An Office Attendance System That Survives Hybrid Working

An office attendance system built around a door device answers one question: did this person enter the building. That was a reasonable proxy for attendance when everyone did, every day. It has not been one for several years, and the register has been quietly diverging from reality ever since.

The symptoms are familiar. Work-from-home days are absences the office manager corrects by hand. Client-site visits show as no-shows. The person who came in at seven and left at eleven for a hospital appointment looks identical to someone who did four hours of nothing. So the authoritative record becomes the spreadsheet somebody maintains alongside the system, which is precisely the thing the system was bought to replace.

The fix is to record the work rather than the entrance. A session is a block of work someone starts and stops, attached to the project it was for, approved by their manager afterwards. Where the person physically was during it is a property of the session, not the thing being measured.

This is why hybrid stops being an exception. A day at a desk, a day at home, a morning at a client's office and an afternoon in the office all produce the same kind of record in the same system, with the same approval path. Nobody maintains a parallel list of who was allowed to work from home on Thursday, because the record does not depend on the door.

• In-office, work-from-home, client-site and field staff on one record.
• No manual corrections for remote days, because they are not anomalies.
• Approval by the manager who knows what the work was.
• Half-days and partial attendance recorded as what they were.

Below the attendance question sit three others that an office runs on every month, and they are usually answered in three different documents. Payroll needs days present, leave taken and any loss-of-pay. Compliance needs a register that matches the state Shops and Establishments requirement and can be produced during an inspection. Anyone billing clients needs to know which hours went to which engagement.

When those come from three separate sources, they disagree, and reconciling them is somebody's last week of every month. Because approved sessions carry both the time and the project, all three are views of one dataset here. The register, the payroll input and the utilisation number cannot contradict each other, because they are the same records totalled differently.

A device-based system has no human judgement in it, so the exceptions pile up on a single administrator who was not there and cannot verify any of them. A team lead who worked alongside the person all week reviews their sessions in a couple of minutes, and the decision is recorded against each session with the approver's name attached.

That distribution is what makes the record defensible in a dispute. A machine reading says a badge was presented. An approved session says a named manager confirmed the work happened, and it says so on a record that keeps the original alongside any correction rather than overwriting it. The audit trail is the point: in a disagreement about hours months later, the question is always what was recorded at the time.

Almost every commercial office in India falls under its state's Shops and Establishments Act, which sets working hours, rest intervals and leave, and requires registers to be maintained and produced on inspection. The rules differ by state, and the practical burden is nearly always the same: produce an accurate attendance record covering the period an inspector asks about.

Sessions are timestamped, approved and exportable in a muster-roll shape, so producing that record is a lookup rather than a reconstruction from a badge log and a leave spreadsheet. The register is generated from the same approved data as everything else, which is the property that keeps it consistent with what payroll already paid.

Removing the door hardware from the attendance path removes the cards that get lost and shared, the readers that fail at satellite offices, the annual maintenance contract and the biometric template store. Access control, if you need it for the building, is a separate decision — and keeping the two separate is worth doing deliberately, because conflating them is how an office ends up with biometric attendance it never explicitly chose.

On the data side the position is straightforward. No fingerprints, no face templates, no Aadhaar numbers, no screenshots, no keystroke logging, no idle scoring. Under the DPDP Act, 2023 the operative principles for an employer are purpose limitation and minimisation, and data never collected is the shortest version of that story — there is no template store to secure, retain or notify a breach on.

Per person, no hardware, no installation. Free for up to 3 users, then ₹199 per user per month with no base platform fee and no seat minimum — so a twelve-person office pays for twelve people rather than clearing a minimum-seat threshold first.

For an office weighing a reader replacement across two or three sites, the comparison worth running is device cost plus maintenance against per-seat software with nothing to install. The free tier is there so a pilot with one team can happen before any of that is decided.