Your company just ran its annual engagement survey. Participation was 72%. The results show that 68% of employees are “engaged” or “highly engaged.” Leadership presents this at the all-hands as a win. HR creates an action plan. Nothing changes. Next year, you do it again.

Sound familiar? It should. This cycle plays out at roughly 85% of large organisations every year, and it’s almost entirely performative.

The employee engagement survey industry is worth over $1.5 billion globally, dominated by players like Gallup, Culture Amp, Qualtrics, and Peakon. These companies have built sophisticated platforms that measure, benchmark, and report on engagement with impressive precision. The problem isn’t the measurement technology. It’s that the thing being measured — “engagement” — is a construct so vague that measuring it precisely gives you precisely nothing useful.

The Construct Problem

What does “engaged” mean? Gallup defines it as being “involved in, enthusiastic about, and committed to” one’s work. That’s three different things bundled into one metric. An employee can be deeply involved in their work (because the stakes are high and failure has consequences) without being remotely enthusiastic about it. Another might be enthusiastic but uncommitted — loving the day-to-day but planning to leave in six months.

Collapsing these distinct experiences into a single engagement score creates a number that feels meaningful but tells you almost nothing actionable. It’s like measuring “health” as a single score combining blood pressure, cholesterol, bone density, and mental health. The aggregate number obscures everything useful.

The Response Bias Problem

Engagement surveys suffer from severe response bias, and most organisations don’t account for it.

Social desirability bias. Despite assurances of anonymity, employees in small teams know their responses can be de-anonymised through demographic crossreferencing. If you’re one of two women in a ten-person engineering team and the results are broken down by gender, your anonymity is gone. People self-censor accordingly, especially on questions about management quality and workplace culture.

Survivorship bias. The people taking the survey are the people who haven’t quit yet. By definition, the most disengaged employees — the ones whose feedback would be most valuable — have already left. Surveying remaining employees about engagement is like surveying restaurant diners about food quality while ignoring everyone who walked out mid-meal.

Participation bias. That 72% response rate? The 28% who didn’t respond are disproportionately likely to be disengaged. They didn’t fill out the survey because they’ve checked out — mentally or literally. By excluding their non-participation from the analysis, you’re systematically overestimating engagement.

Anchoring effects. The way questions are framed influences responses. “I feel valued by my manager” prompts different responses than “My contributions are recognised by my direct supervisor,” even though they’re measuring the same concept. Survey vendors know this and construct questions to produce distributable results — scores that cluster in the upper-middle range, making everyone feel okay without triggering alarm.

The Action Gap

Even if the survey data were perfectly reliable (it isn’t), the track record on acting on results is abysmal.

Research from Quantum Workplace found that only 22% of organisations show measurable improvement on the specific issues identified by engagement surveys. The remaining 78% either don’t create action plans, create plans they don’t execute, or execute plans that don’t address the root causes.

The typical action plan reads like a corporate Mad Libs exercise: “We heard that [communication/career development/work-life balance] is important to you. We’re committed to [improving transparency/creating new programmes/exploring options] in this area.” Six months later, nothing tangible has happened, and employees are more cynical than before the survey was conducted.

This creates what psychologists call the “survey-action gap” — and it’s actively harmful. Asking people for feedback and then ignoring it is worse than not asking at all. It signals that the organisation values the appearance of listening over actual responsiveness.

What Actually Works Instead

If you genuinely want to understand what’s happening in your organisation, there are better approaches.

Stay interviews. Sit down with employees individually and ask: “What keeps you here? What might cause you to leave? What would you change if you could?” These conversations are uncomfortable, time-consuming, and produce messy, qualitative data. They’re also infinitely more useful than a Likert-scale survey. You learn things in a 20-minute conversation that no survey question will surface.

Exit interview analysis. People are more honest when they’ve already decided to leave. Systematic analysis of exit interview themes — not the polished HR summaries, but the actual verbatim responses — reveals patterns that engagement surveys consistently miss.

Pulse checks with follow-through. Short, frequent surveys (3-5 questions) with visible, immediate action create a feedback loop that builds trust. The key word is “visible.” If the pulse check reveals that meeting overload is a problem, cancel some meetings that week and tell people why. Responsiveness is the point.

Operational metrics. Absenteeism, voluntary turnover, internal transfer rates, sick leave patterns, overtime trends — these behavioural indicators tell you more about engagement than any self-report survey. People vote with their feet and their time, and those votes are harder to fake than survey responses.

The engagement survey industry won’t go away because it serves an important corporate function: it lets leadership say “we measured engagement” in board presentations. But if you actually want to know how your people are doing, put down the survey platform and start having conversations. The data will be messier. It’ll also be real.