research
Two ways people buy from an ad: 7 minutes or 3 days
Summary: We reconstructed 51 individual purchase journeys node by node from paid-social traffic. They did not form one funnel with one drop-off pattern. They formed two clearly separated systems: an express path where the ad did the persuading and the buyer purchases in about 7 minutes through land → CTA → product → checkout, and a depth path that runs long-form content and a quiz, often across multiple sessions and days. The two paths want opposite things from a landing page, and a page optimised for one actively obstructs the other.
Standard funnel thinking treats a conversion path as one shape with leaks. Fix the leaks, get more conversions. That model did not survive looking at actual journeys.
Method, and its main limitation
We reconstructed every node — page view, CTA click, quiz step, offer arrival, checkout — for 51 converters in a paid-social campaign: 29 who first saw the control arm, 22 who first saw the test arm, and 1 dual-arm mismatch. Journeys came from a queryable event store, ordered per user.
The limitation is important and we will state it before the findings: this is a converter-only sample. It describes how purchasers finished. It cannot be used to compute population conversion rate, and the paths here are not evidence about how often each pattern occurs among all visitors. It is a study of shape, not of rate.
Supporting rate figures below come from the full campaign cohorts, which are stated separately.
The two systems
The express system. The ad did the persuasion. The visitor arrives already sold and wants the shortest path to a purchase. Median new converter on this path buys in about 7 minutes, moving land → CTA → product → checkout. No quiz, no long-form reading, no return visit.
The depth system. The visitor arrives curious rather than convinced. They read long-form content, complete a quiz, arrive at an offer — often across multiple sessions spanning days. The content manufactures the intent the ad did not.
These are not two segments of one funnel with different speeds. They are different mechanisms. The express buyer's intent was created upstream in the ad; the depth buyer's intent is created on the page.
The attention gap between them
The rate evidence for the depth system comes from the campaign cohorts, and it is the most decision-grade finding we have from this work.
Time on page, holdout (long-form presell) versus test (short pages):
| Reached 2+ minutes | Left under 10 seconds | |
|---|---|---|
| Long-form holdout | 68 / 264 = 25.8% | 105 / 264 = 39.8% |
| Short test pages | 20 / 349 = 5.7% | 217 / 349 = 62.2% |
The long-form pages reached 2+ minutes of attention at about 4.5× the rate. Median dwell on those pages was 1,065 seconds against 301 seconds.
The conclusion we drew, which we did not expect: attention is manufactured by narrative, not demanded by design. Our short pages were not failing to be persuasive — 62% of visitors left before persuasion was possible. There is no CTA placement that fixes a 10-second exit.
Quiz completion tracked the same split: 9.33% on the long-form holdout against 4.35% on the short pages.
The empty-clicks finding
Here is where the two systems produce a genuinely misleading measurement.
Upstream, our short pages looked like they were winning: 35 CTA clicks against 23 quiz completions on the comparison arm. More people were taking the primary action.
By checkout, the advantage was gone: 16 of 333 (4.80%) versus 17 of 345 (4.93%).
The short pages do not fail to make people move. Their apparent upstream advantage simply vanishes downstream, because a CTA click from an unconvinced visitor is not the same event as a CTA click from a convinced one. They are identical in the analytics and different in every way that matters.
If your landing page optimisation stops at click-through, you will systematically prefer pages that produce empty clicks. This is the same trap as optimising a 50%-click / 4%-buy page over a 30%-click / 10%-buy page — you can measure your way into it while every dashboard reads green.
What the funnel numbers say about the depth path
From the same brand's full cohort — 730,101 mature cookies:
- 65,168 completed the quiz — 8.926%
- 5,285 purchased — 0.724% of all entrants
- 98.089% of purchasers completed the quiz first
- Quiz completers purchased at 7.955%; non-completers at 0.0152%
That last pair is a 523× difference, and it is the most abusable statistic in this entire dataset.
It is selection, not causation. People who complete a five-question quiz about a product are people who were already interested. The quiz sorts intent; it does not manufacture it at anything like that ratio. A vendor would present 523× as the ROI of adding a quiz. It is not. It is mostly a measurement of who was going to buy anyway.
One number does suggest the quiz has real function though: only 0.839% of initial non-completers returned and completed within 72 hours. The depth path is largely single-session-or-never — if you do not get them through the narrative on the first visit, they mostly do not come back to it.
Why this breaks a single page design
The two systems want opposite pages.
The express buyer wants price, product, and checkout visible immediately. A quiz gate is pure friction — you are asking someone who already decided to answer five questions first.
The depth buyer needs the narrative before the offer. Show them a price in the hero and they leave, because nothing has established why the price is reasonable.
Our short pages were built for the express buyer: 165–280 visible words, zero interactive elements, price in or near the hero on five of six, product image first on mobile on four of six, median dwell 3.8–10.0 seconds. The holdout they lost to was a ~2,500-word presell.
We did not build a worse page. We built a page for one system and sent it traffic containing both.
Practical implications
Segment by ad intent, not by demographics. The relevant question is whether the creative did the persuading. A skepticism-angle ad and a direct-offer ad produce different buyers, and the split above suggests they should not receive the same page.
Do not gate express traffic behind a quiz. For a visitor who was sold by the ad, a quiz is a 5-step tax on a purchase they had already decided to make.
Do not put a price in the hero for depth traffic. The number needs the narrative first.
Stop deciding on CTA clicks. The empty-clicks finding means upstream metrics can invert by checkout. Measure to the terminal event or accept that you are optimising a proxy that is known to mislead in exactly this situation.
Check dwell distribution before touching layout. If 62% of visitors leave inside 10 seconds, no amount of CTA optimisation addresses your actual problem.
Data notes
Journey reconstruction covers 51 converters; cohort rate figures come from the full campaign populations and are stated separately above. Mobile share of this traffic was 87–88%, iOS 45–47%. We also measured a delivery gap worth knowing about if you run paid webview traffic: landing confirmation reached ~90% on Safari/iOS but only 69.9% on Chrome and 71.4% on Android — a roughly 20-point instrumentation gap that will distort any platform-blind comparison.
Lutiq predicts, for every paid click, which on-brand page to show — and measures lift against a live holdout. Figures above are first-party unless attributed.