The first week after changing analytics is the week people decide whether to trust the new tool, and almost everything that worries them in it is expected. Here is what to expect, in the order you will meet it.
Day one: the number is different
It will be. Two tools on the same site always disagree, and the gap is usually in the tens of percent rather than the single digits. The common causes are blocking — requests that never arrive at one tool and do at the other — and bot filtering, where two products draw the line in different places.
The useful response is not to hunt for the discrepancy but to write down the ratio. In a fortnight you will know whether it is stable, and a stable ratio is a fact about your audience you did not previously have.
Day two: the words mean different things
Before comparing anything, check you are comparing the same quantity. Ours are these, exactly:
- A visit ends after thirty minutes without activity. A gap of exactly thirty minutes still extends it, and no single visit can run longer than a day — so a tab left open overnight cannot produce a three-day visit.
- A visitor is counted once per UTC day. Not once per month: the mechanism that makes the product cookieless also stops it recognising anybody tomorrow.
- A pageview is one page being shown, including navigations inside a single-page app.
- An event is something you asked us to count. Events and pageviews are both metered, which is the industry norm and is worth checking against whatever you were paying before.
A monthly "users" figure from a cookie-based tool and a sum of our daily visitors are different quantities, and the second is reliably larger. Neither is wrong.
Midweek: the reports are smaller, and that is the point
There are fewer screens than you are used to. This is the part that feels like a downgrade for about three days and then does not, because the reports that disappeared are overwhelmingly the ones nobody opened twice.
If something you genuinely used is missing, find that out in week one while the old tool is still running, rather than in month three when it is not.
End of the week: pick a system of record
Decide which tool is the one you quote, say so where your colleagues will see it, and stop reconciling. The failure mode here is not picking the wrong one; it is never picking, and quoting whichever is higher this month.
Before you turn the old one off
Do these while both are still running, because every one of them is harder once an account has lapsed:
- Export the history you actually look at. Not everything — the three or four reports somebody opens. Export options are the first thing to become awkward when a tool stops being paid for.
- Find every place the old snippet is installed. Tag managers, a staging site, an old landing page nobody maintains, the checkout on a different subdomain. Something is always missed, and the one that is missed is always the one that matters.
- Check who is depending on it. A scheduled email, a spreadsheet pulling an API, a dashboard on a wall somewhere. These break silently and are noticed a month later by somebody who assumed the number was still updating.
- Write down the ratio. One line, somewhere findable. In six months, when a figure looks wrong, the ratio is what tells you whether it is.
The procedure itself, including the importer that keeps your history, is on migrating from Google Analytics, and why analytics tools never agree explains the gap in more detail. If you are still choosing, the alternatives page includes the cases where staying put is the right answer.