The Inspection That Stopped Recurring
A weekly scaffold check is set to recur, runs for four months, and then stops — because recurrence happens on completion, not on a date. Nothing is overdue, nothing is flagged, and the series simply is not there any more.
The weekly scaffold inspection was set up properly in March. It recurred, somebody did it every Friday, it was ticked off, and the following week's appeared. For four months this worked exactly as intended.
In July the site agent was away for a fortnight, the inspection was done but not recorded, and the next one never appeared. Nobody noticed, because nothing went wrong: there was no overdue task, no red flag, no notification. The series did not fail. It ended.
This is a property of how recurrence works rather than a defect, and it has a real advantage that comes with a real cost. Knowing which you are getting is the difference between a recurring task you can rely on and one you only think you can.
Two ways to make something recur
Every system that offers recurring tasks picks one of these, and almost none of them say which. They behave identically while everything is going well and completely differently the first time something is missed.
Schedule-driven
- The next occurrence appears on its date, whatever happened to the last one.
- A missed occurrence stays open and becomes overdue.
- Skip three weeks and you return to three overdue tasks waiting.
- The gap is visible — that is the point.
- The cost is a backlog of stale occurrences that people learn to close in bulk.
Completion-driven
- The next occurrence is created when the current one is completed.
- A missed occurrence simply stays open, and nothing follows it.
- Skip three weeks and there is one task, not three.
- The gap is invisible — nothing is overdue that was not already.
- The cost is that a lapsed series looks exactly like a series that has not come round yet.
Recurrence here is completion-driven. A daily job looks for recurring tasks that have been completed and have not yet rolled over, and creates the successor with its dates shifted forward by one interval. A recurring task that has not been completed is not in that set, so nothing is created.
The series did not fail. It ended. Nothing was overdue, because there was nothing left to be overdue.
Why completion-driven is a defensible choice
It would be easy to read the above as a shortcoming, and on a site it very often is not. The alternative has a failure mode of its own, and it is worse in a specific way.
A schedule-driven weekly inspection over a fourteen-month job creates roughly sixty occurrences whether anybody does them or not. Come back from a shutdown and there are four overdue safety inspections for weeks in which nobody was on site. What happens next is universal: somebody closes all four, because closing them is obviously correct and doing them retrospectively is impossible.
And the moment a team has learned to close overdue occurrences in bulk, the overdue flag has stopped meaning anything. That is a more expensive outcome than a series that quietly stopped, because it degrades the signal on every other overdue task on the job.
Completion-driven recurrence never manufactures work that nobody was supposed to do. Its weakness is narrower and more specific: it cannot tell you that a series has lapsed.
What actually gets created
The successor is a copy of the task with a few things deliberately reset, and the details are worth knowing because two of them affect how you should use it.
| Property | On the new occurrence |
|---|---|
| Title, description, assignee, links | Copied from the completed task |
| Status | Reset to open |
| Dates | Shifted forward by one interval, keeping the same duration |
| Baseline dates | Not carried over — each occurrence starts without one |
| Series grouping | Every occurrence points at one root, so the series stays together |
| End of series | Bounded by a recurrence-until date, if you set one |
Intervals are daily, weekly or monthly, with a multiplier — so every third week or every second month are both expressible. And the whole thing is idempotent: a task that has already spawned its successor is stamped, so a job running twice cannot produce two copies of next week.
A recurring task with no dates still terminates
Worth knowing if you set up recurring checklists without dates. A task with no start or due date has nothing to shift, so the next occurrence is seeded from today plus the interval. That matters because it gives the series a date to end on — without one, the until-check has nothing to compare against and a series can run indefinitely. If you use undated recurring tasks, set a recurrence-until date deliberately.
What AWRA OpsHub does today
- Successors are created on completion, by a job that runs daily, so a completed recurring task rolls forward without anybody remembering to.
- Daily, weekly and monthly intervals with a multiplier, so every third week or every second month are both expressible.
- Dates shift forward by one interval and the duration is preserved, so a three-day task stays a three-day task.
- Idempotent — a task that has spawned is stamped, so a job running twice cannot produce a duplicate.
- The series is grouped under one root, so occurrences stay together rather than becoming unrelated tasks with similar names.
- Bounded by a recurrence-until date when you set one.
What it does not do
- A series that is never completed simply stops. There is no occurrence created on a schedule, so a lapse produces no overdue task and no notification — the series is absent rather than late, and absence is not something anything reports on.
- No alert when a series goes quiet. Nothing watches for a recurring task that has had no occurrence in three intervals, which is precisely the condition worth knowing about.
- Baselines are not carried forward. Each occurrence starts without a baseline, so variance is measured within an occurrence rather than across the series.
- No compliance view of a series. How many occurrences should have happened in a period versus how many did is not a figure the system produces — and it is the figure an auditor asks for.
Not ours, by choice
- We will not manufacture overdue occurrences for periods nobody worked. A schedule that creates four overdue safety inspections for a fortnight when the site was closed teaches a team to close things in bulk, and that habit costs you the meaning of every overdue flag on the job.
- We will not auto-complete a recurring task to keep a series alive. A tick on a safety inspection that nobody performed is worse than a gap, because the gap is honest.
An alert when a recurring series has produced no occurrence for a configurable number of intervals, and a per-series compliance figure of expected against actual, are both scope rather than ceilings. The series grouping, the interval arithmetic and the scheduled job all exist and work.
Until a lapse alert exists, the practical control is a monthly review of your recurring series — not the tasks, the series. Ask which ones have produced an occurrence recently, because the ones that have not are the ones that ended without telling anybody.
Using it for things that must not lapse
Recurring tasks are ideal for the rhythm of a site, and they need a wrapper when the rhythm is a safety or contractual obligation rather than a good habit.
-
Separate the two kinds of recurring work
A weekly tidy of the site office and a weekly scaffold inspection are not the same category. One can lapse harmlessly; the other cannot. Label them differently and treat the second group as a list you actively review.
-
Set an until date on every series
Usually practical completion or the end of the contract period. It costs nothing, it stops orphaned series accumulating across finished jobs, and it makes the series a bounded thing rather than an indefinite one.
-
Review the series monthly, not the tasks
Open your recurring safety and compliance series and ask which have produced an occurrence this month. This is the step that substitutes for the lapse alert, and it takes ten minutes.
-
Complete on the day, not on the Friday
Because the successor is created from the completion, a batch of Friday completions produces a batch of next-week occurrences all dated from Friday. The dates drift toward whenever people tick things rather than whenever the work happens.
-
Never tick one that was not done
Worth stating explicitly to a site team, because the incentive is real: ticking it keeps the series alive. A gap in a safety series is a fact worth having; a tick on an inspection nobody performed is a document that will be read out at exactly the wrong moment.
The fourth item is the subtle one and it accumulates. If a weekly check is nominally due on Monday but completed each Friday, the next occurrence is dated from Friday plus a week, and within two months the "weekly Monday inspection" is a Thursday task that nobody decided to move.
What an auditor will ask
It is worth setting expectations here, because recurring tasks look like a compliance record and are not quite one.
The question is always some version of: over this period, this check should have happened this many times — show me that it did. Answering it requires an expected count and an actual count, and the system holds the second but not the first, because a completion-driven series has no view of occurrences that never existed.
What you can produce is a list of occurrences with their completion dates and who completed them, which is genuine evidence and is usually accepted. What you should not do is present a series with no gaps as proof that no check was missed — the absence of a gap in a completion-driven series can equally mean the series stopped and nothing was expected after that.
Before an audit or a site inspection
- Open each safety and compliance series and confirm it has a current open occurrence. A series with none has ended.
- Check the spacing of completion dates. Drift toward the end of the week is normal; a two-month gap is a lapse.
- Confirm the assignee is somebody still on the job. A series assigned to a person who left continues to spawn and is nobody's work.
- Check the until dates. A series past its end date has stopped legitimately and should not be read as a lapse.
- For anything genuinely contractual, keep the signed paper record as well. The task is evidence that it was done and it is not the inspection certificate.
The scheduling logic that these tasks sit inside is in the chain that decides your finish date, the certificate-expiry version of the same lapse problem in the certificate expired and the work carried on, and the post-completion obligations in the twelve months after you leave site.
Our take
Completion-driven recurrence is the right default for site work — it never invents overdue inspections for a fortnight when nobody was on site, and that matters because bulk-closing overdue tasks is a habit that destroys the meaning of the flag everywhere else. The cost is specific: a series that lapses is absent rather than late, and nothing reports absence. So until a lapse alert exists, review your safety and compliance series monthly rather than reviewing the tasks, set an until date on every one, and never tick something to keep a series alive.
See recurring tasks on a live job
Daily, weekly and monthly intervals with a multiplier, successors created on completion with the duration preserved, occurrences grouped under one series, and an end date you set.
Explore task managementFrequently asked questions
When is the next occurrence of a recurring task created?
When the current one is completed. A job runs daily looking for recurring tasks that have been completed and not yet rolled over, and creates the successor with its dates shifted forward by one interval and its duration preserved. The consequence worth understanding is that an uncompleted recurring task spawns nothing — the successor is created from the completion, not from a date, so nothing appears until somebody ticks the current one.
What happens if we miss a week?
The current occurrence stays open, and nothing follows it. You do not accumulate overdue occurrences for the weeks you missed — you have one open task rather than four. The advantage is that a fortnight's shutdown does not produce four overdue safety inspections for a period when nobody was on site. The cost is that a series which lapses entirely is absent rather than late, and nothing reports absence, so it looks the same as a series that has simply not come round yet.
Will we be told if a recurring series stops?
No. Nothing watches for a series that has produced no occurrence in several intervals, which is exactly the condition worth knowing about. Until that exists, the practical substitute is a monthly review of your safety and compliance series — not the individual tasks, the series — asking which have produced an occurrence recently. It takes ten minutes and it is the only thing standing between a lapsed series and an audit.
Why is our weekly Monday check drifting to Thursdays?
Because the next occurrence is dated from the completion rather than from the original schedule. If a nominally-Monday check is ticked each Friday, the successor is dated a week from Friday, and within a couple of months the series has walked to the end of the week without anybody deciding to move it. Completing on the day the work happens keeps the dates honest; if the drift has already happened, edit the dates on the current occurrence and the series will follow from there.
Can we use this as our safety compliance record?
As supporting evidence, yes — you get a list of occurrences with completion dates and who completed them, which is genuine and usually accepted. As proof that nothing was missed, no. An auditor asks how many times a check should have happened against how many times it did, and a completion-driven series has no view of occurrences that never existed. Specifically, do not present a series with no gaps as proof of full compliance, because the absence of a gap can equally mean the series stopped.
Do recurring tasks ever end on their own?
Only if you set a recurrence-until date, which is worth doing on every series — usually practical completion or the end of the contract period. It costs nothing and it stops orphaned series accumulating across finished jobs. It matters particularly for undated recurring tasks: a task with no dates has nothing to shift, so the next occurrence is seeded from today plus the interval, and without an end date there is nothing to stop the series continuing indefinitely.
