The exceptions you tolerate become your operating model

Firms do not hit a scaling wall because the work dries up. They hit it because the workarounds they tolerated at thirty people are load-bearing at eighty.

This is the least discussed constraint in professional services, because nothing about it looks like a problem while it is forming. It looks like competence.

The mistaken diagnosis

When control gets harder as a firm grows, the reading is usually that the business has outgrown its processes and needs more discipline, or more management, or both. The response is a process programme.

The more accurate reading is that the business never had the processes. It had people who compensated for their absence, and that compensation worked so well that nobody noticed it was the process.

Dysfunction does not arrive with a warning light. It arrives as “just one spreadsheet”, “just one manual workaround”, “ask Sarah, she knows how that one works”, and the professional services classic: “we will tidy that up after this busy period”. Then the busy period becomes the business model.

The four exceptions worth naming

Most of it concentrates in four places, and each one is individually reasonable.

The extra client request that never becomes a variation. Small, quick, and refusing it would have felt petty. It is delivered, absorbed, and never priced.

The approval that happens after the work is done. Not a shortcut in anyone’s mind — the work was obviously going to be approved. But the record now shows an approval that post-dates the effort it authorised, which is not an approval.

The quick favour that becomes part of the retainer. Nobody decided this. It was done once as goodwill, twice out of consistency, and by the fourth time it was an expectation with no line item.

The time entry that arrives late, vague, or not at all. Reconstructed from a calendar a fortnight afterwards, rounded, and entered against whichever engagement seems closest.

None of these is a scandal. That is precisely the mechanism — a deviation that produces no immediate bad outcome is not experienced as a deviation, so the standard quietly moves to accommodate it. Repeat that for two years and the tolerated exception is the documented behaviour, whatever the documentation says.

Why thirty people absorbs it and eighty does not

At thirty people, the compensating mechanism is proximity. The people who know the exceptions are in the room, they can see when one is about to matter, and they carry the context between functions personally.

At eighty, three things break at once. The exception-holders are no longer in every room. New joiners learn the workaround as the method, because that is what they are shown. And the volume of exceptions crosses the point where any individual can hold the full picture.

The symptoms then arrive together, which is why they read as a sudden crisis rather than an accumulation: billing slows, WIP becomes an investigation, delivery argues about scope, leadership stops fully trusting the project financials, and a small number of heroic people become the actual dependency. What is visible in the WIP balance is the aggregate of every exception nobody wrote down.

The mechanism, and what it does not require

The structural answer is not the elimination of exceptions. It is that an exception has to be recorded as one.

In practice: the request that is not a variation is still captured against the engagement, with a value and a decision to absorb it, made by someone with authority to absorb it. The retrospective approval is possible but visible, so the pattern shows up in a report rather than in nobody’s memory. The favour has a named owner and a threshold above which it becomes a conversation. The late time entry has an ageing measure attached to it.

None of that removes discretion. It prices it, and it puts the price where someone can see the total. Our invoicing and WIP page covers how exceptions are held against the item rather than in someone’s head, and the capacity side of the same problem is on the resourcing page.

Where tolerance is right

This argument has an obvious failure mode and it should be said plainly.

Tolerance is not the same as dysfunction. Discretion at the edges is frequently what keeps a good client relationship intact, and a firm that eliminates every exception has usually eliminated its judgement along with them. The instinctive concession that saves a renewal is worth more than the margin it costs, and no policy will identify which concession that is in advance.

There is also a real cost to visibility itself. Recording exceptions means people who have been quietly holding the business together will see their compensating work appear in a report, and some will experience that as surveillance rather than support. How that lands depends entirely on what leadership does with the first report — if the answer is a performance conversation, the recording stops being accurate within a quarter.

So the argument is narrow: not which exceptions to remove, but which are priced and which are absorbed silently. A firm that knows the annual value of its absorbed exceptions can decide to keep absorbing them. A firm that does not know is not making that decision at all.

Start with one. Pick the exception your team mentions most often, put a number against a single month of it, and see whether the number changes anybody’s mind.