The rule the whole product is built on
An experienced SEO looking at a 74% click-through rate knows immediately that something is broken. A restaurant owner does not, and neither does someone two years into their first SEO job. They will believe it, and then act on it.
So accuracy here is not code hygiene, it is the product. Every figure carries the window it covers and whether it is complete or a sample. Comparisons refuse to run against periods we do not fully hold. Ratios are recomputed at every level rather than averaged. And when a number cannot be labelled honestly, it is not displayed. An absence is recoverable; a confident wrong number is not.
And the second rule: it has to be understandable
A feature that needs explaining is, for the people we build for, a minus rather than a plus. So terms are written out where they are used, screens say what they are for in a sentence, and anything that stays unclear gets cut rather than shipped with a help article attached.
This is why the product does not try to out-feature the large SEO suites. They build depth for professionals. We build unambiguity for people who are not.
The rules we build to
- No estimated data presented as measurement. Everything here comes from your own Search Console. Nothing is modelled, scraped or bought.
- Nothing tracks you without a yes. Anything beyond the cookies the site needs, including noting which site sent you here, waits until you allow it in the cookie banner. There is no session recorder and no chat widget.
- No selling data. Not aggregated, not anonymised, not "with partners".
- No dark patterns in billing. Cancelling is a button. Your archive survives a failed card for three months. Nothing auto-upgrades.
- No write access. We ask Google for read-only, so there is no version of this product that can change anything on your site.
Where we are now
In active development, and honest about it: the free plan is not a trial (one site, no card, no countdown), and the paid prices are published on the pricing page rather than hidden behind a sales call.
What that period has mostly gone on is unglamorous. In one recent stretch we found and fixed around forty defects of a single class: a number from one period or one sample, labelled as another. Three of them:
| Figure | What the screen printed | What was true |
|---|---|---|
| Group CTR | 4.33% | 0.47% |
| Unique queries | 9,838 | 82,094 |
| Staleness indicator | Nothing: it never appeared | Could never fire: its arithmetic produced NaN |
Those numbers are now held in place by more than 2,400 automated checks that re-read the database and assert that every figure the interface prints still matches it. We mention it because "accurate" is a claim anyone can put on a homepage, and the only version worth anything is the one with tests behind it rather than an adjective.
Where it runs
Servers in Germany, inside the EU. Passwords hashed with scrypt, Google tokens encrypted at rest, everything over HTTPS. The main third parties that touch any of it are listed on our sub-processors page.