Blog

Time clock hours don't match payroll: find the gap

· payroll, time & attendance, recordkeeping

When the two totals disagree, the instinct is to open both files side by side and hunt for a missing shift. Don’t. An hours gap at a small hotel is almost never one mistake — it is two or three structural differences that each have a direction, and some of them cancel. Check in a fixed order instead, cheapest and largest first, letting each check eliminate a candidate rather than confirm a hunch. The shape of the gap tells you which one to start on.

Time clock hours don’t match payroll: which number actually moved

Before anything else, decide which of two problems you have.

Hours differ. The clock says 612.00 and the register says 598.25. Something entered or left the file between the two. That is this post.

Hours agree and dollars don’t. Both say 612.00 and the gross is still wrong. Nothing went missing; the same hours were classified or priced differently — which workweek they fell in, which carried a premium, which rate was in effect on which day. A pay period is a sum over a span of days and an overtime threshold is applied to a week, and the two windows do not share a boundary. None of that can move a total, so none of it belongs on the ladder below; it belongs in the file before it leaves, which is a separate job with its own checklist.

Confusing the two wastes an afternoon, because the fixes have nothing in common.

The ladder: six checks, in this order

Order matters because the point is elimination: these are sorted by how cheap the check is against how big the cause usually is, so the top of the list clears the most ground for the least work.

1. Are both sides covering the same window, in the property’s own days? A clock report run for “the 1st through the 15th” and a pay period ending the 15th are the same fifteen days only if both agree where a day starts. The night auditor clocking in at 10:48 PM on the 15th is the whole of this check. Most properties attribute a shift to the day it started; the question is whether both systems do. If one dates punches somewhere other than at the hotel, the shift files itself into the next period — about one shift of gap at the same end of every period, forever.

2. Is there an automatic deduction sitting between them? One setting to look at, and the largest single magnitude on the list. If one side subtracts a break by rule and the other reports punch-to-punch, the gap scales with the number of qualifying shifts — which is why it reads as a big unexplained number rather than a missing shift. An automatic deduction is an assumption, not a record; establish which side is making it.

3. Count the people before you sum the hours. Eleven names on the clock report and ten on the register is a faster finding than 13.75 hours. The hotel-shaped versions: a new hire clocking in before they exist in payroll, and a person working at two properties you own whose hours were exported twice and entered once.

4. Was every punch approved when the file was generated? If punches sit in an approval queue, the on-screen total and the exported total are different numbers until that queue is empty. This one moves in a single direction: the clock side reads high, in whole shifts, concentrated in the one or two people nobody got to.

5. Did anything change after the file was sent? Compare the timestamp of each punch edit against the time the export ran. An edit that landed afterwards is invisible from both ends — payroll never saw it, and the clock report in front of you already includes it. The tell: re-running the same report gives a different total than the one you wrote down.

6. Are there paid hours nobody worked? PTO, holiday pay, a paid training block: these exist in payroll with no punch behind them, so they only ever make the register read higher — often in clean blocks of four or eight hours, though a part-day allowance will not be. If your gap runs the other way, skip this rung entirely.

What survives all six is usually a punch an approver had to guess at — a missed clock-out closed by a default or a person. Those are real hours; they are simply not reconstructable from the file.

What the shape of the gap tells you

You can often skip to the right rung by looking at the gap before investigating it. It says where to start, not what the answer is.

What the gap looks likeStart at
Payroll lower by a constant fraction of an hour, times roughly the shift count2 — an automatic deduction on one side only
Whole shifts, concentrated in one or two people4, then 3
About one shift, at the same end of every period1 — the two sides date the boundary shift differently
Payroll higher, in whole- or part-day blocks6 — paid hours nobody worked
The clock total changes when you re-run the report5 — edits landed after the file went out
Small, scattered across nearly everyoneneither side is wrong; they round to different increments

That last row deserves a sentence: a few minutes per punch across the whole staff is usually two systems rounding to different increments rather than a missing record, and it recurs until one of them changes. Whether a given rounding practice is permissible is a different question, handled below.

A worked example

An invented 48-room independent in southern Ohio: eleven hourly staff, semi-monthly periods, nobody’s week like the last. The clock report for the 1st through the 15th reads 612.00 hours; the register for the same period paid 598.25. Clock high by 13.75.

Rung 1: both sides dated the shift starting 10:48 PM on the 15th to the 15th. Rung 2: the property deducts no break automatically, so nothing sits between the two sides. Rung 3: eleven names on both. Three eliminations, none longer than one screen.

Rung 4 finds it: two housekeeping punches still in the approval queue when the export ran — a 7.75-hour Tuesday and a 4.50-hour Saturday, 12.25 hours the register could not have contained. That leaves 1.50, and rung 5 has it: one approved edit timestamped two days after the file went out, restoring ninety minutes to a maintenance punch somebody had cut short in error. Rung 6 was never needed.

Resolving to exactly zero is cleaner than a real period usually is, and the useful part is not the arithmetic — it is that five ordered checks beat an afternoon of reading two files in parallel. Settling the clock side before the file leaves takes no time at all: start a free trial and close a period against a queue that was empty before you exported.

What this post is not answering

Every rung above finds hours. None tells you whether those hours must be paid, and that is deliberate.

Whether an automatically deducted break was a real break, whether time spent waiting to be given rooms is working time, whether a rounding increment is acceptable, whether a shift crossing midnight counts once or twice against any threshold — those are wage-and-hour questions. They vary by state, they change, and a reconciliation does not settle them. Take the reconciled numbers to your own payroll advisor or your state labor department. This is a method for finding where two totals diverge; it is not legal advice and not a compliance check.

Where this stays genuinely hard

Two causes of opposite sign cancel. A pending 8-hour shift and an 8-hour holiday block produce a gap of zero and two wrong paychecks. A matching total is weak evidence, not proof — which is why counting people and checking the queue is worth doing in a period that already reconciles.

The clock side is not a fixed number. While punches stay editable, “the total” is whatever it was when you last looked. Write down the number and when you pulled it, or rung 5 is unanswerable.

At this size the gap is always small enough to ignore. Thirteen hours across eleven people is under two hours each, well inside what a busy week swallows. It is also the same thirteen hours next period, and the one after, whenever the cause is structural — and rungs 1, 2 and 6 all are. The arithmetic that matters is annual, not per period.

Start seeing what labor actually costs

Set up your property, put staff on the clock, and read tomorrow morning’s labor number.

14-day free trial · No credit card · Cancel anytime