What actually re-engages people when the perks don't

A few years ago a company I knew responded to a brutal engagement survey by buying a ping-pong table. It sat in the corner of an open-plan floor where everyone wore headphones to survive the open-plan floor; people walked past it to microwave their lunch. Six months later the next survey came back worse, and someone in leadership said, with genuine bafflement, that they had tried.

That is the whole problem in one prop. When engagement sags, the instinct is to add something visible and cheerful — a perk, an app, a wellness week — because it is fast, it photographs well, and it lets leadership feel they acted. The trouble is that the rigorous evidence on these interventions is bleak. The largest study of its kind, an Oxford analysis led by William Fleming covering more than 46,000 UK workers and around ninety different wellbeing offerings, found that mindfulness sessions, resilience training, stress-management courses and wellbeing apps produced no measurable improvement in the people who used them — and a couple of categories trended slightly negative. A separate randomised trial of roughly 33,000 US employees, run out of the University of Chicago and Harvard, found wellness programmes nudged what people said about their habits but moved nothing that mattered: not clinical health, not spending, not absenteeism, not performance.

So if the snack wall is theatre, what is the actual stagecraft? It has a more useful answer than the cynicism suggests. The fixes exist; they are just unglamorous, structural, and mostly free.

Start from why people check out

You cannot prescribe until you diagnose, and the diagnosis is the thing the perks quietly ignore. People do not disengage because they are lazy. They disengage for three reasons that show up again and again across the surveys: economic insecurity, a lost sense that the work means anything, and a trust gap with the people running the place. Deloitte’s recurring survey of Gen Z and millennial workers has found roughly half of each group saying they do not feel financially secure — and financial precarity is a reliable engine of the emotional fatigue that precedes checking out. Underneath that sits the slow drain of meaning: duties get done while the sense that any of it matters leaks away. And over the top sits the trust gap — the distance between what an organisation says and what its people actually experience.

Hold those three causes in mind, because they are the test every intervention has to pass. A perk that does not touch security, meaning, or trust is, almost by definition, going to fail. A ping-pong table addresses none of the three; recognition addresses meaning; a capable manager addresses all three at once. That is why the rankings below come out the way they do — not because anyone polled people’s favourite benefits, but because some moves reach the actual wound and most do not.

Rank them by leverage, not by enthusiasm

Here is the part that matters for anyone who leads a team and has finite energy to spend. The interventions that work are not equally cheap or equally powerful, and the most powerful happen to be among the cheapest. Plotted against each other, the picture is almost the opposite of the perks instinct: the things that move the needle most cost the least, and the expensive, photogenic options cluster near the bottom.

Treat that chart as a map of where to spend, not a precise measurement — the effects come from different studies with different definitions, so the bar lengths are directional. But the ordering is the honest part, and each lever earns its place for a specific reason.

The structural lever: equip the manager

If you change one thing, change this. Gallup’s workplace research, which spans well over a hundred countries, attributes something like 70 percent of the variance in a team’s engagement to its manager. The single largest determinant of whether your people are checked in or checked out is not the compensation band, not the office, not the mission statement — it is the person they report to and how that person leads.

Which makes the rest of Gallup’s findings alarming: managers are the group deteriorating fastest. Their own engagement fell from roughly 30 percent to 27 percent to about 22 percent across a single stretch, the steepest drop on record, with younger and female managers falling hardest. So the highest-leverage point in the system is also the one under the most strain — and a burned-out, unsupported manager cannot manufacture engagement for anyone else.

The fix is not a slogan; it is investment in a specific, trainable craft. For a data or engineering lead that craft is concrete: one-on-ones that are about the person and not just the sprint, feedback that is frequent and specific rather than saved up for a review cycle, removing blockers instead of adding pressure, and shielding the team from organisational chaos rather than transmitting it downward. There is a telling perception gap here — about half of managers believe they give weekly feedback, while only around one in five employees agree they receive it. The gap is rarely dishonesty: a hallway “nice work” does not land as recognition, and a manager who feels communicative can still leave a team feeling unseen.

The cheapest high-ROI move: recognition

Recognition is the lever with the best return on the least money, and almost everyone underrates it. The research that informs the ranking above puts the satisfaction lift from regular recognition at around 22 percent, and finds that roughly two-thirds of workers say they would work harder if they simply felt their effort was noticed. Setting clear expectations — telling people what good looks like — lifts engagement by something on the order of 30 percent. None of that requires a budget line.

The catch is that recognition has to be specific to count. “Great job, team” is wallpaper. “The way you restructured that ingestion job cut our nightly run by forty minutes and it has not paged anyone since” is recognition, because it proves you actually saw the work. For data and engineering teams this matters more than usual, because so much of the best work is invisible by design — the migration that did not break anything, the test that caught the bug before production, the on-call night nobody noticed. If the only work that gets named is the work that goes wrong, you are training your strongest people to feel that competence is met with silence — a fast route to the quiet, on-the-job fraying that hides inside good performance reviews.

Redesign the work, not just the mood

Further down the leverage ranking, but still well above any perk, sits the most underused lever of all: changing the work itself. The strongest positive evidence in this entire field comes from the largest four-day-week study to date, published in Nature Human Behaviour in 2025 — 141 organisations and nearly 2,900 employees across six countries, keeping full pay for about 80 percent of their hours. Burnout fell sharply, by around 0.44 points on a five-point scale; job satisfaction, mental health and physical health all improved; productivity held or rose; and roughly nine in ten companies kept the schedule after the trial. The gains were not magic. They came from less fatigue, better sleep, and recovered capacity to actually do the work.

I rank this lever a notch below recognition only because it costs more — it is real operational change, and not every team can adopt a literal four-day week. But the underlying principle is portable and cheap: reduce the friction and the fragmentation, not just the hours. Knowledge workers get interrupted roughly every couple of minutes and field a hundred-plus emails or messages a day; for engineers, much of the after-hours crunch traces to testing and deployment bottlenecks downstream of how fast code now gets written. Protecting blocks of focus time, killing a standing meeting, fixing the pipeline that forces the weekend deploy — these are workload redesigns a team lead can make without anyone’s permission, and they touch the fatigue that perks cannot reach.

Make it safe to say the quiet part, and tell the truth about AI

The last two levers are about trust, and they are nearly free. Psychological safety — people feeling able to say “I am losing the thread on why this matters” without it counting against them — is what converts silent deterioration into a problem you can still solve. When the channel for dissatisfaction is open, it surfaces as words in a one-on-one. When it is closed, it surfaces as a resignation nobody saw coming. The cost is mostly a manager’s willingness to hear something uncomfortable and not punish the messenger.

And then there is the specific anxiety of this moment: AI. The trust gap here is measurable and self-inflicted. Gallup has found that while about 44 percent of employees say AI is being used in their workplace, only around 22 percent say leadership has actually explained how. That silence is where the fear breeds — Mercer’s recent work shows the share of employees afraid of losing their job to AI jumping from roughly 28 percent to 40 percent, and the proportion who simply “feel good at work” sliding from about 66 percent to 44 percent. The intervention is not a reassuring slogan; it is candour. Telling a team plainly how AI will and will not be used on their work — what it automates, what it augments, where the role is heading — directly drains the uncertainty that erodes trust. People can adapt to a hard truth; they cannot adapt to a vacuum, so they fill it with the worst-case story.

The takeaway

The reason perks keep losing is not that they are bad; it is that they answer a question nobody asked. People do not disengage for lack of foosball. They disengage because they feel insecure, unseen, and unsure anyone at the top is being straight with them. Every lever that actually works is a direct answer to one of those — recognition for the unseen, transparency for the uncertain, a capable manager and a safe channel for all of it.

If you lead a small data or engineering team and want the highest return on the least spend, the order is unromantic and clear. Get good at the manager craft, because it moves more than everything else combined. Name real work specifically and often. Set expectations people can actually see. Make it safe to tell you the truth, then tell them the truth back — especially about AI. None of that needs a budget, an app, or a table in the corner. It needs attention, which is the one thing a perk can never fake.