Why does a working process stop working quietly?
Software that breaks is easy to handle. It throws an error, someone phones IT, it gets fixed. The dangerous failure is the one that keeps producing output. The drafts still appear. They look roughly right. They are simply less useful than they were, and the people using them absorb the difference without saying anything.
That absorption is the problem. A fee earner who used to approve a report draft with two small corrections now makes eight, and does not think of it as a fault. They think of it as Tuesday. The hours the rebuild removed creep back, a few minutes at a time, spread across enough people that nobody owns the trend. Six months after go-live the firm believes the process is fine because nothing has gone wrong, and the saving has halved.
So the honest answer is that you do not know unless you decide to look. The looking has to be deliberate, short and scheduled, or it does not happen.
What causes drift in the first place?
Drift is rarely the system behaving differently. It is the world around the system moving while the system stands still. Four causes account for most of it.
- The inputs changed. A client switched reporting format. A funder added a section. The process keeps producing what was right in week four and nobody told it otherwise.
- The templates moved on. Someone revised a standard clause or saved a better schedule in a different folder. This is the decay described in why templates rot, and a process built on a rotting template inherits the rot. Template management exists for this reason.
- The people changed. Those who understood the process best left or moved teams. Their replacements were shown it once, in passing.
- The work changed. A different kind of matter, a larger scheme, a new sector. The process was designed for the old shape of work.
None of these is a reason to abandon the process. All of them are reasons to inspect it on a schedule rather than on a hunch.
What are the early signs worth watching?
The useful signals move before anyone complains, and all four can be read without a dashboard.
How much of the draft survives review. This is the best single indicator. When the process is healthy, the reviewer corrects judgement, not facts and formatting. When it drifts, the reviewer starts rewriting structure. Ask reviewers each month for a rough sense: light, moderate or substantial rework. A shift from light to moderate across several people is a warning long before anyone calls it a failure.
How often people go round it. Every process has a back door: the old spreadsheet, the previous document, the colleague who will just do it for you. Rising use of the back door means the process no longer fits the work, and people have voted with their feet.
How long the job takes now. You measured this before the rebuild. Measure it again the same way. The unbillable hours calculator uses the same arithmetic you started with, so the comparison stays honest.
How many exceptions appear. A stable exception rate is fine. A growing one means the work has changed shape and the process has not caught up.
Why don't people tell you when it stops helping?
Because telling you costs them more than working around it. Raising a problem means a conversation, a meeting, possibly a review, and in the worst case an implication that they are using the thing wrongly. Fixing it themselves takes ten minutes and nobody asks questions.
There is a quieter reason too. People who were cautious about the change have no incentive to report that it is working less well, because they do not want it fixed. Those who championed it have no incentive either, because they do not want to be the one who oversold it. Both sides stay silent for opposite reasons.
So the check has to be a direct question put to named people, not an open invitation to raise issues. "Is anything wrong with the reporting process?" gets a no. "Show me the last report that went out and tell me what you changed" gets the truth. The same holds for the habit covered in the risk in copy and edit.
What should the check actually look like?
Keep it small enough to survive a busy quarter. Three layers work well.
- Monthly, five minutes. One person looks at volume through the process, the rough rework level from reviewers, and any exceptions logged. Three numbers in an email. No meeting.
- Quarterly, one hour. Read a small sample of real output properly and compare it with what the firm would have sent before. Confirm the templates feeding it are still the current ones. Ask the heaviest users what they change every time, because whatever they change every time should be built in.
- Annually, half a day. Revisit the map made in week one. Has the work changed shape? Have client expectations moved? Is the process still aimed at the biggest cost?
Where a process carries regulatory weight, such as compliance evidence or audit trails, the quarterly sample is not optional. Drift there is not a productivity issue. It is a file that will not stand up when someone asks to see it.
Who owns the check?
A process without a named owner will drift, however well it was built. The owner should be the person accountable for the output, not the one most comfortable with technology. For client reporting that is usually the head of the team that signs the reports. For invoice preparation it is whoever answers for the bills. Their job is small: run the monthly look, chair the quarterly hour, decide what changes. They need the authority to change the process, not only to report that it needs changing. A watcher who cannot act produces a record of decline.
What do you do when the check finds something?
Diagnose before you rebuild. Most drift traces to a single upstream change, and the fix is a template update, a new rule for a new client format, or half an hour with someone who joined after go-live. Treating every wobble as total failure is how firms abandon something that was mostly right.
Occasionally the finding is that a process has outlived its design, which is a legitimate outcome, as set out in calling time on a stalled project. More often the finding is that the biggest cost has moved: reporting is fine now and the queue is in bids, or onboarding, or the write-ups. That is progress. Measuring return sets out how to keep score, and the audit will tell you where the hours go today rather than where they went last year.