The Research · Safe to Great

What actually predicts leadership effectiveness under pressure.

A five-year 360° research programme across pharmaceutical, technology, energy and financial-services organisations. The evidence beneath the Safe2Great system — what moves leadership, what breaks it, and what the aggregate data hides.

1,012
Leaders assessed
360° multi-rater
9,210
Colleagues rating
peers, reports, managers
4
Industries: Pharma
Tech · Energy · Financial
2019–24
Five-year
research programme
Fathom 01 · What we studied

The behavioural layer, not the intentional one.

We analysed 360-degree feedback data from 1,012 leaders rated by 9,210 colleagues across pharmaceutical, technology, energy, and financial services organisations between 2019 and 2024. Fifty-two per cent of the sample sits in Europe, thirty-one per cent in Asia-Pacific, the rest spread across the Middle East and Americas. Every leader was rated on six growth-mindset dimensions, ten protective patterns, and a set of outcome measures covering team performance, psychological safety, engagement, and perceived effectiveness.

The instrument — the Growth Mindset Leadership 360 — is designed to do something most leadership assessments don't. It measures behaviour under pressure rather than general tendencies. It captures growth behaviours and protective patterns in the same frame. And it correlates both against outcome measures rated by the same people, so we can see what actually predicts what.

Fathom 02 · The headline finding

Growth Mindset behaviour predicts leadership effectiveness.

r = .64
Growth mindset behaviour
× leadership effectiveness

That number does more work than it looks like it does. In applied social science, correlations are often reported around .30 for individual predictors of complex outcomes. A value of .64 is among the strongest relationships documented in organisational psychology.

What leaders do under pressure — not what they know, not what they intend, not what their competency framework says they can do — explains roughly forty per cent of the variance in how effective they're perceived to be.

For practitioners, the implication is simple: the behavioural layer is where the leverage is. Personality assessments, competency models, and 360s that measure capability in calm conditions are all useful, but they miss what actually matters. What matters is the repertoire the leader reaches for when something goes wrong.

Fathom 03 · Five findings worth putting on a wall

What the data actually shows.

01

Relational and achievement behaviours predict different outcomes.

For building engaged, psychologically safe teams, the people-focused dimensions dominate. For being seen as promotable, the task-focused dimensions lead. Leaders need the full repertoire — but which dimension carries the weight depends on what outcome you're trying to produce.

Developmental conversations that don't make this distinction tend to produce likeable leaders who don't get promoted, or promoted leaders who leave damage behind.

Lift Others Up × engagement r = .53
Team Up × psych safety r = .48
Aim High × promotion r = .65
Transform × promotion r = .61
02

Disconnected is the most corrosive protective pattern.

Among the ten protective patterns, Disconnected leads on team damage. What makes this pattern particularly difficult is what it does to self-awareness. Disconnected leaders show the largest self-awareness gap in the dataset — they withdraw, detach, and ignore, and they genuinely don't see the damage.

In coaching terms: where a Hippo may know they have a problem but not know how to stop, a disconnected leader often does not perceive a problem at all.

× psych safety r = −.40
× engagement r = −.38
× working together r = −.37
self-awareness gap r = −.52
03

Reactive is the development destroyer.

While Disconnected erodes through absence, Reactive erodes through the wrong kind of presence. Reactive shows the strongest negative correlations with both "challenges others to grow" and "supports others to grow," and it's the only protective pattern that damages the leader's team and their own career simultaneously.

Emotional volatility under pressure is not a quirk. It is a full-system defect.

× challenges others r = −.51
× supports others r = −.45
× productivity r = −.43
× promotion readiness r = −.41
04

"Nice" leadership does not create safety.

Pleasing and Conforming — the patterns we associate with agreeable, accommodating leaders — show surprisingly weak correlations with psychological safety. Being pleasant is not the same as being safe.

What actually predicts psychological safety is a set of specific behaviours rarely labelled "soft": proactive conflict resolution, delivering on commitments, and treating people fairly when they fail. Safety is built through reliability, fairness, and willingness to engage conflict constructively — not through the absence of discomfort.

Pleasing × safety r = −.20
Conforming × safety r = −.17
Conflict resolution × safety r = .49
Fair when they fail × safety r = .43
05

Protective patterns destroy the leader, not just the team.

The internally-focused protective patterns — Doubting, Reactive, Pleasing — show the strongest negative correlations with self-confidence in the entire dataset. The externally-focused patterns (Controlling, Critical, Disconnected) damage teams while leaving the leader's own self-regard largely intact.

Direct coaching implication: the leaders suffering most are often the ones doing the least visible harm, and the leaders doing the most visible harm are often the ones feeling most confident.

Doubting × self-confidence r = −.67
Reactive × self-confidence r = −.54
Pleasing × self-confidence r = −.42
Being pleasant is not the same as being safe.
From the research findings · Safe2Great
Fathom 04 · The longitudinal complication

What the aggregate data hides.

We also ran a 9-month test-retest study with 37 leaders to see whether these behaviours can actually be changed. The aggregate finding is encouraging. The archetype breakdown is less comfortable.

+5.1pp
Growth Mindset
composite (p = .014)
+7.9pp
Transform
(p = .003)
−12.3pp
Conforming
(p < .001)
n = 37
Matched T1/T2
9-month interval

Only Dolphins — leaders without heavy protective armour at baseline — grew across all growth dimensions. Hippos shifted their controlling patterns significantly when confronted with the data, though growth gains were narrower. Snails complied with the workshop target beautifully, but showed no growth elsewhere. Clams went through the motions without emerging from their shells.

The strongest baseline predictor of who would grow wasn't motivation (r = −.28, not significant). It was the absence of Pleasing behaviour (r = −.47, p = .003).

T1 characteristic Correlation with growth p
Pleasing−0.47.003
Resists Change−0.46.004
Dismisses Feedback−0.39.016
Motivated to Develop−0.28n.s.

The practitioner takeaway: standard leadership development works best for the leaders who need it least. For leaders with heavy Pleasing, Disconnected, or Clam patterns, the default programme produces compliance rather than change.

Different armour requires different interventions — and screening for Pleasing patterns at the front end of a cohort is the single most useful piece of calibration you can do.

Standard leadership development works best for the leaders who need it least.
The longitudinal complication · Safe2Great Research
Fathom 05 · What this means for how you design development

Five design principles drawn from the data.

Make protective patterns visible.

Self-perception is systematically biased. The gap between how leaders see themselves and how others experience them is where development begins. Use multi-rater feedback.

Teach conflict engagement as a core skill.

Proactive conflict resolution is the single strongest predictor of psychological safety in the dataset. Most programmes still treat it as a peripheral topic. It shouldn't be.

Stop rewarding "deliver but destroy" leadership.

Safety and performance are not a trade-off. Leaders who build the safest teams also build the most productive ones. Organisations that keep promoting leaders who hit numbers while eroding team capability aren't making a hard choice — they're making the wrong one.

Measure at multiple levels.

Individual 360s can miss aggregate cultural shifts that only become visible in a culture-level survey. Use both. The Safe2Great suite runs across leaders, teams, and organisations using the same underlying model.

Match intervention to archetype.

External armour (Hippo patterns) responds to feedback and confrontation. Approval-seeking armour (Snail patterns) requires safety to experiment with disapproval. Internal armour (Clam patterns) often needs longer timeframes, individual rather than group work, and sometimes a different kind of help than a leadership programme can provide. One-size-fits-all cohort design is the single most common cause of visible compliance without durable change.

Methodology · Sample · Compliance

How the study was run.

Sample. 1,012 leaders rated by 9,210 colleagues across pharmaceutical, technology, energy, and financial-services organisations, 2019–2024. Geographic distribution: 52% Europe, 31% Asia-Pacific, remainder Middle East and Americas.

Instrument. The Growth Mindset Leadership 360 (GML360), measuring six growth-mindset dimensions, ten protective patterns, and a set of outcome measures covering team performance, psychological safety, engagement, and perceived effectiveness. Raters are peers, direct reports, and managers.

Longitudinal substudy. 9-month matched T1/T2 retest with 37 leaders across multiple organisations. Archetype assignment based on T1 protective pattern profile.

Parallel organisational data. Culture for Growth survey at Danfoss Northern Europe Region (n = 174). See the Danfoss case study for applied findings.

Compliance. All Safe2Great instruments comply with European GDPR requirements and with Danish Psychological Association standards for the safe and responsible use of psychometric tools.

Research programme. Ongoing. The dataset continues to grow, and new findings are integrated into practitioner training and programme design.

For the full paper, or a conversation about applying the findings.

The research summary on this page draws on a forty-page paper available to HR leaders, L&D specialists, and executive coaches who want to go deeper — or who want to discuss applying the findings inside their own organisation.

Request the full paper → Talk to Skip →