How we know whether a change worked

Most SEO tools tell you what to change. CommitSEO changes it, checks it on the live page, and then measures what happened. This page is the exact method - including the parts where the honest answer is "we can't tell".

Our promise is evidence, not outcomes. We can't promise a change will raise your traffic. We can promise that every change is verified on your live site, measured against rules written down before it went live, and reported the same way whether it helped, hurt, or proved nothing.

The five steps behind every result

In this order, every time.

Terms are fixed before the change

Deciding how to judge a result after seeing it is how bad measurement hides in good-looking charts.

Before anything is written to your site, we record the success metric (daily clicks), the safeguard metric (server errors), the measurement window, the expected direction and the decision rule
A fingerprint (SHA-256) of those terms is stored with the change, so the terms shown next to a result can be shown to be the ones that existed beforehand
A change recorded after it went live is never presented as pre-registered

The write is verified on the live page

An "OK" from an API is not proof that anything changed.

After writing, we clear caches, wait for them to settle, then fetch the real URL with a cache-bypassing parameter and compare it to what we meant to write
If the live page doesn't show the change, it is reverted automatically
Where Google Search Console shows a later crawl, we show that Google has seen the change too

A matched control group

Traffic moves for reasons that have nothing to do with us - seasons, news, algorithm updates.

We compare the changed page with similar pages on your site that we did not change (matched by page type and traffic volume, and by how closely their daily clicks have moved together)
The measured effect is the change in the changed page beyond what the control pages did over the same days (difference-in-differences)
We use the page's own earlier history to reduce noise (CUPED variance reduction)

A range, not a single number

Daily clicks are noisy; one figure would look more certain than it is.

The window is 14 days before and after the change, with a few extra days of buffer because Search Console data arrives 2–3 days late
The result is a probability distribution with a 95% credible interval
A result is called positive or negative only if that interval excludes zero. Otherwise it is reported as inconclusive

A safeguard that can overrule the result

A page that gets more clicks but starts throwing server errors is not a win.

A rise in 5xx server errors during the window overrides any click result
Measurement results only flag a change for review - a person decides whether to revert it
Every revert restores the exact previous value, checked step by step

What we do not claim

Being clear about the limits is part of the method.

Proof of cause on small sites. With a few dozen clicks a day, two weeks of data rarely separates a real effect from noise. Expect "inconclusive" often - we show it as plainly as a win.

Immunity to outside events. A control group softens seasonality and algorithm updates; it cannot remove them, especially if they hit different pages differently.

That every edit is visible in results. Google rewrites a large share of page titles and descriptions on its own, so some edits change nothing you can observe.

Rankings we can't see. We rely on your own Search Console data, not estimates from third-party tools or scraped results.

About the numbers. Any success figure we publish will state how many changes and how many sites it is based on. Today that sample is small, so we publish the method first and the totals only once they are large enough to mean something.

See it run on your own site

Two free checks, no signup: how AI assistants see you, and what's technically wrong with your homepage right now.

Run the free GEO audit Run the free technical audit