Positive deviance: learning from what already works

The previous article covered four methods for finding root causes, mapping systems, foreseeing risks, and identifying the real constraint. All four are aimed at the same thing: what went wrong. That's also what most organizations reflexively do when a problem shows up – assemble a review team, sometimes literally called a "SWAT team," to autopsy what failed.

That's a reasonable first instinct. But it's also, on reflection, an odd choice of where to spend your analytical time.

Wrong was foreseeable. Right wasn't.

In most operations, especially during change, the probability that something actually works – under the same constraints, the same budget, the same systems the rest of the organization is wrestling with – is considerably lower than the probability that it doesn't. That's exactly why change is hard. But it also means something rarely noticed: the rare cases where something does work carry more information than an average failure does.

A failure could have any one of ten different causes, and autopsying it rarely reveals which. But a success that emerged under exactly the same conditions as the failures around it is, in practice, a proof of existence: it's possible, here's how, under these conditions, with these resources. That's harder to argue away than a theoretical proposal.

The method: positive deviance

This isn't just a theoretical argument. It's the foundation of a method called positive deviance, developed by Jerry and Monique Sternin during an assignment for Save the Children in Vietnam in the early 1990s.

The assignment was to reduce malnutrition among rural children, with a six-month timeframe and a limited budget – too short to address the underlying causes (poverty, poor sanitation, weak agriculture) usually cited as the explanation. Instead of importing an external solution, Sternin asked a different question: were there, in these very villages, children who were well-nourished despite their families being just as poor as everyone else's? The answer was yes. By comparing what these families' caregivers actually did differently, the team found they fed their children more often but in smaller portions, and mixed in shrimp and crabs from the rice paddies – food considered unsuitable for children by local norms, but both accessible and nutritious. The solution already existed in the village. It just needed to be found and spread, not imported from outside.

The method generalizes into three steps: identify "positive deviants" – individuals or teams performing markedly better than comparable peers under the same constraints; find out concretely what they actually do differently, not what they say they do; and spread it by letting others in the same situation discover and try it themselves, rather than lecturing about it from above.

Why it works better than imported best practice

What makes positive deviance powerful compared to bringing in a solution from outside – a consulting framework, an industry standard, "what the successful office in another country does" – is that the positive deviance has already proven it works under exactly the constraints that apply here: the same budget, the same systems, the same culture, the same customers. An external best practice always has to be translated and adapted, and in that translation, the specific detail that made it effective often gets lost. An internal positive deviance requires no translation.

It also solves one of the most underrated obstacles to change: resistance to being told what to do. Pointing to a neighboring team, a sister department, or a colleague who has already solved the problem is a completely different conversation than a consultant or leadership team presenting a solution from above – even when the solution happens to be identical.

Where to find the positive deviants

In practice, the method requires comparing units that should perform similarly but don't: branches, sales teams, production lines, care units, project teams. If one of them consistently outperforms on a measure that matters – without more resources, without more favorable conditions – it's worth a closer look, no matter how unglamorous the measure is.

Worth watching for is the difference between a genuine positive deviance and one that only looks like one, for reasons that can't be copied: better resources that don't show up in the budget, a single unusually skilled individual rather than a replicable practice, or simply random variation that statistically always produces some outcomes at the top even without any real difference. Telling these apart requires the same kind of rigor as the root cause analysis from the previous article – just aimed in the opposite direction.

A reminder, not a universal fix

Positive deviance doesn't solve everything. It requires that a positive deviance actually exists to find – that someone, somewhere in the organization, has already cracked the problem under comparable conditions. That's far from always true. Sometimes the whole organization, the whole industry, is in the same boat, with no internal or external model to lean on.

That brings us to the final, and perhaps most uncomfortable, truth in this series: that some problems can't be solved by analyzing either what went wrong or what went right, because what's about to happen has never happened before. That's the subject of the closing part of the series.