The 33 Ways Targets and Incentives Get Gamed

When a target or bonus scheme goes wrong, it always feels like bad luck. A rogue employee, a one-off, nobody could have seen it coming.

It is almost never bad luck. In July 2026, the UK’s communications regulator, Ofcom, fined Virgin Media – one of Britain’s biggest broadband and TV providers – £28m after finding call-centre agents hung up on customers trying to cancel, parked them on hold for no reason and bounced over a million callers between departments – and that the commission scheme itself “effectively encouraged” the behaviour. The agents were responding, entirely rationally, to how they were paid.

After twenty-five years of watching schemes misfire, I catalogued every distinct way they fail: 33 failure modes, in four families, each with a cause, a real example and a flag for whether it accelerates once it starts. The stories below are all real. Somewhere in here, you will recognise your own scheme – everyone does.

Family 1 – Target design failures

The goal itself is wrong before anyone starts chasing it. My favourite specimen: at the heart of the Wells Fargo scandal sat the “Going for Gr-eight” target – cross-sell eight products to every customer. When CEO Dick Kovacevich was asked how he arrived at eight, his answer was:

“It rhymes with GREAT!”

Dick Kovacevich, on the target design that preceded one of banking’s biggest scandals (TF-07: Arbitrary target selection)

Or take the UK government’s pandemic promise of “100,000 tests per day”. When the number looked out of reach, the counting method changed from tests used to test kits dispatched – and kit requests were reportedly fulfilled at triple the quantity asked for. Target approached; benefit, nil. That is TF-02, incomplete rule definition, performing exactly as the catalogue predicts.

All ten target design failures:

• All or nothing (TF-01) – one tier: hit it or you don’t. Most people have no realistic prospect; those close to the line are tempted into risk-taking. (Research on a US financial services firm found three target tiers significantly out-motivated one.)
• Incomplete rule definition (TF-02) – the rules don’t cover every situation, so the target can be won without the intended outcome. See the Covid testing story above.
• Design input bias (TF-03) – inputs excluded by design to flatter the result – like schools quietly preventing weaker students from sitting A-levels to protect their average grades.
• Excessively high bar (TF-04) – targets seen as insurmountable drive disengagement or worse. Tesco’s management saw their 2014 profit targets exactly that way; the company later admitted overstating profits by £250 million.
• Intentionally low bar (TF-05) – the bar set suspiciously low. Exam grade inflation is the classic: easier exams attract more schools, so the commercial incentive is always downwards.
• Weak or no link to intended positive outcome (TF-06) – hitting the number doesn’t deliver the thing you wanted. The ambulance story below ran on this for four decades.
• Arbitrary target selection (TF-07) – the number is plucked from the air – see “It rhymes with GREAT!” above.
• Individual and organisational success not aligned (TF-08) – pay programmers per line of code and you get lots of lines of code. Nobody said it had to be good code.
• Lack of pace-setting targets (TF-09) – no lap times. Elite runners never wait until the finish line to learn their pace; neither should your team.
• Inappropriate timescale (TF-10) – review frequency mismatched to the thing measured. A school inspected every four years can be a different school entirely by the time you read the rating.

Family 2 – Management failures

The goal may be sound; the way it is run is not. At Wells Fargo, one regional leader had district managers “run the gauntlet” – dressing in themed costumes and running down a line to a whiteboard to announce their sales numbers. Over the same period, sales-practice misconduct tripled. And in the US Air Force nuclear missile programme, nine senior officers were fired after systemic cheating in proficiency exams; investigators noted that leadership’s focus on perfection had led commanders to micromanage their people into it.

• Weak enforcement (MF-01) – rules exist but nobody enforces them – until it’s suddenly nine fired officers and a scandal.
• Negative leadership behaviour role-modelling (MF-02) – leaders visibly tolerating (or performing) the very behaviour the rules forbid. Whistle-blower reports at Wells Fargo were ignored for years.
• Intense management pressure (MF-03) – the number must be hit, whatever it takes – costumes optional, misconduct guaranteed.

Family 3 – Incentive design failures

The reward design does the damage. My own favourite came from a factory I worked with in the 1990s, where the bonus was set so low that someone painted the words “bonus shifter” on the side of a wheelbarrow and parked it in the middle of the shop floor. Message received – by everyone except the scheme’s designers.

The same family covers the opposite mistake: a study of Swiss volunteers found that offering a small financial reward decreased the hours they gave, compared with volunteers offered nothing at all. Pay people for what they loved doing, and you can kill the love.

• Low perceived incentive value (IF-01) – the reward isn’t worth the effort – see the wheelbarrow.
• Extreme rewards or punishments (IF-02) – when the stakes are keep-your-job or lose it, people sell to customers they know can’t pay.
• Non-differentiated incentive types (IF-03) – second prize as a worse version of first prize demotivates everyone who isn’t coming first. A golf weekend beats a three-star hotel stay it doesn’t have to out-rank.
• Winner takes all (IF-04) – one winner, nothing for anyone else – so the rest of the field stops trying.
• Capped rewards (IF-05) – the moment the reward stops, so does the effort. Cap the commission and you cap the motivation.
• Low-attainability driven disengagement (IF-06) – targets perceived as unattainable (whether they are or not) simply switch people off.
• No over-achievement rewards (IF-07) – exceptional performance earns nothing extra – the mirror image of the cap.
• Material rewards for intrinsically motivated people (IF-08) – pay people for what they loved doing and you can kill the love. Swiss volunteers offered a small fee worked fewer hours than those offered nothing.

Family 4 – Dysfunctional behavioural symptoms

What people actually do when the first three families leave the door open – this is where Virgin Media’s agents were operating. My most complete specimen is the UK ambulance service’s old “A8” target: 75% of life-threatening calls attended within 8 minutes. There was no medical basis for the 8 minutes (TF-06). To hit it, control rooms would sometimes dispatch up to four ambulances to a single borderline call (DB-09, resource fire-hosing). And after 43 years of it, the workforce’s faith in targets had corroded entirely (DB-10). One target, three failure modes, four decades.

• Invisible bar-lowering (DB-01) – standards quietly eased so the target gets “hit” – like factories nudging down the machine-speed target so the efficiency number holds up.
• Breaking the law (DB-02) – Volkswagen’s emissions cheating: one executive imprisoned, an estimated $33.3 billion in costs, sales down 24%.
• Rule bending or breaking (DB-03) – staff at a UK emergency call centre used quiet periods to place fake emergency calls to themselves – it did wonders for the Grade of Service figure.
• Using loopholes (DB-04) – technically compliant, spiritually bankrupt: orders placed to hit target, returned the week after.
• User input misclassification (DB-05) – one UK police force was found to have failed to record 80,000 reported crimes a year – clear-up rates looked great.
• Unexpected adverse outcomes (DB-06) – UK school attendance targets excluded closure days – so when snow threatened the figure, the rational move was to close the whole school.
• Output misclassification (DB-07) – insurance claim handlers bonused on throughput approved invalid claims, because rejecting them required paperwork.
• Corrupted reporting (DB-08) – a Sicilian health councillor under house arrest for allegedly altering Covid death figures to dodge red-zone restrictions.
• Resource fire-hosing (DB-09) – see the four ambulances above.
• Apathy and cynicism (DB-10) – when people see through the system, they stop believing in any of it.
• Loss of intrinsic motivation (DB-11) – the inner drive – curiosity, satisfaction, pride – quietly dies.
• Loss of extrinsic motivation (DB-12) – pay a child $2 to tidy their room and watch the payment become expected rather than motivating.

Want targets that survive contact with real humans?

Spotting your scheme in this catalogue is the easy part. The discipline is designing targets and incentives that never make the list – and stress-testing them against all 33 modes before anyone gets paid a penny against them. That is exactly what my GAMED course teaches: diagnose an existing scheme, design the fix properly, then white-hat and black-hat test it before launch.

Newsletter readers: click the button and something extra will land in your inbox within the hour.

Prefer to read first? The full method, with a worked example for every one of the 33 failure modes, is in my book GAMED.