The recognition cascade: why it flows from the top, not the boss
Reply behavior cascades through your direct manager. Recognition behavior does not. It flows from senior leaders two levels up — through visibility, not through the reporting line.
An earlier study on the Happily platform showed that one manager behavior, replying to feedback, cascades downhill through the org chart. A manager whose own boss replies is 2.4x more likely to reply to their own team. The effect runs through the direct reporting line: your boss models the behavior, you copy it, your reports copy you.
That raised an obvious follow-up. Does every manager behavior cascade the same way? We took a second behavior — whether a manager actively recognizes other people — and ran the same chain analysis on 908 managers. The expectation was a smaller version of the same picture: your direct manager's recognition habits would predict yours.
They do not. Recognition behavior transmits, but it does not transmit through your direct boss. The signal that predicts whether a manager is an active recognizer comes from two levels up — the manager's manager's manager. Direct manager influence is not statistically significant at all. Recognition and reply behavior look like the same kind of thing, but they spread through completely different routes.
If you want to build a recognition culture, the intuitive move is to train every manager to recognize their teams. This research says that effort lands in the wrong place. Recognition is a public behavior — people watch what senior leaders do, not what their immediate boss does. The lever is executive visibility, not manager training.
How we measured the recognition cascade
The behavior under study is active recognition: whether a manager regularly recognizes other people on the platform. We reconstructed each company's reporting hierarchy by following the reporting line up from each manager — their direct manager (L1), their skip-level (L2), and the level above that (L3) — and asked whether a manager's recognition behavior is predicted by the recognition behavior of the people above them at each level.
This mirrors the method used in the earlier reply-behavior study, so the two cascades can be compared directly. The difference in results comes from the behavior, not the method.
Finding 1 — The influence multiplier is real but modest
At the headline level, recognition does transmit. A manager whose leadership recognizes is more likely to be an active recognizer themselves. The probability of being an active recognizer rises from 23% when leadership does not recognize to 56% when it does — an absolute lift of 33 percentage points.
| Leadership behavior | P(manager is a recognizer) |
|---|---|
| Leadership does not recognize | 23% |
| Leadership recognizes | 56% |
| Absolute lift | +33 pts |
The logistic model puts the odds ratio at 1.6x (95% CI: 1.15–2.34, p = 0.006). The effect is statistically solid, but it is smaller than the 2.4x multiplier found for reply behavior. The interesting result is not the size of the multiplier — it is where the signal comes from.
Finding 2 — The direct manager does not transmit recognition
Reply behavior cascades through the direct boss. We expected recognition to do the same. It does not.
The first regression model tested whether a manager's direct manager (L1) predicts their recognition behavior. The L1 coefficient is 0.040, and it is not statistically significant (p = 0.19). Whether your own boss recognizes people tells you essentially nothing about whether you do.
| Term | Value |
|---|---|
| L1 coefficient | 0.040 |
| p-value | 0.19 |
| R² | 31.7% |
| Sample | n = 908 |
The model still explains 31.7% of variance, but almost all of that comes from the company baseline, not the direct manager. Recognition culture is set somewhere — just not at the L1 link.
Finding 3 — Influence comes from two levels up
Adding higher levels to the model is where the picture changes. When skip-level (L2) and the level above (L3) recognition behavior enter the regression, the L3 term carries a large, positive, significant coefficient.
| Level | Who this is | Coefficient | Significant? |
|---|---|---|---|
| L1 | Your direct manager | −0.068 | No |
| L2 | Your skip-level | −0.304 | Yes (p < 0.001) |
| L3 | Two levels above your boss | 0.708 | Yes (p = 0.016) |
| L4 | Three levels above your boss | 0.441 | Not tested for significance |
Two terms stand out. The L3 coefficient is 0.71 and significant (p = 0.016): the recognition habits of senior leaders two levels above a manager strongly predict that manager's own recognition behavior. The L2 coefficient is negative (−0.19, p < 0.001 in the skip-level model): when your skip-level recognizes a lot, you are slightly less likely to.
One plausible explanation for the negative skip-level term is substitution. If the manager directly above your boss is a heavy recognizer, the recognition that matters in your part of the org may already be getting done — you have fewer occasions, or feel less need, to recognize people yourself. It is a hypothesis the data is consistent with, not a proven mechanism.
Note that the L3 estimate rests on a smaller sub-sample (n = 81 managers with chains deep enough to have L4 data). The direction is clear and the result is significant, but the precise size of the coefficient should be read with that sample in mind.
Why recognition spreads differently
Reply behavior and recognition behavior diverge because of one difference: who can see the behavior.
Replies are private. When a manager replies to feedback from a direct report, the only person who sees it is that report. You observe your own manager's reply habits because the replies are addressed to you. You cannot observe how a senior leader four offices away replies to their team. So reply behavior can only be modeled by the person directly above you, and the cascade runs link by link.
Recognition is public. When a leader recognizes someone, the whole organization can see it. A senior leader's recognition pattern is visible across the company in a way their reply habits are not. Managers watch what executives do, and they take their cue from that visible behavior — not from their immediate boss, whose recognition habits carry no special weight.
This is the side-by-side picture from the two studies:
| Metric | Reply behavior | Recognition behavior |
|---|---|---|
| Odds ratio | 2.4× (p < 0.001) | 1.6× (p = 0.006) |
| Probability lift | 49% → 81% (+32 pts) | 23% → 56% (+33 pts) |
| L1 (direct manager) | 0.19 (significant) | 0.04 (not significant) |
| L2 (skip-level) | −0.05 (not significant) | −0.19 (significant) |
| L3 (two levels up) | −0.07 (not significant) | 0.71 (significant) |
| Sample | 588 managers | 908 managers |
Same platform, same method, two behaviors. Reply behavior is local — it lives at L1 and dies above it. Recognition behavior is the mirror image — nothing at L1, the signal sits at L3.
What this means for HR
The practical consequence is that culture change is not one problem. Different behaviors need different interventions, and treating them the same wastes effort.
| Decision | What the recognition cascade tells you to do |
|---|---|
| How to build a recognition culture | Get senior leaders to recognize people visibly. Their behavior sets the tone that flows down through visibility — manager-by-manager training is not the lever here. |
| How to build a feedback-reply culture | Train every layer of management. Reply behavior only transmits through the direct boss, so one broken link stops the cascade. Saturation, not top-down messaging. |
| Where to spend recognition-program budget | On executive visibility and modeling, not on cascading recognition quotas to every frontline manager. The L1 link does not carry the signal. |
| How to diagnose a weak culture | Identify the behavior first. A recognition gap points upward to senior leaders; a reply gap points to a specific broken manager link in the chain. |
A one-size-fits-all approach to culture change does not work. Recognition is shaped from the top through what people can see; reply behavior is shaped link by link through direct observation. If you run the same playbook for both, one of the two efforts is aimed at the wrong level of the org chart.
Limitations
This study measures association, not proven causation. A few caveats shape how far the findings travel.
- Correlation, not causation. We cannot prove managers take their recognition cue from senior leaders. Selection effects — organizations promoting recognizers into senior roles — could produce a similar pattern.
- Noisier signal. Recognition counts range from 0 to 16 per month, a wider and noisier measure than the 0–1 reply rate. Some of the difference between the two cascades may reflect measurement properties, not just behavior.
- Small upper-level sample. The L3 effect rests on 81 managers with chains deep enough to carry L4 data. The direction is clear and significant, but the exact coefficient size should be read with that sample in mind.
- Company baseline dominates. Company-level effects explain roughly 30% of the variance in recognition behavior. Culture is largely set at the company level, and some unmeasured company factors may remain.
- Two behaviors only. Reply and recognition are two observable manager behaviors. Whether other behaviors transmit locally, from the top, or some other way is an open question.
Happily Research (2026). The Recognition Cascade: Why It Flows From the Top, Not the Boss. happily.ai/research/recognition-cascade/
References
- Bandura, A. (1977). Social Learning Theory. Prentice Hall. Foundational account of behavior acquired through observation and modeling — the mechanism this study tests for a public versus a private behavior.
- Happily Research (2026). The Leadership Cascade: How Manager Behaviors Flow Downhill. Internal analysis, 588 managers, 180-day window, January 2026.
- Happily Research (2026). The Recognition Cascade: Behavior Transmission for Reply vs Recognition. Internal analysis, 908 managers, January 2026.
Happily turns everyday recognition into a live picture of who models it and where culture is set — so you can aim culture change at the level that actually transmits.
Get in touch