We asked for eight months of Clockify timesheets and got 50 rows
It was not the free plan and it was not an error. 50 is their default page size, and our request for 200 had been silently ignored. The import looked perfect: no errors, no warnings, 50 neatly mapped rows. This is the list of things we learned moving real timesheets — and what you should check, whichever tool you use.
1. The trap that cuts your data without saying anything
Any API import reads in pages, and the loop almost writes itself: ask for 200; if fewer than 200 come back, we are done. That assumption is false anywhere the server may ignore the page size you asked for. Clockify politely ignored ours and answered with its own default of 50. The loop saw „50 < 200”, concluded „finished”, and stopped after the first page.
What this means for you: after any import, count. Do not look for a green message — compare the number of hours in the old tool and the new one, for the same month. If one of them is a small round number, that is not a coincidence.
2. Duplicates
Everyone imports twice. The question that decides whether you end up with double hours is: what makes a row recognisable as already imported? If the fingerprint includes the source, the same hour from a CSV and from the API are two different things. Ours is one per company: the same hour, the same person, the same day, the same project is the same thing — however it got there.
Import one week first, then repeat it identically. If the hours double the second time, stop.
3. The emails do not match. For almost anyone
We expected people to recognise themselves by email. In reality the address someone has in the old tool differs from the new one for nearly everybody. Mapping cannot be automatic, and it is better that it is not — what matters is that the tool shows you the mapping rather than guessing it.
4. Harvest exports have no email at all
A real Harvest export taught us two things the documentation does not: the name arrives split into two columns, and the email is missing entirely. If a tool looks for a single name column and finds nothing, the whole team's hours land on the person doing the import. Check the number of people after an import, not only the total hours.
5. What each platform actually gives you
| Platform | What the key brings |
|---|---|
| Clockify | the whole workspace's hours |
| Toggl | only the key owner's own hours — we could not confirm a whole-team endpoint, and we do not guess at contracts |
| Harvest | export from the interface; no email, name in two columns |
6. The API key
In Clockify: Profile settings → Advanced → Manage API keys. It is shown once. Ask where the key ends up in whatever tool you paste it into — and revoke it when the migration is done.
7. The short list
- Import one week first, not eight months.
- Count the hours in both tools for the same month.
- Count the people.
- Repeat the trial import. If it doubles, stop.
- Only then bring the full history.
- Revoke the key.
Sources
Every figure and behaviour here was measured on our own connectors against real Clockify, Toggl and Harvest accounts in August–September 2026 — not quoted from documentation. Platform documentation changes without notice: one Jira endpoint we used was removed between versions, returning 410. If you find something different, it probably changed on their side.