The Exception Becomes the Rule

Category: Organisations and systems A one-off allowance, made under pressure, quietly hardens into standard practice.

It started, as these things do, with a reasonable yes.

A customer needed an order shipped before their cut-off, and the warehouse had already closed its picking for the day. The account manager asked the team lead whether they could run one more pick, just this once, off-cycle. The team lead looked at the clock, weighed the cost of disappointing a good customer against the cost of an hour’s overtime, and said yes. It was the right call. The order went out, the customer was grateful, and everyone went home feeling like the kind of company that looks after people.

Nobody decided anything that day except to ship one order late. There was no policy written, no precedent declared, no meeting held. There was just a sensible person making a sensible exception to a sensible rule, under pressure, because the rule had met a case it handled badly and a human being chose to override it. That is exactly how exceptions are supposed to work.

A fortnight later the same account manager came back. Different customer, similar bind. “Could we do the off-cycle pick again — like we did for the other order?” And here is the moment the whole chapter turns on, because it is so small you could watch it happen and not see anything occur. The team lead said yes again. Not because the second case was as strong as the first — it wasn’t, particularly — but because saying no would have meant explaining why this customer didn’t qualify for something the last one got. The first yes was now sitting in the room, pointing at him, asking why the rule applied today when it hadn’t applied a fortnight ago.

He did not have a good answer to that, because there wasn’t one. The honest answer was “the first time was genuinely exceptional and this isn’t,” but that answer requires drawing a line in public that says you don’t qualify and they did — an unpleasant thing to do for the sake of a rule nobody is currently defending. So he said yes, and now there were two.

Six months on, the off-cycle pick is simply how the warehouse runs. It has a slot. It has, God help them, a name. New starters are taught it as part of the job. The picking cut-off is still written on the wall, and it still says what it always said, but it no longer describes what happens, because the exception has eaten it. Somewhere in those six months the company quietly committed to a standing capability — staffed hours, expectations, a thing customers now assume — that no one ever decided to take on. There was never a yes to that. There was only a yes to one order, and then a yes that couldn’t say no to the first yes, and then a slope.

The Principle

An allowance granted once under pressure does not stay an allowance. The second request points at the first, the third points at the second, and refusing any of them means being the person who revokes a precedent in public — so nobody does. What was justified once, for a real reason, becomes load-bearing standard practice without anyone ever deciding it should be.

The mechanism has nothing to do with whether the first decision was right. The first decision is usually right; that is what makes the whole thing so hard to catch. The exception was granted because reality threw up a case the rule handled badly, and a sensible person sensibly overrode it. The damage is not in the granting. The damage is in what the granted exception becomes once it exists — a reference point, a thing that can be pointed at, a fact on the ground that the next decision has to argue against rather than for.

A rule asks “should we?” An existing exception flips the question to “why not, when we did last time?” — and those are not the same question. The first one weighs the case on its merits. The second one puts the burden of proof on refusal, in front of an audience, against a thing that already happened and worked out fine. That flip is the entire phenomenon. Once it has happened, the exception is no longer competing with the rule. It is the rule, and the rule is just decoration.

Why It Is Inevitable

It is inevitable because revoking a precedent is a fundamentally different, and worse, act than declining a new request — and the difference is borne entirely by one person at one moment.

Declining a fresh request is clean. You weigh it, you say no, and the person who asked has no standing to feel singled out, because nobody got the thing. But once the exception exists, declining the next request is no longer a refusal — it is a withdrawal. You are not saying “no, we don’t do that.” You are saying “we do that, but not for you,” and that sentence is almost impossible to say without it sounding like a judgement on the person, or a reversal you owe them an explanation for. The first yes converted every future no into a confrontation. Few people will pick that confrontation over a rule that, in the moment, nobody is paying them to defend.

It compounds because the cost structure is lopsided in exactly the way that guarantees drift. Granting the exception again is cheap, immediate, and pleasant — a grateful customer, a smooth interaction, a colleague who got what they asked for. Holding the line is expensive, immediate, and unpleasant — an awkward refusal, a bruised relationship, the faint accusation of being unhelpful or bureaucratic — in exchange for a benefit (the rule staying real) that is diffuse, delayed, and invisible. You cannot point to the chaos you prevented. You can only feel the friction you caused. So the daily incentive, every single time, points toward yes, and a thing that gets a yes every time is not an exception. It is the new normal, arriving one defensible yes at a time.

And it is invisible while it happens, which removes the one force that might stop it. Nobody ever convenes the meeting where the company decides to take on the standing off-cycle pick, because there is no such meeting — the capability assembled itself out of individual decisions, each small enough to make alone and reasonable enough to defend on its own. There is no moment where someone with the authority to say “wait, are we sure we want to own this forever?” is even presented with the question. It never gets asked, because at no single point does anyone do anything that looks like committing to a standing practice. They just say yes to one more order, the way they did last time.

None of this needs a weak manager or a careless culture. The people involved are behaving well. The first granter was right. The second granter was kind, and reasonably reluctant to humiliate a colleague over a cut-off. Everyone downstream simply inherited a practice that already existed and treated it, correctly, as how things are done here. The system produced a commitment nobody chose, out of a sequence of choices every one of which was sound. That is why it is inevitable: the drift is not a failure of the people, it is a property of the slope, and the slope is there whether or not anyone is paying attention.

How It Shows Up

• A practice that everyone follows and nobody can point to a decision for — “we’ve always done it this way,” where “always” turns out to mean “since one Tuesday eight months ago.”
• The phrase “but we did it for X last time” ending an argument that should have been weighed on its own merits.
• An exception with no expiry, no review, and no record of who granted it or why — so there is no longer a way to tell it apart from policy.
• A rule still written down and still cited that no longer describes what actually happens.
• New starters being taught the workaround as though it were the process, with no memory that it was ever exceptional.
• A standing commitment — staffed hours, a customer expectation, an SLA in all but name — that the business is clearly carrying but never formally took on.
• The original justifying reason long gone, while the practice it created carries on unbothered by its own disappearance.

Why It Causes Damage

The damage is not the second pick, or the tenth. The damage is that the organisation is now carrying a load it never chose to carry, and therefore never resourced, priced, or defended.

A decision that is made deliberately gets the things deliberate decisions get: someone weighs whether it is affordable, someone budgets for it, someone owns it and can later change their mind. A commitment that assembled itself out of un-revoked exceptions gets none of that. It is real — customers depend on it, staff plan around it, the cost is genuinely being paid — but it exists below the level at which anyone could decide to stop. You cannot defund a thing that was never funded. You cannot review a thing that was never proposed. The off-cycle pick is now a fixed cost of running the warehouse, and it got there without ever passing through the part of the company that decides what the warehouse does.

It is worse than an honestly-made bad decision, because a bad decision at least had an author and a moment, and can be found and reversed. This has neither. There is no document to amend, no owner to persuade, no meeting to reconvene. To unwind it, someone has to do the very thing the whole mechanism was built to prevent — stand up and revoke a precedent that everyone now treats as normal, against a wall of “but we’ve always done this.” Precedents do not get cheaper to revoke with age. They get dearer, because they accumulate dependents.

And the mechanism is self-fuelling, which is the quiet horror of it. Every exception you fail to revoke strengthens the next exception’s claim, because now there are two prior cases pointing at the refusal instead of one. The reference set only ever grows. So the slope steepens precisely as you slide down it: the more the rule has been bent, the harder it is to hold, until the written rule and the operating rule have parted company entirely and the organisation is being run by a practice it cannot find the decision for. It looks, from the inside, exactly like stability — everyone agreeing on how things are done. It is the opposite. It is a commitment with no brakes, because the only place a brake could go is the one place no individual is willing to stand.

How To Counter It

You cannot counter this by being more careful about the first yes, because the first yes is usually correct and refusing it would cost you the case the exception was for. The leverage is not in the granting. It is in refusing to let the exception become a precedent — in stripping it of the one property that lets it harden, which is the ability to point at itself next time.

• Mark the exception as an exception, out loud and on the record, the moment you grant it. “Yes, this once, because of X” is a different act from a silent yes — it tells everyone in the room that a rule was bent rather than abolished, and it denies the next request the cover of “but you did it before.” An unmarked exception is just an undocumented policy change.
• Give every exception an expiry or a condition by default. “Until the system is fixed,” “for this order only,” “this week.” A thing that has to be deliberately renewed cannot silently become permanent, because permanence now requires somebody to actively choose it — which is exactly the choice you wanted made in the open.
• Make the second request argue on its own merits, not against the first. When you hear “but we did it last time,” that is the precedent doing its work — name it. “Last time was genuinely exceptional; tell me why this one is,” resets the burden of proof back where it belongs, onto the case in front of you, instead of onto your refusal.
• Count the exceptions and treat a rising count as the rule failing, not as friction to smooth away. If the same exception keeps getting requested, the rule is wrong or incomplete — fix the rule deliberately, in daylight, rather than letting a hundred quiet yeses fix it for you in a direction nobody chose.
• Find the standing commitments that nobody decided to make, and force the decision retroactively. Once a year, ask of any settled practice: where is the decision for this? If you cannot find one, you have an un-chosen commitment — surface it, put it in front of whoever should have decided, and let them either own it deliberately or end it. Make the implicit explicit so it can at last be governed.
• Protect the person who holds the line. The team lead who says “no, not this time” is doing the rule’s most important and most thankless work, and the mechanism punishes them with the confrontation it spares everyone who says yes. If you want the line held, the person holding it has to be visibly backed, not left to absorb the awkwardness alone.

What Good Looks Like

Good looks like an organisation that can tell the difference between a rule, an exception, and a decision — and never lets one quietly turn into another.

It looks like exceptions that are granted freely when reality demands them, because the people granting them know the exception will not metastasise: it is marked, it is dated, it will expire unless someone renews it on purpose. It looks like “we did it last time” being treated as a fact to be examined rather than an argument that wins, and like the practices the organisation follows all being traceable to a decision somebody can name. It looks like a periodic, unsentimental audit that goes looking for the commitments nobody chose — and drags them into the light to be either properly owned or properly ended.

It does not look like an organisation with no exceptions; that organisation would have shattered against the first hard case it could not bend for. It looks like one where bending and committing are different, visible acts — where a yes to one order stays a yes to one order, and a standing practice exists only because somebody, somewhere, actually decided to take it on, knew what it cost, and could change their mind. The rule survives its own exceptions, and the business carries only the load it chose to carry. The slope is still there; the difference is that someone is standing on it on purpose, with a record, and a date, and a name.

A Reflective Question

Think of something your organisation now does as standard — a service, an allowance, a way of working that everyone treats as simply how things are. Can you find the decision to do it, with a person and a date attached? Or did it arrive the way the off-cycle pick did, one defensible yes at a time, until it was load-bearing and nobody could remember choosing it — and if so, who would have to stand up and revoke it now, and have you made sure that person isn’t standing there alone?