How it varies by context
At a single, established site, consistency issues most often trace to informal adaptation that has become the real local procedure without anyone updating the written one — a sequence changed years ago to work around a specific piece of equipment, now so embedded that nobody thinks of it as a deviation at all. Erik Hollnagel, Robert L. Wears and Jeffrey Braithwaite’s research on the difference between work-as-imagined and work-as-done is directly relevant here: “Things do not go well because people simply follow the procedures and work as imagined. Things go well because people make sensible adjustments according to the demands of the situation.” At a single site, this adaptation is usually invisible precisely because it works — the site hits its numbers, and the gap between document and practice never becomes anyone’s problem to solve.
Across a multi-site network built organically — where each site has grown its own local expertise over years — the variation tends to be larger and harder to trace, because each site has developed its own version of “sensible adjustment” independently, and those versions have never been compared against each other, only against each site’s own written procedure. Two sites can each be locally optimised and still produce meaningfully different output quality, because the specific adjustments each has made reflect different equipment histories, different supplier relationships, and different individual expertise that never had reason to be reconciled.
At a newly acquired site, the pattern is different again: the written procedure may genuinely be followed closely, but it was designed around the acquiring firm’s typical conditions, which the new site’s equipment, suppliers or workforce may not match — so faithful adherence to the document produces worse results than the informal adaptation a more established site would have made instinctively.
A fourth context worth distinguishing is the fast-growing single company operating several very similar sites built at the same time, using the same design. Even here, where the starting conditions are genuinely as close to identical as manufacturing allows, variation tends to emerge within the first year of operation as each site’s specific team develops its own working rhythm — and because these sites started identically, leadership is often slowest to suspect adaptation as the cause here, precisely because the “everything is the same” assumption is strongest exactly where it is least examined.
A fifth, distinct context is the site operating under a recently updated corporate standard. Immediately after a standard is revised, every site is, briefly, working from the same up-to-date document — but the older, more established sites often have more deeply embedded prior habits to unlearn than newer sites do, which means the initial period after any standard update can produce a temporary widening of variation before it narrows again, purely because different sites are starting the transition from different distances.
Why it happens
Ronald Coase’s foundational observation that coordinating activity within a firm carries a real cost helps explain why this variation persists rather than being corrected quickly. Reconciling every site’s local adaptation against every other site’s, and against the written standard, is itself a coordination task — one that competes for attention against the more visible, immediate demands of running each site’s daily output. Unless someone is specifically charged with this reconciliation, the cost of doing it is easy to defer indefinitely, because no single site’s own reporting ever shows the comparison that would reveal the gap.
This deferral tends to be reinforced by how most manufacturers structure site-level accountability. A site manager is typically measured against that site’s own targets and its own historical performance, not against how closely its practice matches a sister site three hundred miles away. There is, in most reporting structures, no natural moment at which anyone is specifically asked to look sideways across sites rather than backward across one site’s own history — which means the reconciliation Coase’s framework identifies as costly is also, structurally, nobody’s job by default.
A related structural factor is that site visits and audits, where they happen, are usually scheduled against a compliance checklist rather than built to surface genuine cross-site comparison. An auditor checking whether site B follows its documented procedure is answering a different question from an auditor checking whether site B’s actual practice matches site A’s actual practice — the first is common, the second is rare, and it is the second question that would actually reveal the pattern this article describes.
Where this shows up
This perspective is written with manufacturers in mind. It sits within the Asset and Production-Based grouping, and the questions it raises recur in other sectors too.
Questions worth considering
- Which of the current site-level procedures reflect what is actually done rather than what was originally written?
- Where have two sites independently developed different adaptations to what was meant to be the same process?
- Is anyone currently responsible for comparing practice across sites, rather than each site being measured only against its own history?
- For sites built at the same time to the same design, has anyone actually verified they still operate the same way, or has that assumption simply never been tested?
Your team is facing similar challenges?
Subscribe to The Riddle Ledger
Perspectives are sent as they are published, with no promotional email, and you can unsubscribe at any time.