Wage-and-hour data: the same records, two standards, a fifty-four-fold difference
183,429 shifts, one dataset, and an exposure figure that moved by a factor of fifty-four depending on which meal-period threshold applied. The data did not decide the case. It showed precisely what the case turned on.
Counsel heading into mediation sent a list of questions about a produced timekeeping and payroll sample, and needed each one answered with a count, a percentage, and a statement of where in the data the answer came from — not a conclusion about liability.
Answer the question that was asked
Each answer carries the count and, where the question calls for one, the percentage — together with the workbook sheet and column the figure was computed from, so that the other side can reproduce it or contest it.
Match the two datasets before comparing them
Timekeeping and payroll arrived as separate files. 20,004 pay periods could be matched across both; those that could not were identified and counted rather than dropped silently.
Compute the exposure under each threshold, not the favourable one
The same violation tests were run at the five-hour and ten-hour thresholds and again at six and twelve, and both sets of results were reported in the same document.
Report the absences too
Where the records showed no rounding, no time shaving, no rest-break punches and no expense reimbursements, that was stated as a finding in the same terms as everything else.
Where the exposure actually lived
Under a five-hour first-meal and ten-hour second-meal threshold, 14,073 shifts carried a violation — 7.8 percent of shifts over five hours, spread across 4,456 pay periods. Under a six-hour and twelve-hour threshold, the same records produced 260 shifts and 238 pay periods.
Nothing in the data changed between those two runs. What changed was which threshold applied, and that is a legal question rather than one an economist answers. Counsel could see, before mediation, that almost the entire exposure rested on it.
The second finding sat somewhere else entirely. In 5,415 pay periods — 28.9 percent of those that matched on hours — overtime, where it was owed, had been paid at a rate that did not incorporate the bonuses, commissions and shift differentials the payroll data showed. That accounted for 74,209 hours, and it had nothing to do with meal periods.
What we do not do
We do not report the version of a number that helps whoever retained us. Both thresholds were run on the same records and both went into the same document, in the same table, with sources.
We also do not fill a gap by assumption. Forty-nine meal premiums appear in the payroll data, totalling about $1,669; where a premium could not be matched to a timekeeping pay period, the answer says so and gives the count both ways.
The exposure range known before the mediation, not after
Running the violation tests under both meal-period thresholds rather than one — which showed that the difference between a 14,073-shift case and a 260-shift case was a legal question about which threshold governs, not a dispute about the data.
Parties, jurisdictions and identifying facts are altered or withheld. Methods, data volumes and outcomes are not.
Recognise your case in this one?
Tell us which part. We will tell you what was actually decisive in it.