The Hidden Coin: Why the Size of Recognition Doesn't Matter

The psychologists Amit Kumar and Nicholas Epley have spent years documenting a quiet miscalibration in how people give. When someone performs a small act of kindness, a thank-you note, a compliment, a cup of hot chocolate handed to a stranger, they consistently underestimate how much it lands. The giver fixates on the modesty of the gesture. The recipient feels the warmth. The positive impact is real, and it is hidden from the person who created it.

Workplace recognition is the same gesture, made repeatedly, at scale. On Happily, an employee can give a peer "coins" worth 3 or 5. We analyzed 37,116 of those coin events across 68 companies to ask a sharper question than "does recognition help." Earlier Happily research already settled that one: recognition is the single strongest workplace predictor of well-being. The more useful question is whether its value is hidden in ways that change how people should use it.

Two findings survive rigorous testing, and both are forms of underestimation. The size of the coin is irrelevant: a 3-coin lifts well-being as much as a 5-coin. And giving pays the giver about as much as receiving pays the receiver. Meanwhile, four in ten employees go a full year without a single coin. The cheapest, most evenly helpful gesture is the one most often skipped.

Finding 1: the size of the coin does not matter

If recognition behaves the way givers treat it, the 5-coin (reserved for the bigger contribution) should land harder than the 3-coin people hand out casually. It does not. The cleanest test compares employees who received only 3-coins against those who received only 5-coins, so the denomination is the only thing that differs. The two groups are statistically indistinguishable on both well-being measures.

The collinearity between how many 3-coins and 5-coins a person receives is severe (r = 0.92), so a model with both counts is uninterpretable. Instead we hold total recognition volume constant and ask whether a higher share of premium 5-coins adds anything. It does not. What predicts well-being is the volume of recognition, not its denomination.

Finding 2: giving pays the giver

Recognition is almost always framed around the receiver. Kumar and Epley's work suggests the giver is also a beneficiary, and also the last to know it. We tested the giver's own well-being. Entered together, giving predicts the giver's well-being at least as strongly as receiving predicts the receiver's, across both measures.

The obvious objection is that happy, popular people simply give more and also feel better. To address it we matched each giver to comparable non-givers in the same company, with the same role, tenure, and platform activity, and crucially the same amount of recognition received. Givers still score higher.

People give recognition expecting to make a colleague feel good. The data says they are also helping themselves. We report this as suggestive: the regression coefficients are marginal and an observational design cannot rule out reverse causality. The matched comparison is what gives it weight, because it holds constant the recognition a person received.

Most people are getting none

If the cheap, frequent gesture is what matters and giving also benefits the giver, then under-recognition is forgone impact on both sides. It is widespread. Four in ten employees received zero peer recognition over a full year. Recognition pools on a minority: within a single company the average Gini coefficient of coins received is 0.60, and the top 10% of recipients capture 39% of all coins.

What recognition does not do

An honest account of the analysis has to include what failed, because it shaped the conclusions. We tested several stronger claims and could not support them.

• No per-event mood jolt. A within-person model predicting next-period happiness from this period's recognition, controlling for prior happiness, is null (effect near zero, p around 0.9). Recognition is not a switch that flips a person's mood for the next two months.
• No effect on eNPS. Recognition does not move within-person eNPS in any specification.
• Retention is not testable here. An exit record on this platform conflates an individual leaving with a whole company offboarding, so the raw "recognized people leave more" pattern is an engagement-and-churn artifact and disappears once activity is controlled. We do not draw a retention conclusion.

Read together, recognition behaves like a frequent, low-cost climate signal whose worth is underweighted, not a high-dosage lever you can pull once for a big effect. A small within-person association does survive (receiving, p=0.014, net of each person's own baseline), which is what keeps the cross-sectional findings from being pure selection.

What this means

The practical implications are unusually concrete, because the cheap behavior is the effective one.

Limitations

• The core findings are observational. Company fixed effects, matching, and a within-person baseline reduce reverse causality but do not eliminate it. This is associational evidence consistent with a causal reading, not a randomized experiment.
• We have no survey of what givers expected versus what receivers experienced. We infer underestimation from the structure of the impact (size-irrelevance, giver parity, concentration), which is the Kumar and Epley pattern demonstrated behaviorally rather than a direct replication of their expectation-versus-experience gap.
• Per-act magnitudes are modest. The argument is that a cheap, frequent behavior is systematically misweighted, not that one coin transforms a person.
• The sample is companies and employees active on Happily with both recognition and survey usage.