Measuring wrench time without a stopwatch
Formal work sampling is disproportionate for most departments. Lighter methods give a good enough figure and do far less damage to trust.
The classical method for establishing wrench time is work sampling: an observer records, at random intervals, what each technician is doing, and the proportions are inferred from a few hundred observations. For a separate product-side perspective on distinguishing useful time data from artificial activity, this product guide provides additional detail.
It is statistically sound and, in a department of six people who all know each other, it is socially catastrophic. There are lighter methods that produce a figure good enough to act on.
Self-recorded sampling
For a defined period — two weeks is usually enough — technicians record their own time against the categories on a simple sheet or in a phone form. Wrench, travel, waiting, preparation, admin. For a wider operational and compliance reference, consult OSHA maintenance and reliability resources.
The objection is that self-reported data is unreliable. In practice it is reliable enough for this purpose, provided two conditions hold: the categories are few and unambiguous, and the result is reported only in aggregate. What you need is whether waiting is five percent or twenty-five, and self-reporting distinguishes those comfortably.
Every additional category reduces compliance and increases misclassification. If you cannot decide which bucket something goes in, neither can the person filling in the sheet.
Job timestamps you already hold
Where jobs are started and completed in a system, the interval between them is already recorded. This is noisier than deliberate sampling — it includes whatever happened during the job — but it is free and continuous.
The useful comparison is between that interval and the estimated task time. The difference across many jobs is the non-task overhead, and while it does not break down into categories, it does give you the total and its trend.
A walk-round census
The blunt version, used once to get a baseline: at several random moments across a fortnight, note what each available technician is doing. Twenty or thirty observations give a rough split.
This is closest to classical work sampling and needs the most careful handling. It should be announced, its purpose explained, and its results shared with the team before anyone else sees them. Done covertly it will be discovered, and the department will spend the next two years unable to collect any time data at all.
Establish the baseline before changing anything
The temptation is to fix the obvious problems immediately and measure afterwards. That leaves no way to demonstrate that the changes worked, which matters when the next request is for a stores clerk or a key cabinet.
Measure first, publish the figure internally, make one or two changes, measure again in six months. The comparison is the argument for further investment, and it is far more persuasive than an assertion that things feel better.
Expect the number to be uncomfortable
A first measurement commonly lands lower than management expects and roughly where technicians expected. That discrepancy is itself useful information — it usually means the team has been aware of the structural losses for years without a mechanism for reporting them.
The response determines everything that follows. If the reaction is to question effort, the exercise is over. If the reaction is to ask which of the losses is biggest and what would remove it, the team will typically supply a detailed and accurate answer immediately, because they have been living with it.
Re-measure on a slow cycle
Annually or after a significant change — a new site, a new system, a reorganisation of stores. Continuous measurement of wrench time is neither necessary nor healthy; it converts a diagnostic tool into a monitoring one, with the predictable effect on data quality.