More than half of new launches ask you to act and give you no evidence at all

By Ben Peetermans · 2026-08-09 · UXXRAY launch cohort, n=1086

A launch page is a stranger asking for something. Your email, your card, your afternoon. The question a visitor answers in about four seconds is not "is this well designed", it is "why should I believe you".

We measured how many new products answer that question. Of the funnels that ask for something, 55% offer no evidence whatsoever: no review, no rating, no customer count, no logo wall, nothing a skeptical visitor could check.

Finding 1: most of them ask, and most of those cannot back it up

Two facts, measured separately, then combined.

70% of the launch cohort makes a real ask: a signup form, a free trial, a pricing page, a demo request, a checkout. That is the point of a launch page, and most of them do it.

63% carry zero social proof, measured across the whole cohort. Not a thin testimonial. Not one logo. Nothing.

Overlap those and you get the number that matters. Among the funnels that do ask, 55% have nothing to support the request.

	Count	Share
Launch cohort, zero social proof	688 of 1086	63%
Sampled for the overlap (every third funnel)	362
Makes a conversion ask	252	70%
Asks, and has zero proof	138	55% of askers

The zero-proof funnels that ask for nothing are brochures. They owe nothing, so they are excluded from the headline: a page with no ask and no proof is not making a promise it cannot keep.

This replicates an earlier finding of ours at roughly ten times the sample. In July we measured 94 launches and found 74% with no proof. At n=1086 it is 63%. Same shape, smaller number, much larger sample. We publish both.

Finding 2: the difference is density, then length

The obvious theory is that better funnels are longer, and longer funnels simply have more room for tactics. That theory is half wrong, and the half that is wrong is the useful half.

Measuring conversion tactics per page removes the room advantage:

Band	Funnels	Tactics per page	Pages walked
Sparse	518	1.25	1.63
Thin	282	2.29	2.26
Partial	199	2.36	2.79
Strong	77	2.58	2.99
Excellent	10	2.24	3.60

That is not a gradient. It is a cliff and then a plateau.

1.25

Sparse funnels sit alone at 1.25 tactics per page. Every band above them lands between 2.24 and 2.58, and the differences up there are noise. So there are two different problems wearing the same score.

Below the cliff, your pages are empty. Adding steps will not help. A second empty page is still empty. The fix is density: put something on the page a visitor can act on or check.

Above the cliff, density is solved and length is the constraint. Sparse walks 1.63 pages, Strong 2.99, Excellent 3.6. Once each page is doing work, more steps means more work done.

Most launches are below the cliff: 518 of 1086, or 48%.

Finding 3: the top band is basically empty

Ten funnels out of 1086 reached Excellent. Not ten percent. Ten.

They walk 3.6 pages against a cohort average of 2.12, and carry proof on more than one step. Nothing exotic. They do the ordinary things, and they do them on more than one page.

Band	Funnels	Share
Sparse	518	47.7%
Thin	282	26.0%
Partial	199	18.3%
Strong	77	7.1%
Excellent	10	0.9%

Cohort average score: 39.9 / 100. Average walk depth 2.12 pages, 3.58 tactics detected.

Finding 4: 1 in 20 blocks the robots reading it

Of 796 domains attempted, 42 refused automated access outright and 45 errored. That is a 5% hard block rate among products launched this month.

Worth sitting with, given who reads the web now. A launch page exists to be found and understood, increasingly by an agent doing the reading on someone's behalf. One in twenty new products has closed that door on its own funnel.

We label a refusal as blocked and never score it. A fabricated zero for a site we could not read would be worse than an absent row.

What to actually do

1. Put one checkable fact on the page. A count, a named customer, a rating with a source. 55% of asking funnels have none, so this alone moves you past most of the cohort.
2. Get one page above about 2 tactics. Below the cliff, adding pages does nothing. Density first.
3. Then add the step. Once pages carry weight, a pricing or comparison step is what separates Strong from Thin.
4. Let robots read it. If an agent cannot walk your funnel, it cannot recommend it.

Method

Deterministic extraction: DOM, computed CSS, string matching. No model anywhere in the measurement path. Every detected tactic carries the exact on-page string it fired on. If a fact is not provably on the page it is not recorded, and a site we could not read is labelled blocked rather than scored zero.

Each funnel is walked the way a first visit goes: landing, then pricing, signup or checkout where they exist. Social proof is proof_points on the teardown profile. The ask is any of lead_form, free_trial, demo_docs, pricing_tiers, express_checkout or outbound_offer in the detected pattern set.

Reproduce it

• Corpus totals: GET https://api.uxxray.com/api/v1/corpus/stats
• Ranked corpus: GET https://api.uxxray.com/api/v1/league?limit=5000
• One funnel, every measured fact: GET https://api.uxxray.com/api/v1/teardown/{slug}
• Tactic definitions: GET https://api.uxxray.com/api/v1/patterns

FAQ

What counts as social proof?
A measurable on-page claim a visitor could check: a review count, a star rating, a customer or user count, a trusted-by logo wall, a named case study. Marketing adjectives do not count, because they are not checkable.

Why exclude the brochures from the headline?
A page that asks for nothing owes nothing. The interesting failure is a funnel that requests an action and offers no reason to grant it, which is why the denominator is funnels that ask.

Is 55% bad?
It is normal, which is the point. A single checkable fact puts a launch ahead of most of its cohort, and that is a cheap edge.

Does a high score mean a good page?
No. It means broad coverage of the conversion tactics that apply to that funnel type. A deliberately minimal page can score low and convert well.