Equity didn't keep your people. The sentence under the number did.

You wrote the letter yourself, or you had HR draft it and you signed off. Profit-sharing, phantom equity, a bonus pool tied to EBITDA — whatever the mechanism, you put a real number in front of people and you wrote something proud and plainspoken above it: this is your success too, we're a team, let's chase the wins together. Six months later, the same people left at roughly the same rate as before. You assumed the number wasn't big enough. A new working paper out of Harvard Business School says the number was probably fine. It's what you wrote underneath it.

The paper is "Making Equity Incentives Actionable: Internal Communication and Broad-Based Employee Ownership," posted in May 2026 by Sam Karasik, Ethan Rouen, and Ashley Whillans, all of Harvard Business School. It's a working paper — not yet peer-reviewed — but it's a field experiment run on real payroll records, not a survey of intentions, and I read the full text. The setting: an anonymized private-equity-owned building-materials supplier in the southeastern U.S., 12 locations, 292 employees, all just granted stock appreciation rights the PE firm estimated would be worth roughly four months' salary at a future sale. The 12 locations were randomly split into two groups, each sent a shareholder letter explaining the identical program with the identical numbers. The only difference was what came after. One version defined the stake around abstract values — "chase the wins," "know what good looks like," "make room for everyone to be themselves." The other defined it around specific, checkable behaviors — follow safety protocols and report hazards immediately, show up and support one another, pursue training, treat the equipment like it's yours.

Over the following six months, tracked in personnel records, 25.7% of the norms-letter group had left the firm, against 38.9% of the values-letter group — a 13.2 percentage-point gap, same equity plan, same company, different four bullet points. That's real and sizable, and I want to be precise about how hard it's been tested: because only 12 locations served as the units of random assignment, the authors ran the most conservative check available, exact randomization inference, and it came back at a two-sided p of .121 — short of the conventional bar. A wild cluster bootstrap put it at .064. All 12 leave-one-location-out checks held the same direction. That's the authors' own reporting: with 12 sites, "statistically significant" and "large and directionally consistent" aren't quite the same sentence, and they say so.

Broad-based equity compensation gives rank-and-file employees a financial claim on firm value, but it does not clarify how employees should act on that claim.From the paper's abstract, Harvard Business School Working Paper 26-085

Three months into the experiment, Hurricane Helene hit the region and closed every one of the company's 12 facilities. That timing turned out to matter more than the letters alone. The retention gap barely showed up in the first three months — it opened specifically in the hurricane period and grew afterward. A follow-up culture survey shows the same delay: no difference between groups in July, a real and significant gap by November, after employees had lived with both the letter and the storm. The paper's own reading is that concrete behavioral guidance matters most exactly when routine breaks down and people are genuinely unsure what to do next — not in the calm month when the letter first landed.

Here's why that lands differently for you than for almost anyone else reading it. Your management team debates whether to do profit-sharing, how big the pool should be, which vesting schedule is fair. Almost nobody debates what to write under the number, because it feels settled — thank people, invoke the mission, sound like a leader. This study didn't test whether to give equity. It tested what happens when the same company hands out the identical stake and changes only the four sentences below it, and the boring version — the one naming actual Tuesday behaviors instead of team spirit — kept people through the worst month of the year.

The part the research doesn't say, but that I'll say plainly: the instinct to write the values version isn't laziness, it's that naming four specific behaviors feels small next to a real financial stake — beneath the size of the check you just described. That instinct runs backwards. The bigger the number you're handing someone, the more precisely they need to know what you expect them to do about it, and that need gets sharper, not softer, the moment something goes wrong.

The concrete move this week: pull the last thing you sent your team about pay, equity, or bonus — the announcement, the memo, the all-hands deck. Count the sentences that describe how you want people to feel about the company against the sentences that name a specific thing you want them doing. If the ratio runs the way it usually does, rewrite it. Not longer — just replace the values language with four or five concrete behaviors that, if everyone actually did them, would move the number you showed them.

This is the gap between vision and direction, and it's exactly what the Leadership Profile is built to surface — how you communicate expectations when the pressure is on, not when things are calm. It's a $47 assessment at /wiring-map.