Numerical targets and what they produce

In 1954 Peter Drucker formulated management by objectives, which consists of aligning individual and organisational goals through discussion rather than through prescribing tasks. The original idea stresses the person's participation in defining their own objectives.

The device spread widely and went through many reformulations, up to the objectives and key results methods popularised by technology firms from the 2000s. In many implementations, the discussion has shrunk to a target set in advance, which takes practice away from its original formulation.

Edwin Locke and Gary Latham accumulated over several decades experiments showing that a precise and difficult goal produces higher performance than a vague instruction, on simple and measurable tasks.

That work is among the most replicated in work psychology. Its authors themselves note that the effect fades on complex tasks and where the execution strategy is not known in advance. In those situations, a learning goal often yields better results than a numerical outcome goal.

A critique published in 2009 by Lisa Ordóñez and colleagues, under the title "Goals Gone Wild", catalogues the undesirable effects of overly narrow goals. Locke and Latham replied by disputing the scope of those examples, and the debate has since been about conditions rather than the principle.

Formulated by the economist Charles Goodhart about monetary aggregates, it states that a measure ceases to be a good measure as soon as it becomes a target. The common formulation, more general than the original statement, is due to the anthropologist Marilyn Strathern.

A neighbouring formulation, due to Donald Campbell, concerns social indicators and predicts their corruption as they are used for decisions. Both describe the same mechanism in different fields.

Work on indicators in public services, such as that of Gwyn Bevan and Christopher Hood on the targets of the English health system, documents activity shifting towards what is measured: waiting times reduced at the expense of what is not counted, selection of the most favourable cases, administrative reclassification.

Those effects presuppose no bad faith: they result from the rational allocation of limited effort towards what is assessed. Steven Kerr gave a formulation of it as early as 1975 that has become a classic, on the folly of rewarding one behaviour while hoping for another.

Multiplying indicators reduces the vulnerability of each but increases the collection burden and makes the overall picture harder to read.

Combining numerical indicators with qualitative judgement is the most common response. It reintroduces the subjectivity indicators were meant to remove, which is why the debate on this point does not close.