Blog Compliance
A clocking record that cannot be quietly rewritten
Most time-tracking tools store hours in a table anyone with database access can edit. That is fine until the day somebody asks you to prove a number.
The question behind the question
When a labour inspector, an auditor or an opposing lawyer asks for your working-time records, they are not really asking for a spreadsheet of hours. They are asking something harder: can you show that these hours are the ones that were recorded at the time, and not the ones somebody decided last week they should have been?
Almost every tool on the market answers the first question and quietly fails the second. Clocking in writes a row; clocking out updates it; a manager fixing a mistake updates it again. The row you are looking at is the last state of a value that has been overwritten an unknown number of times, and nothing in the file says so.
That is not a scandal — it is just how software gets written when nobody asks the second question early enough. But it means the export you hand over is a statement of what your system currently believes, not evidence of what happened.
Append-only: the one design decision that changes the answer
The alternative is to stop storing states and start storing events. Clocking in is one row. Starting a break is another row. Clocking out is another. A correction is not an edit — it is a new row that says «this replaces that one, and here is who said so and when».
Nothing is ever updated and nothing is ever deleted. The hours you see on screen are not stored anywhere: they are worked out by reading the events, the same way a bank balance is the sum of the transactions rather than a number somebody keeps up to date.
The practical difference shows up the moment somebody disputes a day. With a mutable table you can say «it says seven hours». With an append-only record you can say «she clocked in at 08:02 from the Getafe site, went on break at 13:10, came back at 14:05, clocked out at 17:30, and on the 3rd her manager corrected the clock-out from 17:30 to 18:15 because she had been on a customer site» — and all six of those statements are rows, not recollections.
Three layers, and what each one actually protects against
Vendors tend to say «immutable» and stop there. It is worth being precise, because the three usual mechanisms defend against completely different attacks and none of them covers all three.
- Database triggers. A rule inside the database that rejects any UPDATE or DELETE on the records table. This stops the realistic accident: a badly written migration, a support script run against the wrong environment, a developer fixing one row by hand. It does not stop somebody with administrative rights who drops the table and rebuilds it.
- A hash chain. Each record carries a fingerprint of the previous one for that same person, computed with a secret that never appears on a screen, in an export or in an API response. This is what covers the case triggers cannot: if a table were emptied and rewritten, the chain would not add up. It catches deletions, which is the failure mode triggers are blind to.
- A verification anyone can run. The chain is worth nothing if only the vendor can check it. The check has to be a button in the product, runnable by the customer, that says how many marks were read, how many were altered and how many are missing. A record only its keeper can verify verifies nothing.
What honest immutability does not give you
It does not stop somebody clocking in for a colleague. It does not prove the person holding the phone is who they say they are. It does not make a wrong entry right. What it gives you is narrower and more useful than the marketing usually suggests: whatever went in is still there, unchanged, and anything that was corrected shows both versions and who signed the correction.
That is the claim worth making, and it is the one you can defend in a meeting.
How Fichaloto does it
Every clock-in, break, return, clock-out, amendment and annulment is a separate row that the database itself refuses to modify or delete. Each carries a SHA-256 fingerprint chained to the previous one for that person, salted with a per-account secret that is not exposed anywhere. A Check the chain button on the team clocking screen walks the whole thing and reports two distinct results: records whose content no longer matches their fingerprint, and gaps where a record is missing.
Deleting a day is not deletion: it adds an annulment record. The day stops counting and disappears from the totals, but the register still shows it existed and who annulled it. A deletion with no trace is indistinguishable from an entry that never existed, and that is exactly the difference you need to be able to demonstrate.
Frequently asked questions
Is an append-only record legally required?
No jurisdiction we know of specifies the storage design. What Spanish law (article 34.9 of the Estatuto de los Trabajadores) and most equivalents require is a daily record, kept for four years, available to staff, their representatives and the labour authority. Append-only is not a legal requirement — it is how you make the record worth producing.
What happens when somebody genuinely makes a mistake?
They ask for a correction, saying what actually happened, and somebody with permission approves it. The correction is a new record pointing at the old one. Both stay. That circuit is the point: a record each person can edit at will is not a record.
Can I export it?
Yes, and you should be able to. In Fichaloto the raw register downloads as a spreadsheet and reads through the REST API, fingerprints included, so you can keep your own copy or verify it outside the product. It is read-only in both: writing a record by hand would forge a link in the chain.
All of this is built into Fichaloto. See how it works.
Keep reading
How to choose
Twelve questions worth asking before you sign
None of these twelve questions is technical. Every one of them is answered in a sentence, and you can ask for the answers in writing before signing anything.
Time and attendance
The clock ran all night. Now what?
Forgetting to clock out — or to punch out, if your team says it that way — is the single most common incident in any time-and-attendance system, and the way a product handles it tells you most of what you need to know about it.
Capacity
Planned, clocked, charged: three numbers that should match and usually do not
Most companies can tell you what they planned and what they invoiced. Very few can tell you, for a given week, whether those two things are related.