Morale has dropped and I can't tell why.

What to do when morale has dropped and you can't tell why

New York, two thousand and twenty-one onwards. Thasunda Brown Duckett has just become chief executive of TIAA — the one-point-four trillion dollar retirement and asset-management firm serving the academic, medical, cultural, and research sectors. She arrives from JPMorgan Chase, where she had run the consumer banking division of fifty thousand employees. The diagnostic problem at TIAA is not new and not unusual. The firm has seventeen thousand employees, well-established employee surveys, all-hands feedback channels, and structured manager-to-employee one-on-ones. None of that apparatus is reliably telling the chief executive what is actually happening to morale across the firm.

The structural intervention Brown Duckett builds is a recurring small-group session. She names it Coffee and T. T for Thasunda. T for me.

Each session caps at ten people. The intentionality is the qualifier. Each group has something in common — a function, a level, a tenure cohort, a geography, sometimes simply people who joined TIAA in the last six months. The cap and the commonality are both load-bearing. Ten is small enough that quiet voices speak. The commonality means people walk in already knowing they share a referent, which lowers the cost of saying what's actually happening.

The structural question Brown Duckett asks, every session, is the one she has named publicly across multiple primary interviews. What would be the one thing that you'd change if you were in my shoes? Finally I get to be the CEO and I can get something done? Joseph Fuller at Harvard Business School, commenting on the practice in a CNBC piece in October two thousand and twenty-four, characterised it as the kind of open-ended question in a deliberately safe setting that produces information no formal apparatus can match.

The question routes around the fine answer that a manager-to-employee one-on-one almost always produces. It gives the employee permission to imagine themselves in the seat the question is being asked from. The threshold for saying something useful drops.

The sessions are recurring on Brown Duckett's calendar. The cadence is the discipline. A one-off small-group session produces a courtesy answer. A recurring practice produces the signal a chief executive can act on, because employees learn the loop closes — the next session reflects what the previous session surfaced.

When morale has dropped and you can't tell why, the first move is sometimes not to add something. The first move is to look at what's already running and ask whether the formal apparatus is failing to carry the signal — and whether a smaller, parallel, recurring channel needs to exist instead.

So let's go to the office and work through it.

Work out whether the formal channels are failing to carry the signal

"Morale has dropped and I can't tell why."

The feeling is uneasy.

You can sense the mood in the room shift downward across the last few weeks, and the standard prompts — one-on-ones, team meetings, the manager's open door — are surfacing nothing useful. The team is being polite. You are being told everything is fine.

Two choices. They look identical when the team starts going quiet. They need different first moves.

When the team has withdrawn into silence

Choice one: the team has gone isolated. People have quietly stopped talking to each other outside of the formal meetings. The one-on-ones aren't surfacing anything because people are too cautious to tell their boss the truth. Whatever's draining the room is sitting between people, not inside any one of them.

If that's the read, abolish your one-on-ones for a month and institute a peer buddy system. Pair people off and ask them to meet each other, not you, weekly. Chris Argyris, the Harvard organisational theorist who spent forty years on what he called undiscussables — the things teams can't say out loud — found that people will tell their peers what they will not tell their managers. Once peer conversations are running, the real concerns surface laterally and reach you a different way.

When a recent knock is still raw

Choice two: something specific has happened. A bad week, a rough launch, a client blowing up at the team. The room is sitting with a recent injury, and the average is dragging.

If that's the read, cancel the last hour of work today and send everyone home early with something small and unexpected. A bottle of wine, a handwritten note, something that costs nothing and signals everything. Daniel Kahneman's peak-end rule — what people remember of an experience is shaped by its peak moment and its ending, far more than by its average — says a positive surprising end will reframe the preceding week in memory more effectively than any amount of analysis or apology.

Isolated, or injured. Same quiet team. Two different first moves.

How to hear the team before you act

Three tools. The discipline is to set up the listening infrastructure before you try to fix anything.

Surface what the team can't say in a one-on-one

The first is

Team Health Check — sometimes called the Squad Health Check, after the Spotify model that codified it.

Henrik Kniberg and Anders Ivarsson published the Squad Health Check Model in a twenty-fourteen post on Spotify's engineering blog, codifying a practice the company had been running internally for several years across its squad-and-tribe organisation.

The reason the Team Health Check exists is that managers asking how's the team doing? in a one-on-one almost always get the same answer. Fine. The question puts the person asked in the position of either reassuring you or starting an argument. Most people choose reassurance. The tool replaces the open question with a depersonalised, anonymous traffic-light read across a small set of named dimensions, run on a routine cadence. The format does the work the conversational question can't.

The unique insight is that the aggregate is the diagnostic, not any individual answer. One person rating clarity red is information about that person. Eight out of ten rating clarity red is information about the system. The Health Check pulls the system's reading out of the noise of individual variation.

What you get is a recurring depersonalised diagnostic the team can complete in ten minutes, and a board the team can look at together where the colours show up before any individual is on the spot.

So. How to run it.

Setup. Pick five or six dimensions worth tracking. Support, clarity, pace, mission, autonomy is the canonical set, but the dimensions worth using are the ones that matter to your team. Anonymous form, three colours per dimension: green, amber, red.

Fill. Everyone completes the form individually, in private, before the meeting. Anonymity matters. The first time a team runs it with names attached, the colours skew green and the diagnostic dies before it starts.

Read. The team looks at the aggregate together. Not the individual responses — the colour mix per dimension. Clarity is mostly red. Mission is mostly green. Pace is split.

Act. Each red dimension gets a discussion and a commitment. The reds are the diagnostic; the commitments are what closes the loop.

Run it again four to six weeks later. Same form. Watch the reds. If the same dimension stays red across three rounds running with no movement, the commitments aren't landing — and that itself is the next conversation.

Make telling you worth the effort

The second is

Feedback Loops.

The lineage on this one is shared rather than singular. Norbert Wiener formalised the cybernetic feedback loop in the nineteen-forties as the fundamental unit of self-correcting systems — sensor, comparator, actuator. W. Edwards Deming carried it into management practice in the nineteen-fifties as Plan-Do-Check-Act. John Boyd formalised the OODA loop — observe, orient, decide, act — in the nineteen-seventies for military decision-making, generalised into management since. Three different fields, the same shape: a system without feedback that closes is a system that drifts.

The reason Feedback Loops matter when morale is falling is that listening infrastructure is only as useful as the response that follows. If people flag a concern through your Health Check, your buddy system, or any channel, and nothing visibly returns to action, the next round will flag nothing. The diagnostic dies because the team learns it's wired to nowhere.

The unique insight is that the loop closes at the return-to-action moment, not at the response moment. Telling people you've heard them isn't the loop closing — it's the start. The loop closes when the system actually adjusts what it does next based on what it heard. A team that flags a concern and watches the next quarter's plan change because of it keeps flagging. A team that flags a concern and watches you nod, then ship the same quarter's plan unchanged, stops flagging.

What you get when feedback loops close — when action returns to action, modified — is a self-correcting system. The diagnostic stays alive. The team trusts the channel because the channel produces visible system change.

So. How to run it.

Action. Apply the smallest possible move. Not a programme, not a strategy — one specific change small enough to ship quickly and read clearly.

Observe. Watch — collect signal, not opinion. What did the team actually do differently? What changed in the rhythms? Not what do they think about it — what shifted.

Reflect. Interpret — what does the signal mean? Was the change the one you expected, or did something else happen? The reflect step is the work most teams skip; without it, the next action is just guessing again.

Adjust. Change the next action — close the loop. Either persist with the same move, modify it, or stop. The loop is closed when the system has visibly adjusted, not when feedback has been received.

Feedback becomes input only when the system returns to action — the closure is the point.

Shield agreed work from churning priorities

The third is

OKRs — Objectives and Key Results.

Andy Grove worked out OKRs at Intel through the nineteen-seventies and wrote them down in High Output Management in nineteen eighty-three. John Doerr, who had been a young engineer under Grove at Intel, brought the practice into Google in nineteen ninety-nine as one of the company's earliest investors, and the framework spread from there into the mainstream of contemporary management.

The reason OKRs matter to a morale conversation isn't the goal-setting work they're usually associated with. It is the shielding work they do underneath. When morale drops and the cause turns out to be churning executive priorities — the team has been asked to switch what they're working on three times in a quarter — the team doesn't have a stable thing to point at when asked why they're not on this week's latest fire. OKRs give the team something to point at.

The unique insight is the structure of the pair. The Objective is the qualitative direction; it sets the where. The Key Results are the measurable outcomes; they set the how we'll know. A team that has agreed both — written down, visible, locked for the quarter — can absorb a request to switch direction by asking which of our key results does this replace? The question reframes the interruption. It is no longer will you do this thing? It is what are we agreeing to stop doing?

What you get is a team protected from churn without having to refuse upward. The OKRs do the refusing.

So. How to set them.

Frame. One sentence, qualitative, ambitious. Become the team customers actively recommend. Not three objectives. One.

Define. Three to five measurable outcomes that, taken together, would mean the Objective has been met. Numbers, not adjectives. Net Promoter Score above forty by quarter end. Renewal rate above ninety per cent.

Run. Weekly check-in on movement against the Key Results — fifteen minutes, no more. Quarterly setting and quarterly scoring. The check-ins are where the OKRs stay alive; without them the document goes cold.

Score. At quarter end, score each Key Result honestly. Grove's heuristic was that scoring around zero-point-seven across the set was the right ambition; consistently hitting one-point-zero means the targets weren't ambitious enough.

When churning priorities start arriving mid-quarter, point at the OKRs. We agreed these. Which one are we replacing?

A precedent: reading morale as energy that has migrated

That's the toolkit. One more story before we close.

The Brown Duckett story we opened with read the morale problem as a signal-routing failure — the formal apparatus wasn't carrying what the team was actually saying — and built a different channel that would. The story we close with reads the morale problem differently again: as a question of where the team's energy has migrated to, and whether the company has noticed.

Early two thousand and eleven. Michael Seibel and Justin Kan are running Justin.tv, the general-purpose live-video platform they co-founded in two thousand and seven with Emmett Shear and Kyle Vogt. The company is profitable. It is also, by its own founders' assessment, plateauing.

The roadmap conversation has been running for months. Seibel has been championing Socialcam, a mobile video-sharing app the company has spun up to chase the post-iPhone consumer-social wave. The data says mobile social is the larger addressable market. The data says Socialcam is the rational bet.

Underneath the roadmap conversation, something else has been happening. The StarCraft II beta released in February twenty-ten. Through twenty-ten and into twenty-eleven, the engineers around Emmett Shear have been playing it, watching it streamed, building features for the gaming category of Justin.tv that nobody asked them to build. The gaming vertical is around three per cent of the platform's traffic. It is also, as far as the founders can see, around a hundred per cent of where the team's spare-time energy is going.

Shear has been lobbying internally to pivot the whole company toward gaming. Seibel has been resisting, because the data still says mobile social. The impasse holds for months.

What breaks it is a diagnostic move. Seibel stops asking which is the larger market? and starts asking where has the team's energy actually migrated to? The answer is not on the roadmap. The answer is in the spare-time behaviour. Engagement on the main product hasn't failed; it has migrated. The team has, in effect, already pivoted itself.

In March twenty-eleven, the founders split the company. Socialcam spins out under Seibel. The gaming vertical relaunches as Twitch.tv on the sixth of June twenty-eleven under Shear. Three years later, on the twenty-fifth of August twenty-fourteen, Amazon acquires Twitch for nine hundred and seventy million dollars.

So.

The deeper point: listening infrastructure first, fixes second

Brown Duckett read the morale problem as a signal-routing failure and built a different channel — small, recurring, named, with a question employees could actually answer. Seibel read the morale problem as an energy-migration that the roadmap hadn't caught up to and followed the migration into the strategy. Two different cuts at the same question — what is morale telling you that the formal apparatus isn't?

When morale has dropped and you can't tell why, the temptation is to run a culture intervention on the thing that is losing its energy. Sometimes that is the right move. Often it isn't. The alternative is to ask what the formal apparatus is failing to surface, and where the team's actual energy has already gone instead.

The Health Check tells you what the team can't say out loud. The Feedback Loop is what makes telling you worth their effort. The OKRs are the shield that protects what they've agreed to do from what they haven't.

Listening infrastructure first. Fixes second.

So. Your Next Move from this playbook.

Which dimension of your team's health would everyone privately rate red, but no one wants to say so out loud — and what's the cost of another quarter pretending it's amber?