Impact & Trust, in plain sight.
Two numbers sit behind every recommendation we make. Here's what they mean, where they come from, and why you can lean on them — without a single equation.
Impact
How much an element tends to move the needle on conversion, read from public evidence. A high Impact says “this choice generally works.” It answers the first question you actually have: should I do this?
Trust
How much you should believe that Impact — based on how much real evidence stands behind it, how fresh it is, and whether the sources agree. It answers the second question: how sure can I be? High Impact with low Trust is a lead, not a law.
Read them together, never alone.
Impact tells you the direction. Trust tells you how hard to commit. The combination is the decision.
Proven winner. Copy it with confidence.
Promising but directional. Worth testing yourself.
Reliably weak. Avoid — the evidence is clear.
Unknown. Gather more signal before betting.
Built on public evidence — three layers deep.
Every Impact score blends these. We only ever use public, legally gathered signal — never private analytics, never a competitor's dashboard.
What real users say
Sentiment mined element-by-element from public app-store and community reviews — at scale, not a handful of cherry-picked quotes.
What it adds: Points the direction: are people reacting well, or quietly walking away?
Published conversion studies
A curated library of public A/B and CRO findings, each mapped to the element it actually speaks to, carrying its measured lift.
What it adds: Grounds the read in what's been measured, not what feels true.
Calibrated heuristics
Deterministic quality checks — contrast, clarity, length, busyness — tuned against a growing corpus of real apps.
What it adds: Gives a defensible read even when nobody has written a word about an element yet.
Those three layers don’t work in isolation. They sit inside a much wider evidence base we gather on every app we read — all of it public, all of it cited back to source:
The more of this we can see for an element, the more the two scores have to stand on.
How the Impact number is formed.
Impact is a blend of the evidence layers that exist for an element, resolved into a single 0–100 read. Hard measurement carries more weight than a rule of thumb — a published study outranks a hunch. But when a layer is simply missing, we lean on the ones we have rather than guessing.
The upshot: an element is never punished for evidence that doesn’t exist yet. A screen with only a heuristic read still gets a fair Impact — it just won’t get a confident Trust to match. Which is exactly the point of having two numbers.
Trust is the number most tools fake.
Ours has to be earned. It rises with the amount of real evidence we've actually processed — and it stays honest when that evidence is thin.
How much evidence
One review or one lone study keeps Trust modest. A chorus — hundreds of mentions, several independent studies pointing the same way — earns real confidence. Trust measures how sure we are of the read, never whether the crowd happens to be right.
How fresh it is
Reviews fade with age; published CRO findings age slowly; a heuristic we just measured is current by definition. We take the strongest, most recent source rather than dragging a fresh signal down to sit next to an old one.
Whether it agrees
When the layers point the same way, Trust climbs. When they pull against each other, we temper it on purpose — honest uncertainty beats false confidence every time.
Confidence bands.
The engine, run live.
These aren't mock-ups. They're computed by the exact same engine that powers the product — run the moment this page was built.
Pricing model
Reviews + two studies- 67% positive · 22 mentions
- 2 supporting studies (+12%, +8%)
- the real artifact, observed
App icon
Heuristic only- clarity + contrast checks
- no studies, no reviews yet
- corpus still being built
Look at the second one: a perfect Impact, but low Trust, flagged directional. That’s not a bug — that’s the system refusing to oversell a read it can’t yet back up.