The observer effect is the change in a system caused by the act of observing or measuring it. In physics, measuring a particle disturbs it. In human systems, being watched or measured changes behaviour. The effect is not always large, but it is pervasive: when you add a metric, a review, or a dashboard, you change what people do. The strategic question is whether that change is desirable — alignment with the goal — or distorting — gaming, hiding, or optimising for the measure instead of the outcome.
In understanding and analysing, the observer effect warns that data is not neutral. The way you measure, what you choose to track, and who knows they are being tracked can alter the phenomenon you are trying to understand. Experiments and KPIs are not passive windows on reality; they are interventions. Design measurement and feedback so that the induced behaviour is what you want, or at least so you can separate the effect of observation from the underlying state.
In leading and organising, the effect is central to management. What you measure gets attention; what gets reviewed gets done. The lever is real — use it to focus the organisation on the right things. The risk is that the act of measuring or reviewing shifts behaviour in unwanted ways: people optimise for the number, avoid the activity that is not measured, or perform for the observer rather than for the outcome. Leaders must choose what to observe and how, knowing that the choice will change behaviour.
Section 2
How to See It
You see the observer effect when behaviour or results shift after a new metric, review, or measurement is introduced — and the shift is clearly a response to being observed rather than to the underlying task. The diagnostic is the timing and the nature of the change: does it track the introduction of observation, and does it look like response to the observer (gaming, showmanship, avoidance) rather than improved underlying performance?
Business
You're seeing Observer Effect when a team's output improves the week before a quarterly review and drops after. The underlying capacity may be unchanged; the observed level is a response to the review cycle. Or when a new KPI is added and behaviour shifts to hit the KPI while other dimensions suffer — the act of measuring changed the system.
Technology
You're seeing Observer Effect when A/B test results change once teams know which variant is winning and start optimising for it, or when monitoring and alerting change how engineers deploy (e.g. avoiding deploys before weekends to avoid being on call). The measurement and visibility alter the process being measured.
Investing
You're seeing Observer Effect when a company's reported metrics improve after they become public or after a fundraise — not because the business improved but because management is optimising for what investors see. The act of being observed (reporting, due diligence) changes what is reported and sometimes what is done.
Markets
You're seeing Observer Effect when a policy or regulation is announced and behaviour shifts in anticipation — for example, pulling forward or delaying transactions to avoid or exploit the change. The announcement itself changes the system before the policy is in force. The observer (the regulator, the market) has changed the phenomenon.
Section 3
How to Use It
Decision filter
"When you introduce a metric, a review, or any form of observation, ask: how will this change behaviour? Will the change align people with the goal or distort behaviour (gaming, hiding, optimising for the measure)? Design observation so that the induced behaviour is desirable, or correct for the distortion, or accept that you are not measuring the undisturbed system."
As a founder
Choose what you measure and review with care. The observer effect means that whatever you track will get attention and will shift behaviour. Use that to focus the team on outcomes that matter — but be alert to gaming. If a metric can be gamed without improving the real outcome, the observer effect will push people to game it. Prefer outcome metrics that are hard to fake, or balance multiple metrics so that no single one can be optimised at the expense of the whole.
As an investor
When you ask for data or run due diligence, assume that the act of asking changes what you get. Companies may present what they think you want to see; they may also change behaviour because they know they are being observed. Triangulate with independent sources and with metrics that are harder to manipulate. Do not assume that "what we see" is identical to "what would have been there without us."
As a decision-maker
In performance and strategy reviews, separate the effect of the review from the underlying performance. Some improvement is real; some is the observer effect (preparation, timing, selective reporting). Use unannounced checks, different observers, or lagging indicators that are harder to game. Design the rhythm of observation so that it reinforces the behaviour you want rather than ritual or gaming.
Common misapplication: Assuming that because you measure something, you see the truth. Measurement and observation change the system. The number you get is a product of the undisturbed state and the effect of being observed. Do not treat it as a pure readout of reality.
Second misapplication: Avoiding all measurement to prevent distortion. The observer effect is a reason to design measurement well, not to abandon it. The alternative to good measurement is no measurement or bad measurement — and decisions without data have their own problems. Use the model to improve what and how you observe, not to stop observing.
Campbell stressed that what you measure and how you review it shapes culture and behaviour. He advised leaders to measure the right things and to run reviews in a way that reinforced honesty and improvement rather than theatre. His approach acknowledged the observer effect: the review is not neutral; it teaches people what matters and how to behave.
Netflix's culture of "context not control" and high transparency is a deliberate use of the observer effect: by making goals, performance, and feedback visible, the company tries to align behaviour with objectives. The design choice is to observe broadly and openly so that the induced behaviour is alignment rather than hiding or gaming in the dark.
Section 6
Visual Explanation
Observer Effect — The act of observing or measuring changes the system. What you see = underlying state + effect of observation. Design measurement so the induced behaviour aligns with the goal.
Section 7
Connected Models
The observer effect sits with models about measurement distortion, expectations, and incentives. The connections below either describe related distortions (Goodhart's Law, placebo, expectation effect), extend the idea (Heisenberg, Pygmalion), or help separate signal from noise.
Reinforces
Goodhart's Law
Goodhart's Law: when a measure becomes a target, it ceases to be a good measure. The observer effect is the mechanism: the act of making something a target (observing it, rewarding it) changes behaviour so that the measure no longer reflects the underlying thing. The reinforcement: both say that measurement and targeting alter the system. Design measures and incentives so the induced behaviour is what you want.
Reinforces
Pygmalion Effect
Pygmalion effect: expectations affect performance; people rise (or fall) to what is expected. The observer effect is broader (any observation changes behaviour); Pygmalion is the case where the observer's expectations are the mechanism. The reinforcement: both say that the observer is not passive — what you expect and what you observe shape the outcome.
Tension
[Placebo Effect](/mental-models/placebo-effect)
Placebo effect: the intervention of "being treated" can change outcomes even when the treatment is inert. The observer effect: the intervention of "being observed" can change behaviour. The tension: in both cases the act of intervention (treatment or observation) is part of the causal chain. Separating "real" effect from placebo or observer effect requires design — blinding, control groups, or multiple measures.
Tension
Expectation Effect
Section 8
One Key Quote
"What we observe is not nature itself, but nature exposed to our method of questioning."
— Werner Heisenberg
The quote applies beyond physics. In human systems, what we observe is not the undisturbed system but the system exposed to our method of observation — our metrics, our reviews, our questions. The observer effect is that exposure. Design the method of questioning so that what is exposed is what you need to see, and so that the exposure does not distort the system in ways that defeat your purpose.
Section 9
Analyst's Take
Faster Than Normal — Editorial View
Assume observation changes behaviour. When you add a KPI, a review, or a report, assume it will shift what people do. The shift can be good (focus on the right outcome) or bad (gaming, hiding, teaching to the test). Design for the good and guard against the bad. Prefer metrics that are hard to game; balance multiple metrics so no single one can be optimised at the expense of the whole.
Use the effect deliberately. What you measure gets attention. So measure what you want to prioritise. The observer effect is a lever: choose what to observe and how, and you shape behaviour. The risk is choosing poorly — measuring the wrong thing or measuring in a way that induces gaming. The discipline is to align the measure with the real goal.
Triangulate when stakes are high. When you are the observer (due diligence, performance review), assume that your presence and your questions change what you see. Triangulate with independent sources, lagging indicators, or unannounced checks. Do not assume "what they showed me" is "what would have been there if I had not been looking."
Do not abandon measurement. The observer effect is a reason to design measurement well, not to avoid it. The alternative is decisions without data — which have their own errors. Use the model to improve what and how you observe; do not use it to justify no measurement.
Section 10
Test Yourself
Is this mental model at work here?
Scenario 1
A company adds a KPI for 'customer contacts per week.' Support staff start making short, low-value calls to hit the number. Customer satisfaction drops.
Scenario 2
A team's output is higher in the week before a board meeting and lower the week after. The CEO says the team is inconsistent.
Scenario 3
A researcher runs an experiment and finds no effect. A colleague says 'maybe the subjects changed their behaviour because they knew they were being studied.'
Section 11
Summary & Further Reading
Summary: The observer effect is the change in a system caused by observing or measuring it. In understanding and analysing, treat data as non-neutral — measurement is an intervention. In leading and organising, use the effect to focus behaviour on the right outcomes, and guard against gaming and distortion. Design what you measure and how you review so the induced behaviour aligns with the goal. Pair with Goodhart's Law, placebo effect, Pygmalion effect, expectation effect, Heisenberg uncertainty, and signal vs noise.
The original studies at Western Electric where worker productivity changed when researchers observed them. The effect is often cited as the birth of the idea that observation changes behaviour in human systems.
Goldratt on how metrics and targets distort behaviour in organisations. Practical treatment of observer-effect-style dynamics in operations.
Expectation effect: what we expect shapes what we see and how we behave. The observer effect: observation changes the system. The tension: the observer brings expectations; those expectations can shape both what they notice and how the observed respond. The combined effect is "observer plus expectation" — design for both.
Heisenberg's principle: in quantum mechanics, measuring one property of a particle disturbs another. The observer effect in physics is the same idea: measurement disturbs. The lead: the principle formalises that in some domains you cannot observe without changing. In human systems the analogue is that you cannot measure without affecting behaviour — so choose what to measure with care.
Leads-to
Signal vs Noise
Signal vs noise is separating true signal from random variation. The observer effect adds a layer: what you measure may be "signal plus observer-induced change." To get cleaner signal, design observation to minimise distortion, or use measures that are harder to game. The lead: observer effect is a source of systematic noise (or bias); account for it when interpreting data.
