How Incentives Shape Behaviour — in Companies, Markets and Societies

Systems communicate through more than rules

Organisations publish mission statements, companies define values, states create rules and teams set objectives. People do not respond only to those formal statements, however. They also observe which actions are rewarded, tolerated or sanctioned in practice.

A company may declare long-term customer relationships to be a central value while paying salespeople almost entirely on short-term revenue. An organisation can emphasise quality while evaluating employees primarily on volume, or praise collaboration while tying promotion mainly to individual metrics. In each case, the formal message and the practical incentive are different.

People do not respond mechanically to every reward, but incentive structures still alter which choices appear sensible, risky or worthwhile.

Economic agency theory addresses a related problem: one person or organisation wants another actor to behave in its interest but cannot fully observe or control that actor's decisions.^[1]

That creates one of the hardest questions in organisational design: how can a system make individual decisions align as often as possible with the underlying collective objective?

1. People respond to what is rewarded

In simple tasks, an incentive can operate quite directly. If a worker produces more units and receives higher pay in return, the link between additional effort and additional reward is easy to understand. A sales commission works similarly when revenue is a good proxy for the performance the organisation actually wants.

The problem becomes harder when performance has several dimensions.

Suppose an employee is judged only by how many cases they close in a given period. The system creates a clear incentive to increase completed cases, but it may simultaneously reduce the willingness to take on difficult work, document decisions thoroughly or spend extra time on complicated customers.

That does not necessarily indicate a personnel problem. The organisation may simply have rewarded one part of the job so strongly that other valuable activities became comparatively unattractive.

Research on multitasking and performance pay describes precisely this effect: strong incentives for activities that are easy to measure can redirect effort away from tasks that are harder to quantify but still important.^[2]

A basic property of incentive systems follows from this. People do not automatically optimise an organisation's abstract overall goal; they respond to the signals through which success becomes concrete and consequential for them.

2. A metric changes once it becomes a target

Metrics are indispensable in complex organisations. Without measurement, it is difficult to know whether processes work, performance improves or resources are being used effectively.

Metrics are not neutral observers, however. Once people know that a particular number affects pay, career progression or evaluation, the metric itself begins to shape behaviour.

This problem is often associated with Goodhart's Law: once a measure becomes a target, it can lose some of its value as a reliable measure.^[3]

A call centre might, for example, use short call duration as a proxy for service quality. Once employees know that the number affects their evaluation, they have an incentive to end calls faster whether or not the customer's problem has actually been solved.

Something similar can happen when schools are judged almost entirely by standardised test results or when a sales organisation focuses only on the number of appointments created. The metric can improve while the underlying objective — learning quality or commercially meaningful opportunities — improves far less.

The problem is therefore not measurement itself, but the distance between the real objective and the measurable proxy chosen to represent it. The larger that distance becomes, and the stronger the personal consequences attached to the metric, the greater the risk that people optimise the proxy without improving the underlying goal to the same extent.

3. Good incentives for one task can damage the whole system

Real organisational performance rarely consists of a single task. A manager may be expected to grow revenue, control costs, reduce risk, develop employees, retain customers and build long-term opportunities at the same time.

Those activities are not equally measurable. Revenue is visible and can be quantified quickly; developing an employee or building durable customer trust is much harder to reduce to one number.

If the measurable part of the job receives most of the reward, that activity gains a structural advantage. A manager may understand perfectly well that coaching, risk management and long-term customer development matter, yet still rationally spend more time on the activity that directly influences bonus or promotion.

Research on multitasking incentives has examined this problem for decades. Strong performance incentives for selected measurable dimensions can crowd out activities whose value is more difficult to quantify.^[2]

The result can be paradoxical: the more rigorously an organisation optimises the wrong partial measure, the further it may move away from the outcome it actually wants.

4. Money can change the meaning of an action

Financial incentives invite an intuitive assumption: if unwanted behaviour becomes more expensive, it should occur less often; if desired behaviour is rewarded more strongly, it should occur more often.

That logic works in many settings. People do not interpret money only as a mathematical price, however; a financial incentive can also change how they understand the social situation around an action.

A well-known field experiment by Uri Gneezy and Aldo Rustichini illustrates the effect. At several day-care centres, some parents regularly arrived late to collect their children. The centres introduced a monetary fine for late pickup, which under a simple economic model should have made lateness less attractive.

Instead, late pickups increased. When the fine was later removed, behaviour did not simply return completely to its earlier level.^[4]

One plausible explanation is that the meaning of the situation changed. Before the fine, a social norm may have dominated: arriving late meant making other people wait and violating an informal obligation. A fee can partially recast that situation as a transaction in which lateness becomes something that can be purchased at a price.

The experiment does not establish that monetary penalties are generally counterproductive. The broader literature instead shows that financial incentives depend strongly on context and on how the people involved interpret them.^[5]

The deeper point is that an incentive can change not only the cost of an action, but also its social meaning.

5. Extrinsic incentives meet motivation that already exists

People do not work only for money. They may find a task interesting, want to help others, seek recognition, care about professional standards or derive part of their identity from doing good work.

Economists and psychologists therefore often distinguish between extrinsic incentives — such as pay or sanctions — and intrinsic or social motivation.

These forms of motivation do not operate independently. Financial rewards can reinforce existing motivation by signalling recognition or compensating extra effort, but under other conditions they can displace social or intrinsic motives.^[5]

Incentive design is therefore more complicated than the formula that more money simply produces more performance.

A bonus may be interpreted as evidence that a particular contribution is genuinely valued. In a different setting, the same bonus may communicate that the activity is apparently expected only when additional payment is offered.

The effect of an incentive therefore depends not only on its size, but on the expectations, norms and relationships into which it is introduced.

6. Markets are complex incentive systems

Markets can also be understood as large systems of distributed incentives.

When the price of a scarce resource rises, the decision environment changes for many actors at once. Consumers have a stronger reason to reduce use, producers may find additional capacity more profitable, companies search for substitutes, investors finance new production and engineers gain a stronger incentive to develop more efficient technologies.

No central authority has to coordinate all of those decisions. The price changes the relationship between cost and benefit for many participants simultaneously.

Prices do not operate in isolation, however. Taxes alter relative costs, subsidies influence investment, liability rules shift risk and entry restrictions can determine which competitors are allowed into a market at all. Legal and institutional frameworks therefore help shape which business models become economically attractive.

A market is consequently more than supply and demand. It is a network of prices, property rights, rules, risks and expectations to which firms and individuals respond.

7. Systems often produce exactly the behaviour they reward

Many problems that initially look like questions of organisational culture also have a structural dimension.

A company may complain that employees think too short-term while bonuses and promotions depend almost entirely on quarterly results. An organisation can demand more ownership while punishing mistakes far more heavily than passivity. A manager can ask for early risk reporting while placing anyone who brings bad news under exceptional pressure.

Employees in those situations receive contradictory signals. The formal message may be to surface problems early, while practical experience teaches that doing so creates personal disadvantages.

People do not have to be cynical or disloyal to respond. When a system produces certain consequences consistently over time, participants learn which choices are safer or more advantageous within that system.

That is why apparent culture problems are often worth examining through another lens: the behaviour may be less a reflection of individual attitude than a rational response to the incentives already in place.

8. Incentives change not only actions, but information

This creates a particularly difficult problem for organisations. They depend on information, yet the people who produce that information are themselves responding to incentives.

A salesperson evaluated heavily on pipeline size may have an incentive to classify uncertain opportunities more optimistically. A project manager whose career depends on staying on schedule may communicate risk more cautiously, and managers may delay bad news when negative information is punished more strongly than late reporting.

An incentive system can therefore influence not only what people do, but also what the organisation is able to learn about itself.

At that point, Goodhart's Law becomes more than a measurement problem. When personal consequences are attached to a metric, the process that generates the metric can itself change.^[3][6]

An organisation may believe it is observing performance objectively even after the production of its data has begun adapting to the evaluation system.

9. People learn systems — and begin to optimise them

Incentive systems do not act only at the moment they are introduced. People observe over time how the system actually works.

They learn which rules are strictly enforced and which barely matter, which metrics drive promotions, which exceptions are tolerated and which officially undesirable behaviours remain practically consequence-free.

That learning can be useful. Employees may become better at recognising which activities the organisation genuinely values and align their work accordingly.

The problem begins when the measurement and rule system becomes the main object of optimisation rather than the underlying objective. This is often described as gaming: actors adapt their behaviour so that they perform well under the rules without necessarily advancing the original purpose of the system.

Research on regulatory systems shows, for example, that firms can respond very precisely to the measurements to which regulatory consequences are attached.^[6]

That need not involve fraud. Quite often it is completely rule-compliant and rational from the perspective of the actor.

That is precisely what makes the problem difficult. A system can teach people to comply with its rules perfectly while drifting away from its purpose.

10. More control does not automatically solve the problem

A natural response to gaming, distorted metrics or delayed information is to increase control. Tighter monitoring appears, at first, to be the obvious corrective.

Control is itself part of the incentive structure, however. It creates cost, can reduce discretion and changes what people pay attention to. An employee who spends substantial effort proving formal compliance may begin optimising for the control process rather than the underlying result.

That does not make monitoring unnecessary. It means measurement, incentives and human judgement need to be combined so that no single element dominates the system.

A randomised public-sector study published in 2026 in Pakistan provides a useful example. Researchers compared different performance-pay designs for agricultural extension workers. Objective performance incentives mainly increased intensive effort through repeat visits to existing contacts. A system based on supervisors' subjective assessment, combined with light oversight of the supervisors, produced more geographic expansion and more training events while also reducing favouritism in bonus allocation.^[7]

The important lesson is not that one of those designs is universally superior. It is that different combinations of measurement, control and reward generated different patterns of behaviour.

11. The strongest incentives often sit outside the compensation plan

When companies discuss incentives, the conversation often centres on salary, commission and bonuses. Many of the most powerful signals emerge elsewhere.

Employees observe who gets promoted, who receives attractive assignments, whose mistakes are tolerated and whose judgement carries weight in important decisions. They see which behaviour creates status, who receives greater autonomy and which kinds of risk the organisation genuinely rewards or punishes.

Those observations accumulate over years. An organisation in which aggressive internal competition repeatedly leads to promotion develops differently from one in which collaboration is visibly connected to responsibility and career progression.

Culture and incentive structure therefore begin to merge. Organisational culture does not arise only from shared values or formal mission statements; it is also shaped by the behaviour an organisation repeatedly rewards, tolerates or sanctions.

For that reason, informal incentives can become more influential over time than any formal bonus scheme.

12. AI makes good incentive design more important, not less

Artificial intelligence and automation give this problem a new dimension. Machines can evaluate metrics continuously, agents can optimise processes autonomously and algorithms can make decisions at speeds that human organisations could previously not match.

That can make management more precise, but it does not remove the basic question of what the system should optimise.

If an AI system is told to optimise customer service, the objective still has to be defined. Shorter handling time, higher satisfaction, fewer escalations, lower cost and stronger retention may all be desirable, but they are not the same goal.

Maximising one can damage another. A system focused only on reducing call duration may process customers faster while resolving their problems less effectively; a system focused on suppressing escalations might simply push difficult cases further down the line.

This is fundamentally the same multitasking problem organisations have faced for decades.^[2]

Automation changes its speed and scale. A person may optimise a bad metric across hundreds of decisions; an automated system can apply the same flawed objective millions of times.

AI therefore does not eliminate the incentive problem. It makes precise objective design and effective controls more important.

13. Good incentive systems do not begin with the reward

When designing an incentive system, the intuitive starting point is often to ask how desired behaviour should be rewarded. A better approach is to begin one step earlier by defining the actual outcome the system is meant to produce.

The next question is which behaviours influence that outcome and which of them can be observed reliably. This creates an immediate limitation: not everything that matters to an organisation can be measured as easily as revenue, output or handling time.

Only then should designers ask what side effects may emerge when particular metrics are tied to personal consequences. Where can a measure be improved without improving the real outcome? Which valuable activities might be displaced because they are harder to quantify? Which intrinsic or social motivations already exist, and could a new incentive support or damage them?

The incentive system itself also needs a feedback loop. If people begin using the rules differently from what designers expected, or if metrics gradually lose informational value, the mechanism must be able to change.

Incentive design is therefore less about finding the perfect reward than about designing a system that can align behaviour, measurement and learning.

People respond to systems — but not mechanically

The statement that people respond to incentives can easily become an oversimplified theory of human behaviour. People are not optimisation machines that make identical choices whenever the financial conditions are identical.

Norms, relationships, identity, morality, habits and personal preferences matter as well. A new incentive therefore never enters a neutral environment; it is interpreted through social meanings that already exist.

A bonus can motivate, a fine can deter and a metric can provide orientation. Under other conditions, the same instruments can crowd out motivation, encourage gaming or redirect attention toward the wrong dimension of a task. Empirical research consequently finds both successful incentive schemes and cases of crowding out or unexpected behavioural change.^[5]

The relevant question is therefore not whether people respond to incentives, but how a particular incentive interacts with the social and institutional environment already in place.

Systems reveal their real priorities through consequences

Organisations can explain at length which values they consider important. Over time, employees pay closer attention to what actually happens when they behave in a particular way.

A company can demand long-term thinking, yet if careers depend almost entirely on short-term results, short-term behaviour remains attractive. An institution can demand transparency, but if bad news repeatedly creates personal disadvantages, people gain an incentive to make problems visible as late as possible.

The distance between declared values and lived incentive structure can therefore become substantial.

For many organisational or social problems, a simple analytical question is useful: what behaviour would be understandably attractive to a rational actor inside this system?

That question does not explain every human decision. It does force us to look beyond individual motives and examine the structure within which those motives are translated into action.

We often get not the behaviour we ask for — but the behaviour we make attractive

Incentives are among the less visible components of a system. Buildings, organisation charts, laws and process descriptions can be seen directly; incentives emerge from the interaction of rewards, costs, risks, expectations and informal consequences.

That is precisely why they can be so powerful. They influence countless individual decisions without a central authority having to instruct every action explicitly.

Well-designed incentive systems can bring individual interests and shared objectives closer together. Poorly designed systems can instead encourage people to optimise metrics rather than outcomes, withhold relevant information or comply formally with rules while increasingly bypassing their purpose.

The central lesson is therefore not that people always do what they are paid to do. Human behaviour is far more complicated than that.

What matters more is that people learn which choices a system consistently rewards, enables or punishes, and adjust at least part of their behaviour accordingly.

When the resulting behaviour repeatedly diverges from the desired outcome, an organisation should not automatically begin by trying to change the people. The underlying problem may lie in the design of the system itself.