The Chain From a Punch to a Payment

Take one employee, one week, and follow the number through every link between the terminal and the bank. It takes an afternoon, it needs no special access, and in most organisations it finds at least two places where the figure changed and nobody can say who decided it should.

The recordkeeping discipline in “The Chain From a Punch to a Payment” should also apply to workforce technology. When a team evaluates see the service here in relation to remote employee monitoring software, it should document purpose, access, retention and deletion, then preserve the source entry, approvals and correction history needed to explain the final figure.

The exercise works because wage and hour problems are almost never decisions. They are settings — a rounding rule chosen when the system was installed, an automatic deduction that made sense for a different shift pattern, a threshold applied to the wrong period. Each was reasonable once and none of them is anybody's job now.

For an independent reference relevant to “The Chain From a Punch to a Payment”, consult the OECD hours-worked data. Use it to test record quality, working-time definitions, access, retention and exception handling against the organisation’s real payroll process.

The links, named

1. The raw event: what the terminal, app or sheet actually recorded, with its timestamp.
2. Rounding and grace periods applied to that event.
3. Scheduled rather than actual times, where the system substitutes one for the other.
4. Automatic deductions, usually for breaks.
5. Edits and corrections by a supervisor or by payroll.
6. Approval, which is where unapproved time can silently disappear.
7. Aggregation into a week, or into whatever period the system actually uses.
8. The threshold at which a premium starts, and the period it is measured over.
9. The rate the hours are multiplied by.
10. What the statement says, which is the only link the employee sees.

Ten links. Most organisations can describe three of them and have never looked at the other seven.

Getting the raw event

The first thing to establish is what the system recorded before anything touched it, and this is harder than it sounds. Many systems display the rounded value everywhere, including in the reports, and the raw timestamps sit in a table that is only reachable through an export.

Ask for the export. If nobody can produce one, that is itself the finding: the organisation cannot see what its own system was told, which means it cannot check any link downstream of the first one.

Comparing link to link

Lay the values side by side in the order above, one row per link, and mark every row where the number differs from the row before it. Some differences are intended and correct. The point is that each should be explainable by somebody, in a sentence, without looking anything up.

For every change between one link and the next, write down who decided it, when, and why. Where nobody can answer, the setting was inherited rather than chosen, and it needs a decision now.

That sentence is the whole method. It separates the settings an organisation stands behind from the ones it has merely been running.

The three that go wrong most often

Rounding that is not symmetrical. A rule that rounds the start of a shift forward and the end backward removes time in one direction only, which is a different thing from rounding and is usually visible in the aggregate within a month.

Automatic break deduction. A thirty-minute deduction applied to every shift over a length, regardless of whether the break was taken, is the single largest source of unpaid time in most organisations that have one.

The wrong period for the threshold. Where a premium is owed on hours beyond a weekly figure, applying that figure to a two-week pay period makes a heavy week followed by a light one disappear entirely.

Edits, which need reasons

Every correction should carry a reason and an author. Systems usually support this and the field is usually blank, which means a record was changed and nobody can say why.

Count them before reading any of them. A supervisor making four hundred edits a quarter is not correcting mistakes; the process is producing something that needs correcting, and the edits are a symptom rather than the problem.

Approval as a silent filter

In many systems time that is not approved is not paid, and nothing tells the employee that their hours went unapproved. Minutes vanish at this link more quietly than anywhere else, because nothing was deducted — something simply never arrived.

Pull the list of unapproved entries for the period and look at it. If there are any, somebody worked time that nobody has either paid or refused, which is the worst of the available states.

Writing the chain down

The output of the exercise is one page: each link, the value at it, and a line saying who owns the setting. Keep it, because its value is in comparison — rerunning it next year against the same employee, or the same week against a different site, shows what has drifted.

It is also the document that makes this a system question rather than a payroll question. A chain with four unexplained links is not a payroll error; it is six settings nobody owns, and it is the one artefact capable of getting them owned.