Psychological Mechanisms of KPI Gaming

Why serious people game serious numbers

When you discover that a sales rep has massaged a CRM, your first reflex is moral: "they are dishonest." That is almost always wrong. KPI gaming is rarely the act of cheaters — it is the act of rational humans facing a poorly designed measurement system.

Understanding the psychological mechanisms that make gaming inevitable is the indispensable prerequisite to designing measurement systems your teams will have neither the desire, the need, nor the opportunity to corrupt.

The DDC model: Detection, Desirability, Cost

Social psychologist Dan Ariely and behavioural economist Nina Mazar synthesised in The (Honest) Truth About Dishonesty (2012) a predictive model of gaming. An actor will game a metric when the three DDC conditions are met:

The most malleable and most neglected variable is the third one. Teams usually manage Detection (audits) and Desirability (bonus alignment). They forget the moral cost — which is almost entirely a matter of internal narrative.

Rationalisation: the invisible engine

Your sales reps do not wake up thinking: "Today I will cheat." They tell themselves a story in which their behaviour remains consistent with their self-image. Here are the 6 canonical rationalisations of KPI gaming.

1. "Everyone does it"

Downward social norm. If the top performer massages numbers, the whole team allows itself to massage them. One person gaming → 30 days → the entire team games.

2. "The system is unfair anyway"

The actor frames themselves as defensive. Gaming becomes unfairness correction ("I earned this deal, it was miscounted"). Appears systematically when the measurement system is perceived as arbitrary.

3. "It's in the client's / company's interest"

Noble rationalisation. "I'm signing this deal at a loss to retain the customer." Often 30% true, 70% false.

4. "The KPI is broken, not me"

Displacing moral responsibility onto the system designer. And it is partially true.

5. "I'm just doing what I was told"

The instruction defence. Particularly common when management repeats: "Hit the number, I don't want to know how."

6. "The numbers don't reflect my real work"

The actor feels an emotional proxy gap (see module 1). They compensate by artificially inflating the proxy so it matches what they believe they actually deliver.

The moral tolerance curve: Mazar–Amir–Ariely model

The canonical experiment: participants solve math problems with self-reported scores (so cheating is possible). Results:

• No audit: 80% of people cheat a little. 0% cheat a lot.
• Bigger gain: the average cheating stays low — people cheat little, but persistently.
• Moral reminder (signing an honour pledge): cheating drops to zero.

Business implication: most KPI gaming in your teams is not extreme — it is systematic micro-gaming (5–10%). Punitive policies miss the target (they fight an imagined massive gaming). Moral-reminder and intent-transparency policies are massively more effective.

Asymmetric intentionality bias

Phenomenon documented by Knobe (2003): humans assign intention to a bad act much more readily than to a good act — for the same causal configuration.

Applied to KPI gaming:

• When someone else games a KPI: "they're cheating, it's intentional."
• When I game a KPI: "it's a pragmatic optimisation, I'm playing the system's game."

Until this bias is named in the team, no one recognises themselves in the gaming — even though statistically, everyone does it.

The motivation trap: crowding-out theory (Frey & Jegen, 2001)

Swiss economist Bruno Frey demonstrated that an extrinsic payment destroys intrinsic motivation on tasks with an ethical or creative component. This is the crowding-out effect.

Concretely: your sales rep did their job with pride (intrinsic motivation). You add a bonus on a narrow KPI (extrinsic motivation). Pride fades. The KPI becomes the only reason to do the job. And therefore the only reason to game it.

graph LR
    A[Intrinsic motivation<br/>pride, meaning, craft] --> B[Quality behaviour]
    C[Bonus on narrow KPI] -.crowd-out.-> A
    C --> D[KPI-only optimisation]
    D --> E[Gaming + off-KPI degradation]

This is one of the deep reasons why teams without individual bonuses (Patagonia, Basecamp, some scale-up teams) maintain behaviour more aligned with real value — they have not triggered crowding-out.

The 4 stages of team-level gaming

Gaming does not appear — it installs in stages. Here are the 4 stages observed in consulting practice and reproducible in most organisations.

Stage 1 — Silent exploration (weeks 1–4)

One or two actors test the metric. They are not gaming yet — they are mapping. "If I requalify this lead as X, what happens?"

Stage 2 — Local adoption (months 2–4)

The hack spreads by word of mouth in a subgroup. Management sees nothing: the aggregate numbers have improved (a sign that it works).

Stage 3 — Tacit normalisation (months 5–9)

The hack becomes the way to do things. New joiners learn it as if it were part of the job. The reported KPI no longer reflects operational reality — but no one knows, except the actors.

Stage 4 — Measurement collapse (month 10+)

The business reality (real revenue, real churn, real satisfaction) diverges from the KPIs. Management discovers in a strategic meeting that the numbers were lying. Crisis.

The later management detects this, the higher the cost of rebuilding. Any organisation above 200 people has at least one KPI at stage 3.

Behavioural markers to watch for

How do you know a KPI is starting to drift? Here is a 7-point checklist (from the audit/assurance literature + internal post-mortems like ENRON):

• The KPI improves faster than the underlying indicators (revenue, NPS, retention) it was supposed to predict.
• Inter-person dispersions tighten (everyone converges toward the "perfect" score).
• Suspicious spikes at cut-off dates (end of month, end of quarter).
• New joiners hit the KPI abnormally fast.
• You hear internal phrases like "you've got to know how to play the game."
• CRM tools / dashboards show new ad hoc categories that are undocumented.
• Top performers are the least able to explain their method.

Three boxes ticked = enhanced monitoring. Five boxes = you are at stage 3.

The counter-model: motivation by purpose

Convergent work by Daniel Pink (Drive, 2009), Edward Deci & Richard Ryan (Self-Determination Theory) and Adam Grant (Give and Take): humans who understand the real purpose (the "why") of the measured work game far less. Not because they are more moral — because the emotional proxy gap closes.

Direct implication: before any KPI rollout, explain what the KPI tries to capture. Not the formula. The underlying value. This kills 60% of gaming by default.

Conclusion

KPI gaming is not a moral defect — it is a predictable psychological mechanism. Detection, desirability, moral cost, rationalisation, crowding-out: you can predict and neutralise the drift if you understand the human engine. The next module shows how these mechanisms manifest specifically in a sales team under quota.