What the Timesheet Is Not Telling You

Why this matters

Almost every shop that has worried about theft has worried about parts. Time is the larger number, in almost every shop, by a wide margin: labour is the biggest line in a service business, and half an hour a day across a crew is a payroll cost that never appears as a missing item on any shelf. It is also almost never framed as loss at all, which is exactly why nobody measures it.

What makes this material dangerous is the second fact. Most of the gap you are about to find is not dishonesty, and the instruments you will use to look at it cannot tell you which part is. An owner who reads a telematics report as a confession does more damage in a week than the underlying gap does in a year.

The spectrum, and both ends of it are real

The honest version runs from deliberate at one end to entirely legitimate at the other, and the middle is where most of the hours actually sit.

• Deliberate padding. Hours logged that were not worked. The clearest case and the rarest.
• The long lunch, the personal errand, the extra stop. Real, usually small per instance, and usually a drift rather than a decision.
• The slow-walked job. The tech who knows a two-hour job will not be followed by another one if he finishes at noon. This is a scheduling problem being expressed as a time problem.
• Genuine dead time the shop created. Waiting on a part, a bad route, a dispatch gap, a customer who is not home. The clock is honest and the shop caused it.
• The customer who talked for forty minutes. Frequently the difference between a renewal and a churn, and it looks identical on a report to a tech sitting in a truck.
• The tech who takes longer because he does it right. Torques the connection, pressure-tests instead of assuming, cleans up, shows the customer the reading. The slowest numbers in the shop, and often the lowest callback rate.

The last three are not loss. They are the shop's own cost structure, its customer relationships and its quality, and every one of them shows up in the data as time that is not billable.

What the instruments actually record

This is the section worth keeping. The tools are good at duration and location and blind to intent and difficulty.

Two consequences. First, the instruments corroborate each other or they prove very little: a punch plus a vehicle position plus a customer-visible arrival stamp is a strong record, and any one of them alone is not. Second, no combination of them separates a forty-minute job done properly from a forty-minute job done slowly. That question is answered by the work, not by the clock, and the rest of this card turns on it.

Location tracking also carries an employment-law layer worth knowing before you switch it on: vehicle tracking rules vary by state, several require notice to the employee, and the disclosure and consent mechanics live in their own card rather than this one.

The management trap: the fastest tech as the standard

Take a job type your fastest tech does in 1.4 hours where the crew median is 2.1 hours. The median is 50 percent over the fastest time, which is normal and not a finding.

Now set 1.4 hours as the standard. Everyone at or above the median fails it, and by definition that is half the crew. You have manufactured a shop-wide performance problem out of one person's outlier, and the response you get is not faster work, it is shortcuts, because the only way to hit a number set by the best person in the building is to stop doing some of the job.

Before you believe 1.4 hours is good, check the fastest tech's callback rate on that job type against the crew's. Sometimes it is genuinely good and the finding is that he should be training the others. Sometimes the speed is the callbacks, and you were about to make it the standard.

The usable standard is built off the median or the second quartile, with the top performer treated as a training resource rather than a benchmark.

A worked read of one technician's month

A shop runs a 22-working-day month, eight paid hours a day. The owner runs billable hours per paid hour and stops on one name.

Tech C averages 5.1 billable hours against 8.0 paid. The crew median is 5.9 against 8.0. The gap is 0.8 hours a day, and over 22 days that is 17.6 hours, which is most of two working days of paid time a month. That is the number that makes an owner angry, and it is the number that is about to turn out to mean something different.

Decompose the 0.8 using the day's other components, each compared against the same crew median:

The three differences sum to 0.8 hours, which is the whole gap.

Now read it. Drive accounts for 0.6 of the 0.8, which is 75 percent of the gap, and drive is assigned: Tech C carries the rural quadrant three days a week because the dispatcher gives it to him. That is the shop's routing decision showing up in an individual's productivity number.

What is left that is actually about Tech C is 0.2 hours a day. Over the same 22-day month that is 4.4 hours, and it sits inside the spread of the rest of the crew. The two days a month the owner was angry about were three quarters his own dispatch board and one quarter noise.

Here is the part that matters, and it is where most owners stop too early. This decomposition has proved that the gap is explained. It has not proved that Tech C worked hard, and no further slicing of the same data will. The instruments have said everything they can say. The next question is about the work: first-visit completion rate, callback rate, rework hours, and whether customers ask for him. If those read well, there is nothing here. If they read badly, you have a performance conversation, and it is about outcomes rather than about 0.2 hours a day, which is a number no reasonable person can defend themselves against anyway.

Run the same decomposition on the crew, and the shape changes

The owner in that example went looking at one name. Point the same arithmetic at the whole crew and a much larger number appears.

The median tech bills 5.9 of 8.0 paid hours, so 2.1 hours a day per tech are not billed at all. Across six field techs over the same 22 days that is 277 hours in a month.

Those two figures are not the same kind of thing, and reading them as if they were is the error this whole card is about. The 17.6 hours is one person's gap measured against his own crew. The 277 hours is the crew's total unbilled time measured against a paid hour, and most of it is drive, supply runs and legitimate unbilled work that no shop drives to zero. Still, the scale is the finding: the structural number is roughly sixteen times the individual one, and it moves in response to routing density, supply-run policy, first-visit completion and how many trucks cover a quadrant. None of those are conversations with an employee.

That is the honest ranking of where the hours are. The part that felt like theft was the small one.

What to do with a genuine outlier, in order

1. Check the assignment before the person. Route, job mix, customer mix, truck. Most outliers are assigned into existence, as above.
2. Check whether the rule was ever stated. Whether the drive home in a company truck is paid, whether a supply run is on the clock, whether waiting on a customer is billable. Unstated rules produce padding across the whole crew, evenly, and that pattern is a policy gap rather than a person.
3. Have the conversation curiously, not accusingly. "Your days are running longer than the others on the same job types, walk me through a typical Tuesday" gets you the answer. "Where were you between two and three" gets you a defensive employee and no information, and you cannot unring it if the answer was a parts counter with a queue.
4. Only then treat it as conduct, and only where the evidence is specific, corroborated across two independent sources, and about a described act rather than a ratio.

Where a specific act is alleged rather than a pattern of slowness, the sequence for establishing it before anybody is spoken to is a separate card in this group, and it applies here exactly as it does to parts or cash.

Measure the work instead of the hours

The structural fix, and the one that makes most of this card unnecessary.

Hours are an input. A shop that manages inputs gets people who are good at looking busy. Measure output instead and the time question mostly dissolves: jobs completed per day per tech, first-visit completion rate, callback rate by tech, revenue per field hour at shop level. Those numbers reward the careful tech and expose the slow-walker without anyone needing to review a map.

Two cautions, because this is easy to do badly. Any single output measure will be gamed: rank on jobs per day alone and you get rushed work and callbacks, so you carry completion and callback rate beside it or not at all. And the measures need a floor before they can be read as a ranking, because one bad week on a small number of jobs is noise. A monthly read on a tech with fewer than about twenty completed jobs in the period is a conversation starter and not a finding.

The shops that get this right end up doing less monitoring, not more, and the reason is that they stopped asking a question the timesheet was never able to answer.

References

• U.S. Department of Labor, Wage and Hour Division, on hours worked under the Fair Labor Standards Act, including compensable waiting and travel time
• See related: GPS and Telematics Without Treating Techs Like Suspects, which owns the disclosure and trust half of vehicle tracking
• See related: Time Theft and the Buddy Punch: A Decision Tree, which owns the punch-integrity case
• See related: Before You Accuse Anyone: The Sequence; FLSA Overtime Rules for Trade Businesses