Products
Pulse01Digaso02RevealWhy03Rosto04Penguins05
StudioCompanyContact
Start a project
03RevealWhy · Evidence engine for the web

See what visitors notice. Prove why it converts.

RevealWhy reads predicted and real attention, the content itself and every conversion — then tells you what to change and proves the change worked. The math decides; the AI only phrases what is true, and says “I can't tell yet” when the data can't carry the claim.

Start free on revealwhy.com How it works
  • 1,000 sessions / month free
  • Setup in 2 minutes
  • Self-hostable
app.revealwhy.com/pages/checkout
Attention · above the fold
1ST FIXATION · 0.4s
Provenverdict #4187 · checkout

Moving the trust badges above the fold lifted checkout starts.

+18.4%checkout starts
96%confidence
12,480sessions
14druntime
Illustration · sample data, not a customer result
Built to be inspectedSelf-hostable end to endNo raw input captureYour own LLM keyTenant-isolated
The difference

Analytics tells you what happened. RevealWhy tells you why — and proves it.

Heatmaps, funnels and dashboards describe behaviour. They leave the causal step as your guess. RevealWhy closes it with a measured verdict.
Every other tool

The what

Where they clicked, how far they scrolled, where the funnel leaks. Accurate, but mute on cause.
The gap

The guess

A hypothesis, a redesign, a hope. Weeks later a metric moved — but you can't say it was you.
RevealWhy

The proven why

It finds the likely cause from attention and content, proposes the fix, tests it, and returns a labelled verdict you can trust.
The substrate

Six layers of meaning under every page.

Most analytics has events. RevealWhy captures meaning — a grounded, per-site substrate that lets an agent reason, cite evidence, and be graded against reality.
01

Predicted attention

What visitors are likely to notice — predictive eye-tracking from a saliency model, before a single visitor arrives.

02

Real attention

Click, scroll, dwell and on-screen visibility from live traffic, calibrating the prediction against reality.

03

Content & element identity

The unit of reasoning: this headline, this image, this CTA — tracked as it changes.

04

Conversions & visits

The outcome to optimise — auto-detected per page, stitched into one visit, in money where it can be.

05

Session replays

Concrete evidence the AI can cite — “watch these six sessions” — not just a number.

06

Per-site memory

A longitudinal record of what was tried and what won, so the agent gets truer on your site over time.

Day one → week one

Watch the prior get overwritten by reality.

Any chatbot will hand you twenty confident suggestions for free. Ours arrive labelled as a guess — and then real behaviour overrules them.
Simulated · minute one

The prior

Digital twins read your page and predict what visitors will notice and skip. Useful immediately — and marked Simulated everywhere it appears, because that is exactly what it is.

Observed · week one

The correction

Measured behaviour replaces the guess. Some priors are confirmed, some are plainly wrong — and RevealWhy tells you which. Every finding carries the basis it earned.

The overwrite is the product. An opinion that never meets reality is just content.

The loop

Propose. Test. Prove. Learn.

Every suggestion eventually earns a verdict, so the system grades its own predictions against reality. That is the difference between an opinion and an instrument.
01 · Propose

Find the likely cause

High attention on the hero, low clicks: an attention–action trap. Fixes ranked by effect × traffic × value.
02 · Test

Ship the variant

The change goes live through the SDK as a controlled A/B — human-approved before it ships.
03 · Prove

Read the verdict

The real outcome, labelled by basis: proven, controlled or principle. Never louder than the evidence.
04 · Learn

Get truer over time

Every verdict feeds back, calibrating future suggestions for your site.
The output

A verdict, not a dashboard to interpret.

A conclusion you can act on and defend — with the confidence, the sample and the evidence attached.
Provenverdict #4187 · checkout

Moving the trust badges above the fold lifted checkout starts.

+18.4%checkout starts
96%confidence
12,480sessions
14druntime
Basis: measured A/B · Cause: attention–action trap▸ Watch 6 replays
Illustration · sample data, not a customer result
Not yetverdict #4192 · pricing

“I can't tell yet — the sample is too small to be honest about this.”

The hardest behaviour to get from an AI is to decline to over-claim. We made it a product rule.

Needs ~2,100 more sessionstracking →
Illustration · sample data
Three doors

However you change your site, something should check the work.

Creators

You change the words

Edit a headline, move a CTA, reorder the page. RevealWhy tells you which of your changes actually won.
no code · approve from the dashboard
Developers

One script tag

Drop the sensor in before </body>. Conversions are detected automatically — no event taxonomy to design.
<script src="…/insight-recorder.js">
Agents

Your coding agent asks

RevealWhy speaks MCP: tools your agent can call to read findings, drop-offs and the evidence behind every claim.
mcp · list_findings · get_finding
Where it sits

Measure-only tools show. Test-only tools prove. RevealWhy does both.

CategoryFinds the whyProves the fixGenerates the changeLearns per site
Behaviour analyticsContentsquare · Hotjar◐ partial— no— no— no
ExperimentationVWO · Optimizely— no● yes— no— no
Attention predictionAttention-AI tools◐ partial— no— no— no
RevealWhyThe evidence engine● yes● yes● yes● yes

Stop guessing why. Start proving it.

Drop in one script. Twins give you a labelled prior in minutes, findings arrive as traffic does, and a verdict lands once a change has been tested. Free to start, self-hostable when you're ready to own the stack.
Start free — 1,000 sessions/mo Read the methodology