Every experiment starts with a hypothesis. Except the one about the page.

Growth teams run twenty experiments a month and learn from ten — because half the failures could have been hypothesis or could have been bad copy and you can’t tell. Lytms scores the page before the traffic, separates the variables, and doubles your effective learning rate.

A verdict reads like this

One sentence. The thing your buyer is reading.

Every Lytms scan synthesizes one verdict — the highest-leverage sentence the product makes about your page. Here’s the verdict from a recent scan of Stripe. Yours arrives in about 2 minutes.

Verdict · Stripe8.0^/10

Stripe's specific proof — Hertz, URBN, £540 million — never reaches buyers who leave before scrolling past the generic claims that lead.

The reality of running experiments without a content score.

Twenty experiments a month sounds like velocity until you realise half the failed ones are confounded. Without a content score, you can’t tell whether the test failed because the idea was wrong or because the page was.

What it looks like now

A failed test, and no way to tell why.

The pricing-page experiment failed. Was the new pricing structure wrong, or did the landing page explain it poorly? You don’t know because you never scored the page.

You spend a week rewriting copy that was already fine, or shipping a hypothesis that was wrong from the start. Either way, the learning rate stays half what it could be.

What Lytms gives you

Score the page before the traffic.

Paste both variant URLs. Get the scores back in about 2 minutes, before you allocate a visitor. Variant A scores 7.4. Variant B scores 6.1. You already know B is starting at a disadvantage.

Fix the content first. Then test the hypothesis. When the test fails, you can isolate the variable that caused it — because the other one was measured before launch.

What growth teams actually use Lytms for.

Pre-flight

Score every variant before traffic allocation.

You write two variant pages for the next test. Before either gets a visitor, you score both.

Variant A scores 7.4. Variant B scores 5.8. You either fix B to match or acknowledge the confound out loud — instead of pretending the test was clean.

Stage: Pre-experiment

Post-mortem

Read the failure report with a number, not a guess.

The test failed. You check the page score. 8.1 — the hypothesis was wrong, move on. 5.5 — the copy killed it, rewrite, retry.

Your post-mortem stops being a debate and starts being a routing decision. You learn from twice as many tests because the confounds are isolated.

Stage: After every test

Backlog triage

Decide which experiments are worth running this quarter.

You have a backlog of 30 hypotheses. Which pages are the lowest-scoring? Which have the most upside on a single move? Lytms gives you the prior so the quarter’s plan stops being intuition.

Pages scoring under 6.0 get rewritten before they’re tested. Pages above 8.0 get the hypothesis-only treatment. The mix gets sharper.

Stage: Quarterly planning

What your growth team stops doing. What it starts doing.

Stop doing

Running confounded experiments and pretending the result was clean
Spending a week rewriting a page that was already above-bar
Shipping the same hypothesis twice because the post-mortem was a debate
Letting your CRO calendar fill with low-leverage tests
Telling product the test failed without naming which variable caused it
Re-running last quarter’s losing experiments in case it was just bad luck

Start doing

Scoring every variant before traffic allocation
Separating hypothesis failure from copy failure in every post-mortem
Prioritising the backlog by where the score upside is highest
Telling product exactly which variable to re-test next
Doubling your effective learning rate with no extra experiments
Closing the loop between content quality and conversion rate, in numbers

Start with the free scan.

One URL. About 2 minutes. The truth your team has been guessing at.

Score your homepage →See pricing

Free, forever. No card. No follow-up email blast.