Trust Networks: why most of your most-trusted people aren't managers

Every organization runs on two structures. One is the org chart: formal, documented, and the basis for almost every people decision a company makes. The other is the trust network: who asks whom for advice, whose judgment people seek out, who gets recognized and recognizes back. It is undocumented, and it rarely matches the org chart.

We set out to measure the second structure directly. Using a year of peer-feedback and recognition data from 3,446 employees across 31 companies, we built a per-company trust score and asked a simple question: when you rank everyone by how much their colleagues trust them, who is at the top?

The answer: 72% of the most-trusted decile are individual contributors, not managers. This pattern held across 31 companies and survived a conservative statistical correction. But the finding has a second half that matters just as much, and most summaries miss it.

How we measured trust

Trust is hard to self-report and easy to fake. So we did not ask anyone how trusted they felt. We measured two behaviors that only happen when trust already exists.

The first is peer-feedback solicitation. On Happily, an employee can request feedback from a specific colleague. Choosing someone as a feedback giver is a deliberate signal: the requester is saying "I value your judgment." We counted, for each person, how many distinct coworkers requested feedback from them over the year.

The second is reciprocal recognition — pairs of people who recognized each other in both directions. One-directional recognition blends genuine trust with simple visibility; a bidirectional tie is a stronger signal that the relationship is real.

We combined the two into a single composite, standardized within each company so that a high score means "trusted relative to this company's own norms," not "works at a company with a chatty culture."

Peer-feedback carries 60% of the weight because being chosen as a feedback giver is the cleaner trust signal. Recognition carries 40%: it captures more people (51% of the workforce versus 40% for peer-feedback) but mixes trust with activity volume. A sensitivity check with equal weighting produced near-identical rankings, so the split is not load-bearing.

Finding 1 — The top of the trust ranking is mostly ICs

We ranked every employee by trust score within their company, then looked at who occupied the top 10%, 20%, and 30%.

Read this carefully, because it cuts both ways. "72% of the top decile are non-managers" is the headline, and it is true. But the workforce is 80.5% non-manager. So managers actually appear in the top decile at 28%, against a 19.5% baseline — a 1.4× over-index. Seniority does correlate with trust. Promotion is a real, if imperfect, trust signal.

What the headline gets right is scale. Because the base population is so heavily IC, the count of trusted non-managers is large: hundreds per organization. A company picking change ambassadors only from its management layer is ignoring the majority of the people its workforce actually relies on.

Finding 2 — The two distributions overlap heavily

If managers were uniformly more trusted, the two groups would form separate peaks. They do not.

Managers lean higher, and the gap is statistically unambiguous (Welch's t-test, p = 1.4 × 10⁻¹⁰). But the effect size is small: Cohen's d = 0.29. The practical consequence is the number that should change how you build talent shortlists — 35.2% of non-managers score above the manager median. One in three individual contributors is more trusted than the typical manager.

Finding 3 — Trusted ICs hold real network position

Being chosen often could, in principle, just mean a person is liked by a small, chatty clique. To rule that out we measured eigenvector centrality — the network-science signature for influence, weighting a connection more heavily when it comes from someone who is themselves well-connected.

Among the top trust decile, managers (n=48) averaged an eigenvector centrality of 0.158; non-managers (n=138) averaged 0.114. A trusted non-manager sits at roughly 70% of a trusted manager's centrality, and once Bonferroni-corrected the difference is not significant (p = 0.62). Trusted ICs are not popular within a corner of the org. They are genuinely central.

Per-capita versus absolute: how trust actually flows

This finding has two readings, and choosing the right one depends on the decision in front of you.

Per capita, a top-trust manager is about 40% more central than a top-trust IC. If the question is "does trust track seniority?", this view says yes.

In aggregate, non-managers carry roughly twice the total network throughput, because there are nearly three times as many of them in the top decile. If the question is operational — who should carry a change rollout, seed a pilot, or move information across teams — the answer is that the path almost always runs through a non-manager.

Hidden influencers are everywhere

To make the pattern concrete, we drew the peer-feedback network for one representative mid-size company: 204 active employees and 448 trust ties.

The three labeled nodes are this company's top-3 trust scorers. Each was chosen by 8 or more distinct colleagues as a feedback giver over the year. None is a manager. They sit in the densest region of the network, with ties that cross the org chart into several managerial clusters.

This was not a lucky pick. We repeated the top-5 ranking across 6 representative companies.

In every one of the six companies, the top-5 trust scorers were entirely non-managers, with trust z-scores from +1.8 to +5.1 — genuinely exceptional positions. In the largest company shown (742 active employees), five individual contributors absorbed more peer-feedback requests and reciprocated more recognition than any of the company's 140-plus managers.

What this means for HR

The trust network is not a curiosity that sits alongside the org chart. It is the structure through which culture, information, and change actually move. Three decisions should use it directly.

The simplest version of the rule: a non-manager with a trust z-score of +3 is a better change ambassador than a manager with a trust z-score of 0. The title is not the signal. The behavior is.

Limitations

This study measures the existence and shape of trust networks, not their downstream effect. A few caveats shape how far the findings travel.

• Recognition can be circular. People who send more recognition tend to receive more back, so reciprocal recognition partly measures activity. Weighting it at 0.4 mitigates this but does not remove it.
• Peer-feedback is sparse. Only 40% of employees were sought as a giver at least once in the year. A negative trust score should be read as "no signal," not "distrusted."
• Tenure is uncontrolled. The 365-day window mixes new hires with long-tenured ICs, and managers tend to be more tenured. Tenure may partly confound the manager-IC gap.
• Company size matters. Larger companies give hidden influencers more room to stand out; smaller companies may show flatter distributions.
• Regional skew. The 48 companies are predominantly Southeast Asian workforces. Generalization beyond that is not yet validated.
• No outcome test. We have not yet linked trust scores to eNPS, attrition, or engagement. That is the natural follow-on study.