Checklist Completion Rate and What Completion Means
Why this matters
Checklist completion rate is filed under quality and it is not a quality number. It measures whether a habit holds: did the person who started a checklist finish it. That is worth knowing, and shops that confuse it for a measure of the work end up defending a number in the high nineties while the callbacks climb. The blind spot is simple and total. A checklist filled in from the truck seat at the end of the day, without a single check being run, is indistinguishable from a perfect one. The rate reads them the same way, because to the rate they are the same event.
What the number is actually made of
Of the checklists started in the window, the share marked finished. Numerator and denominator are both checklists, not jobs, not items and not technicians.
The time anchor is usually the checklist's own start, which is normally the job's creation or scheduling date. That has a consequence worth naming: a checklist started on the last two days of the window and finished on the first day of the next one counts as unfinished here and as nothing at all there. On a monthly read that is a handful of checklists permanently taxed out of every period, which is fine as long as it is consistent, and misleading the one month somebody reads the last week on its own.
The silent exclusion is the big one. A job that finished with no checklist attached is in neither the numerator nor the denominator. It is invisible to this rate entirely. That means a shop can drive completion rate to 100 percent by starting fewer checklists, and the number will congratulate it the whole way. The count of uncovered jobs is a separate figure and it has to be read next to this one. See related: Jobs That Finished With No Checklist at All.
The fork that decides what a high rate means
Everything above depends on one piece of configuration, and it is worth going and checking rather than assuming.
If checklists are attached automatically when a job is created on a job type that carries a template, then a checklist nobody ever opened still exists, sitting at zero items done. It lands in the denominator and drags the rate down. The number can see a never-started checklist, and that is the useful arrangement.
If a technician has to create one, a checklist nobody opened was never created, so it is in neither side. The worse the habit gets, the better the rate looks, and the shop with the weakest discipline posts the strongest number. The tell is a denominator that moves around far more than job volume does: if completed jobs held steady and checklists started fell by a fifth, the rate that rose over the same period rose for the wrong reason.
Find out which one you have before you read the number twice. The test takes one job: create one, do not touch the checklist, close the job, and see whether anything appeared in the denominator.
The two failure modes it genuinely catches
Abandoned partway. A checklist with some items marked and the rest blank is the clearest signal in the whole measure. Somebody started, got interrupted, and never came back, or reached an item they could not complete and stopped there. The second case is the valuable one: when several unfinished checklists all stall at the same item number, that item is the problem, not the technician. It is asking for a reading nobody can take, a tool nobody carries, or access nobody has.
Never started. Only visible when checklists are auto-attached, per the fork above, and it looks different from abandonment: zero items done rather than some. Treat the two separately. Abandonment is a workflow fault and usually has a fixable cause; a zero-item checklist is a habit or a belief that the checklist does not apply to this call.
Both of these are worth acting on, and both are behaviour. Neither tells you a single thing about whether the equipment was left working.
What it cannot catch, and the cross-checks that can
A checklist can be completed without being performed, and no amount of reading the completion rate will separate the two. Three cross-checks will, and all three run off data you already have.
Elapsed time per checklist. Take the time between the first item marked and the last, for the finished checklists only. Build the distribution, not the average. What you are looking for is the left tail: checklists closed in a time that is physically impossible for the checks they describe. Compare it against the checklist's own item count, not against a fixed number of minutes, since a six-item checklist and a twenty-item one should not share a floor.
Uniformity. Real work produces ragged timing. A tech spends four minutes on a measurement and eleven seconds ticking that the area was left tidy. When every item on a checklist carries the same timestamp to the minute, the checklist was filled in one pass. When the same technician's checklists all take almost exactly the same elapsed time across very different jobs, that is the same finding with more steps.
Items that never record a failure. An item that has never once come back bad is either a check that is not needed or a check nobody is running, and you can tell which by going and finding an instance where the condition genuinely was bad. This is the item-level half of the problem and there is more to it than one cross-check can carry. See related: A Checked Box Is Not a Passed Check.
Worked example: one month, corrected
A shop closes a month with 140 completed jobs, 44 of which carried no checklist at all. Those 44 are in neither side of this rate. Templates are attached automatically on the job types that carry one.
Separately, 96 checklists were started in the month, anchored to job creation rather than completion, so those two counts do not add to 140 and should not be read as though they do. Of the 96, 81 were marked finished: a published completion rate of 84.4 percent of the checklists started in the month.
Now the cross-checks, run on the 81 finished checklists.
Elapsed time. Median across the 81 is 38 minutes. Eleven of the 81 closed in under 3 minutes. The template on those jobs carries 16 items, so under 3 minutes is under 12 seconds an item, including the two that require a meter.
Uniformity. On 9 of those 11, the first and last item carry the same timestamp to the minute, meaning the whole checklist was entered in one sitting rather than worked through.
Where they came from. Nine of the 11 short closes came from two technicians, 6 from one and 3 from the other. Those two started 14 and 11 checklists in the month respectively. Those are counts, not rates. A completion or compliance rate read on fewer than about 30 checklists started in the window is a headcount wearing a percentage sign, and at 14 checklists one file moves the figure by more than 7 points. So this points at where to go and look, and it does not rank anybody.
The correction. Treat the 11 as unverified rather than finished. Defensible completion rate: 70 of the same 96 checklists started, 72.9 percent. The published 84.4 percent and the defensible 72.9 percent differ by 11.5 points on the identical 96-checklist base, and the entire difference is checklists that were filled rather than run.
The item cut. Across the 81 finished checklists, one item recorded a failure zero times all month. The shop's own job notes for the same month show two jobs where that exact condition was found bad and corrected on site. On both of those jobs the item is marked done, with nothing recorded. That is not a discipline problem with two technicians, it is an item whose result nobody is capturing.
Two findings, two owners. The 15 checklists that were started and never finished (96 started less 81 marked finished) belong to the driver table below, and at the defensible 72.9 percent the rate sits well under the 90 floor set in the next section, so the first questions are which job type they sit in and which item number they stall at rather than whose name is on them. The 11 filled-in-one-pass checklists are the separate finding, and that one is a conversation held on the elapsed-time evidence rather than on a suspicion. Hold it before anybody reports the month as an improvement over the last one.
The target to set, and what to compare it against
There is no credible cross-shop benchmark for this number and anyone quoting one is quoting their own configuration. A shop that attaches a three-item closeout list to every job and a shop that attaches a twenty-item commissioning sheet to installs only are not measuring the same thing, and the second one will always read lower. So the comparator is your own trailing history, not an industry figure.
Set the standing target at 95 percent of checklists started in the period, and treat anything below 90 as a process fault rather than a people fault until the driver table says otherwise. That target is a starting point to tune: if your templates are long and your call types are short, you may settle at 90 and be running a tighter shop than someone posting 98 on a three-item list. What matters more than the level is that the figure is flat. A five-point move in a month is worth an hour of somebody's time whichever way it went, because a rise is as likely to be the filled-not-run driver as it is to be an improvement.
Two conditions genuinely change how you read the level. If the checklist is customer-facing - left with the customer, attached to the invoice, or used as the basis for a warranty registration - completion is partly a billing gate and the rate will sit high for reasons that have nothing to do with discipline. Read the elapsed-time distribution harder in that case, because the incentive to close the file is stronger. If the checklist feeds a regulatory or manufacturer record, an unfinished one is not a soft failure at all; it is a missing document, and the right handling is a stop rule at job close rather than a number reviewed at month end.
What actually moves this number
Six things move a completion rate, and five of them move it down. The one that moves it up is the one you least want, which is why the rate is read with the elapsed-time distribution beside it rather than alone.
| Driver | Direction | The tell that separates it from the others |
|---|---|---|
| Template attached to a job type it does not fit | Down | Unfinished checklists cluster in one job type and stall at the same item number |
| A technician who does not open them | Down | Unfinished checklists cluster on one person and sit at zero items done, not part done |
| Checklist too long for the call | Down | Unfinished ones are part done across everybody, and elapsed time on the finished ones is a large share of the visit |
| Job closed from the office before field entries landed | Down | The checklist finishes, but its last timestamp is after the job's completion, often the next morning |
| Template edited mid-period | Down | A cliff on one date, and every stalled checklist predates it |
| Checklists filled rather than run | Up | Finished checklists closing in implausibly short or suspiciously uniform elapsed times |
Read the driver table before you read the rate, because three of the five downward drivers are fixed by changing the checklist or the schedule, one by an office rule about when a job may be closed, and only one by talking to a person. Shops reliably reach for the conversation first.
References
- See related: Service Quality Checklist Discipline, and Write a Checklist That Actually Gets Used, for building the checklist this number measures
- See related: A Checked Box Is Not a Passed Check, for the item-level rate and why a failure does not move it
- See related: Jobs That Finished With No Checklist at All, for the coverage figure this rate cannot see
- See related: The Checklist That Grows Every Time Something Goes Wrong