency positioning and software categories, which is why two competent people can use them to mean opposite things.
The commercial pressure behind the terms. Software categories need names. A company selling marketing automation has an interest in a vocabulary where nurturing captured leads inside its tool is the central act. A company selling brand consultancy has an interest in the opposite emphasis. Neither is dishonest, and neither is neutral.
What the standards bodies actually say. The one measurement standard that addresses the word “lead” defines it as a targeting audience that has provided no direct indication of purchase intent, and advises against using leads as a success metric where the campaign goal is to prompt action. That is not how the marketing industry uses the word, which tells you how far the vocabulary has drifted from anything normed.
Why this matters practically rather than academically. When a meeting stalls on whether an activity is “demand gen” or “lead gen”, the argument is about labels rather than mechanisms. Reframing to the underlying question, are we reaching people who are looking or people who are not, resolves it in one sentence.
The two questions that replace the vocabulary entirely. Is this person looking for a solution right now? And can I be found when they look? Everything useful in this debate is downstream of those two, and neither requires a label.
The measurement asymmetry, and what to do about it
This is the practical heart of the subject, and it explains a pattern you have probably lived through.
The asymmetry. Capture produces an unbroken chain: click, form, enquiry, all inside a reporting month. Creation produces an effect that arrives later, through people who never clicked, and lands in a channel your analytics calls direct.
What that does in a review meeting. Every channel goes into one table with a cost-per-lead column. Capture wins every row. Creation is cut. Two or three quarters later the pipeline thins, and by then the delay exceeds the memory of the review cycle, so nobody connects the two events.
The wrong fix. Trying to attribute creation. Multi-touch attribution models do not solve this, because the effect they need to capture produced no touch. Building a more elaborate model on missing data produces confident numbers rather than correct ones.
The right fix, and it is organisational rather than technical. Judge the two activities on separate documents, with separate metrics and separate horizons. A creation programme reviewed quarterly on reach, branded search volume and unprompted mentions is being judged fairly. The same programme in a monthly cost-per-lead table is not.
What to actually track for creation. Branded and direct search volume over quarters. Share of enquiries that mention you unprompted, which your sales team can record in ten seconds per call. Total enquiries against total spend over rolling twelve-month windows, ignoring attribution entirely.
The honest limitation. None of these proves causation, and you should say so rather than dress them up. They are consistent with an effect, they are cheap, and they are better than a precise number that measures the wrong thing.
How to allocate between them
There is no universal ratio, and anyone offering you one is selling a framework rather than answering your question.
Start with whether capturable demand exists. Search volume in the words your buyers use, and your current impression share against it. Two facts, both available this week.
If demand exists and you are not capturing all of it. Capture is your entire answer for now. It is faster, cheaper and lower risk than anything else available, and creation can wait.
If you are capturing all of it and still short. This is the case for creation, and it is the only case where it is obviously right. You have hit the ceiling of the existing market, and the only way up is to enlarge it.
If almost nobody is searching. Common on narrow or new categories. Capture has nothing to work with, and the choice is creation or a different go-to-market entirely, which is a real option worth considering honestly.
The proportion question, answered as well as it can be. Once both are running, the split is a cash-flow decision more than a marketing one. Creation is an investment with a delayed and uncertain return; how much of it you can carry depends on your runway, not on a published ratio.
The one rule worth holding. Never fund creation out of the capture budget while capture still has unharvested demand. That trade is almost always negative, and it is the most common way a working account gets broken.
Where to go next
You want the experimental evidence in full. Paid advertising vs organic.
You are splitting brand and performance budget. Brand vs performance in B2B.
You want the stage vocabulary and its limits. TOFU, MOFU, BOFU.
You are choosing which channel captures. LinkedIn Ads vs Google Ads, and Meta Ads or Google Ads.
Your cost per lead is the number in dispute. Cost per lead.
You want to know which metric governs the budget. ROAS vs MER vs CAC vs LTV.
In short
- They are two jobs, not two stages. Capture reaches people who are looking. Creation reaches people who are not.
- Capturing demand you already dominate adds little. A marketplace that switched off brand advertising kept 99.5% of the clicks through organic results.
- That result inverts for a company nobody knows. Its own authors say it “may not be true for small and new entities that have no brand recognition”.
- Creation produces effects invisible to clicks. In one randomised experiment, 78% of the sales lift came from people who never clicked, and 93% happened in physical stores.
- The asymmetry is structural. Capture reports in days on metrics you already have; creation reports in quarters through effects that produce no touch.
- Which is why creation loses budget arguments it should win, and why the pipeline thins two quarters after the decision.
- Do not try to attribute creation. Judge it on separate documents, with branded search volume, unprompted mentions and long-window ratios.
- Never fund creation from the capture budget while capture still has unharvested demand.
Find out whether demand exists before deciding whether to create it. Book a diagnostic, or see how we approach B2B paid acquisition.