Ninety-four studies. More than 23,000 workers. No overall relationship between electronic performance monitoring and better performance โ not on speed, not on output volume, not on helping behaviour, not on reducing counterproductive work behaviour (Ravid et al., Personnel Psychology, 2023).
The same meta-analysis found monitoring reliably moves four other things: stress up, job satisfaction down, fairness perceptions down, privacy-invasion feelings up.
So the intervention has a consistent effect. It just isn't the one on the invoice.
Employee monitoring is one of the few operational purchases where the accumulated evidence points in the opposite direction from the sales deck โ and where the cost of being wrong lands on the part of the organisation that is hardest to repair. If you are a Head of Operations renewing an ActivTrak, Hubstaff, Time Doctor, or WorkiQ contract this quarter, that is the sentence worth sitting with before you sign.
The Evidence Base Is Larger Than the Business Case
Most operational software gets bought on a vendor case study and a two-week trial. Monitoring is unusual in that it has been studied for three decades across call centres, warehouses, back offices, and remote knowledge work, and the results have been pooled.
That pooling is what makes the null result meaningful. A single study showing no effect is noise. Ninety-four studies aggregating to no overall effect on performance means the effect, if it exists, is small enough and inconsistent enough that no operator should plan around it (Ravid et al., Personnel Psychology, 2023).
Note what is not being claimed here. The finding is not that monitoring never changes behaviour. It is that across the accumulated literature, monitoring does not reliably produce more or better work. People adjust to being watched โ they just adjust in ways that show up in the metric rather than in the outcome.
That distinction is the whole problem. Activity data is easy to collect and easy to display. Output is neither. When you install a system that measures the first and reports it in a dashboard labelled "productivity," you have not measured productivity. You have measured what your monitoring tool can see, and then renamed it.
What Employee Monitoring Actually Buys You
Strip the marketing and monitoring software delivers three things reliably: activity telemetry, a management-visible dashboard, and a sense of control for whoever signed the contract.
Only the first is a fact. The second is an interface decision. The third is a feeling, and it is the one that closes the sale.
I want to be fair to the buyer here, because the motivation is usually reasonable. A Head of Operations running 200 people across hybrid arrangements has genuinely lost a signal they used to have. Walking the floor gave them a coarse but real read on where work was piling up. Monitoring software is sold as the replacement for that walk.
It is a poor replacement, for a specific structural reason: the floor walk was aggregate and ambient, and the dashboard is individualised and permanent. Those are different instruments with different effects on the people being observed, even when the stated intent is identical.
The Transparency Paradox: Privacy Made a Factory Better
The most instructive counter-example is not a survey. It is a field study inside a large mobile-phone factory in China, where researchers gave some production lines visual privacy โ literally, curtains around the work area โ and compared them to matched, fully observable lines.
The lines with privacy produced fewer defects (Bernstein, Administrative Science Quarterly, 2012).
The mechanism matters more than the result. Under full observation, operators ran the officially sanctioned process, because deviating from it in front of managers is a career risk. Behind the curtain, they ran the process they had actually improved โ local workarounds, informal fixes, small deviations that made the line work better and that nobody had written down.
Observation did not make people lazy. It made them compliant. And compliance suppressed exactly the informal problem-solving that was generating the quality gains.
Now map that onto knowledge work. The equivalent of the sanctioned process is the visible activity: hours in the tool, messages sent, tickets touched. The equivalent of the curtain is the unobserved half-hour where someone thinks about the problem instead of typing at it. Monitoring systems reward the first category by construction, because it is the only category they can see.
The Strongest Counter-Argument, and Where It Holds
There is a legitimate version of this technology, and dismissing it wholesale would be sloppy.
Aggregated, anonymised workload data answers questions an operator genuinely needs answered. Which function is absorbing the most rework? Where did cycle time degrade after the process change? Is the support team's queue depth a staffing problem or a routing problem? Those are diagnostic questions about the system, and telemetry can inform them.
The meta-analytic result does not say that data about work is useless. It says that monitoring individuals does not make those individuals perform better (Ravid et al., Personnel Psychology, 2023).
The line, then, is not surveillance versus no surveillance. It is aggregate diagnostics versus individual scrutiny โ and almost every product in this category is sold as the first and configured as the second. The demo shows a department heat map. The default deployment ships with per-person activity scores, idle-time alerts, and a manager view that ranks names.
Ask your vendor directly whether individual-level identification can be disabled at the data layer, not just hidden in the UI. The answer tells you which product you are actually buying.
Why the Renewal Gets Signed Anyway
The current wave is not theoretical. TD Bank began rolling out WorkiQ tracking to monitor employee browser use and internal chat activity in June 2026, and the reporting around it made clear how thin the legal guardrails are โ Canadian workers have limited statutory protection against workplace surveillance, and the same is true across much of the United States (Reuters, 2026; Canadian HR Reporter, 2026).
When a large regulated institution normalises a practice, mid-market procurement follows within two quarters. The reasoning is rarely examined: a bank with a compliance department did it, so it must be defensible.
That inference is weaker than it looks. A bank monitoring fraud investigators has a supervisory and regulatory obligation that a 220-person services firm does not. And legal permissibility has never been evidence of operational value โ the expansion of workplace surveillance has tracked the falling price of the tooling far more closely than it has tracked any published performance case.
The cost that does not appear in the renewal quote is the one the meta-analysis measured: stress, satisfaction, and fairness perceptions all move the wrong way (Ravid et al., Personnel Psychology, 2023). In a market where mid-market retention is already expensive, deploying a system with a null performance effect and a reliable trust cost is a poor trade before you count the licence fee at all.
There is also a second-order effect worth pricing. Monitoring changes what employees believe the organisation values, and they optimise accordingly. If the visible metric is activity, activity is what you will get more of โ sessions kept open, messages sent, tools left running. None of that is fraud. It is a rational response to a stated preference, and it degrades the very data the dashboard was installed to produce. Within two quarters, the metric measures compliance with the metric.
Three Moves Before the Next Renewal
Separate the diagnostic use case from the scrutiny use case โ in writing
Before renewal, write one page naming the specific operational questions the data will answer, at what level of aggregation, reviewed by whom, at what cadence. If a question can only be answered by identifying an individual, it belongs to a performance-management conversation, not a dashboard. Anything the page cannot justify gets switched off. Most deployments lose two-thirds of their enabled features in this exercise, and nobody notices.
Set the data-use policy before the tool goes in, not after
The predictable failure sequence is: deploy, discover a manager is pulling individual reports, then write the policy in response to a complaint. By then the trust cost is already booked and the policy reads as damage control. Publish retention period, access list, permitted uses, and prohibited uses before the first agent installs. Then hold to it, because the second violation is the one people remember.
Audit what the dashboard is displacing
Ask your managers a direct question: since the tool went in, has anyone spent less time in one-to-ones because they can see the activity feed? That substitution is the real cost. Activity data creates the sense of knowing how someone is doing, which reduces the felt need to ask. A manager who checks the dashboard instead of the person has traded a high-bandwidth signal for a low-bandwidth one and will not notice until someone resigns.
One Decision This Quarter
Open your monitoring contract and find the individual-level features. For each one, name the operational decision it has informed in the last six months โ a specific decision, with a date.
If you cannot produce that list, employee monitoring is not measuring productivity in your organisation. It is buying visibility and expensing it under a different name.
The factory lines that produced fewer defects were the ones nobody could watch. That result is fourteen years old, and the industry selling you the opposite conclusion has had the entire interval to produce evidence for it. Ninety-four studies in, it hasn't.
