What the research finds
V. F. Ridgway’s research on the dysfunctional consequences of performance measurement, published in Administrative Science Quarterly, found that single, multiple and composite performance measures can each produce behaviour that undermines the outcome they were meant to protect — his documented examples include employment interviewers who, evaluated on interview volume, conducted faster interviews and placed fewer people, and investigators who, given a monthly case quota, selected easier cases as the deadline approached. Ridgway’s own warning remains the clearest single statement of the risk: “Quantitative measures of performance are tools, and are undoubtedly useful. But research indicates that indiscriminate use and undue confidence and reliance in them result from insufficient knowledge of the full effects and consequences… The cure is sometimes worse than the disease.” Applied to a multi-site manufacturer, a standard that each site is measured against can itself generate the behaviour that produces inconsistency, if what is measured does not fully capture what actually matters.
The mechanism Ridgway describes is worth spelling out for a multi-site context specifically: where each site is measured primarily on its own throughput or its own cost-per-unit, sites have a structural incentive to solve local problems locally, in whatever way most efficiently protects their own numbers — which frequently means the fastest available fix rather than the one most consistent with how the standard prescribes it, and each such fix, individually invisible to central reporting, is a small contribution to the aggregate drift the standard was meant to prevent.
Nelson P. Repenning and John D. Sterman’s research on the capability trap, published in California Management Review, offers a second, largely independent line of evidence: organisations under pressure default to working harder — more correction after the fact — rather than working smarter, in ways that are self-reinforcing because visible firefighting is rewarded while invisible prevention is not. A site absorbing the cost of inconsistency through constant local correction, rather than through a genuine fix to what caused the inconsistency, is exhibiting exactly this pattern, and the correction work itself becomes a hidden cost that never appears as a line item tied to the original inconsistency.
Applied across a multi-site network, this dynamic has a compounding property that a single-site view would miss entirely: if each site is independently caught in its own version of the capability trap, each is independently under-investing in the kind of process improvement that would also, incidentally, bring it closer into alignment with every other site — which means the capability trap is not simply costing each site individually, it is actively working against the network-wide consistency the shared standard was meant to achieve.
Erik Hollnagel, Robert L. Wears and Jeffrey Braithwaite’s research on the gap between work-as-imagined and work-as-done offers a third perspective, and a genuinely different one: some of what looks like costly inconsistency between sites is local adaptation that is actually functional, keeping each site running under conditions its written standard never anticipated. Their finding — that “things go well because people make sensible adjustments according to the demands of the situation” — complicates any simple accounting of inconsistency as pure cost, since eliminating it indiscriminately risks removing exactly the adaptation that was quietly protecting output.
Ronald Coase’s account of why coordinating activity inside a firm carries a real cost provides a fourth, more structural lens: reconciling standards across sites is itself coordination work, and that work competes for attention against each site’s more immediate, visible demands. Where nobody is specifically resourced to do this reconciliation, its absence is not a decision anyone made — it is simply what happens by default when a genuine cost has no natural owner.
Coase’s lens also helps explain why the other three mechanisms compound rather than simply add together. A measurement design that distorts site-level behaviour (Ridgway), a capability trap that consumes prevention capacity with firefighting (Repenning and Sterman), and functional local adaptation drifting from documentation (Hollnagel, Wears and Braithwaite) would each, individually, be a containable problem for an organisation that actively invested in reconciling site practice on an ongoing basis. Coase’s point is that this reconciliation is not free, and in a growing, resource-constrained manufacturer it is often the first coordination cost to be deferred precisely because deferring it produces no immediate, visible consequence — until the other three mechanisms have compounded for long enough that the eventual cost of reconciling is considerably higher than it would have been at the outset.
Where they agree, where they diverge
All four converge on a single underlying point: the total cost of inconsistency across sites is larger than what any individual site’s own reporting captures, because each piece of that cost — distorted measurement, hidden firefighting, unreconciled local adaptation, deferred coordination work — sits structurally outside what a single site is positioned to report on itself. None of these costs belongs cleanly to one site; each exists in the space between sites, which is precisely where organisational reporting tends to be weakest.
Where the research diverges is on what to do about the inconsistency once its cost is visible. Ridgway’s and Repenning and Sterman’s findings both point toward inconsistency as something to actively reduce, through better-designed measures and more deliberate investment in prevention. Hollnagel, Wears and Braithwaite’s findings complicate that conclusion directly — some proportion of site-to-site variation is functional adaptation rather than a defect, and removing it without understanding why it exists can degrade the very outcomes consistency was meant to protect. This is not a contradiction the research resolves; it is a genuine open question that depends on distinguishing, site by site, which specific instances of inconsistency are costly drift and which are legitimate, situation-specific adaptation — a distinction none of the four studies, individually, provides a general method for making.
What remains unresolved
None of the four bodies of research, individually or together, specifies how a leadership team should distinguish functional adaptation from costly drift in their own specific case — that judgement remains organisation-specific. What each does establish, reliably, is that the true cost of inconsistency sits mostly in the gaps between sites rather than inside any one of them, and that a leadership team relying solely on each site’s own reporting is very unlikely to see the total cost accurately, however diligent any single site’s reporting happens to be.
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
- Where standards differ between sites, which differences are costly drift and which are sensible adaptation to local conditions — and who decides?
- What does each site’s own reporting leave out about the effort spent firefighting or working around those differences?
- How long has the work of reconciling them been deferred, and what would it cost now compared with a year ago?
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.