method
Where our numbers come from,
and where they stop
Every attribution tool works from partial evidence. Most of them hide the partial part. This page walks through what we can actually see, what we cannot, and the rules that decide what happens at the boundary.
what a click can prove
What click attribution can see
When someone clicks an ad, the platform stamps the landing URL with an identifier. Google's is called a gclid, and Meta and TikTok each ship their own. Our script reads it, and from that moment the visitor's whole journey through your funnel is joined to the campaign, and where the platform allows it, to the single ad that bought the click. That join is deterministic. There is no model in it and no probability, and nothing gets inferred from people who merely resemble your buyers.
The visitor is attributed once, at entry. A person who arrives on their phone from a Tuesday campaign and buys on their laptop on Thursday counts as that Tuesday campaign the whole way down. Tools that re-derive the source at every step count that journey wrong, and those tend to be the considered, expensive purchases.
what a click cannot prove
How much traffic ad blockers hide
Independent estimates put the share of visitors that ad blockers stop from being measured at 15 to 30%, on any analytics script, ours included. Those are estimates rather than measurements, published by analytics vendors including Sleek Analytics and Kissmetrics, and the true figure swings hard by audience: a developer tool sees far worse than a recipe site. Serving the script from your own domain does not beat them. Blockers match request shapes, and a tool that plays cat and mouse with them has chosen a side against its own users. So a slice of your paid traffic is invisible to us, and the size of that slice matters to every number on your dashboard.
Most tools mention this in a support article. We put a number on it, per site, using the one instrument that still works when scripts are blocked: asking. A one-question survey on your conversion page, with your fixed answers and no free text field. Someone the script never saw can still tell you where they came from.
the reconciliation
Survey answers and click data, kept apart
The survey and the click data answer the same question from different sides, so we show them side by side and never blend them. Self-reported memory is evidence, and it is a different kind of evidence from a click. Averaging the two would manufacture a third number that neither source stands behind.
The useful part is how a missing click gets filed. Say a customer answers Google and we saw no click. Google clicks are the kind we should see, so that gap points at measurement loss, an ad blocker or a stripped parameter or an untagged campaign. Now say a customer answers a podcast. There was never a click to see. That is demand working where clicks cannot look, and it belongs in a different column. Every tool we studied adds those two numbers together and labels the sum unattributed. Separated, one of them is a to-do list and the other is a budget insight.
standing rules
Six places this product refuses to guess
- Does it ever sample my data?
- No. A report that cannot read every event in its range says so and stops. Scaling a slice up to look complete would put exact-looking digits on a guess, so it does not happen here at any plan size.
- What happens to ad spend that matches no visitor?
- It is shown as exactly that: money that joined to nobody. It is never spread across funnel steps to make the cost columns look tidy, because spreading unattributable spend produces a cost per conversion that looks authoritative and is invented.
- How do I know how precise a cost figure is?
- The report says which level the spend matched at. A cost matched to a single creative and a cost matched at account level are different claims, and an account-level match is reported as one. It usually means campaign tagging is the limiting factor rather than the arithmetic.
- Can it tell me whether pausing a campaign caused a change?
- No, and it will not imply otherwise. When spend on a source stops, the report counts conversions over equal-length windows either side of the stop and shows both numbers with the date. Attributing the change to the pause would need a controlled experiment this product does not run.
- Where do creator names in the reports come from?
- From the ad platform, or nowhere. A creator handle is recorded only when the platform reports one. An ad title containing an at sign is a title rather than an identity, and a per-creator report that names the wrong person is worse than one that names nobody.
- Can the written summaries invent numbers?
- No. Written summaries may only use figures the funnel engine already computed, and every number in an answer is checked against that set afterwards. An answer containing a figure we cannot match is discarded whole rather than shown with a warning.
The same rules, running on your traffic
Beta access is free while pricing is decided. The script installs in one tag.