Organisations short of innovation usually go looking for ideas — hackathons, innovation weeks, suggestion schemes. The ideas are almost always already present. What is missing is the environment in which someone is willing to raise one.
An innovation event assumes the constraint is idea supply. In most organisations it is not. The constraint is the cost of raising something that might fail, or that implies a current approach is wrong.
That cost is set by leadership behaviour and by whether the last person who raised something was heard. An event cannot lower a cost that is set the other fifty weeks of the year — which is why the ideas generated in the room so often go nowhere afterwards.
Each figure compares employees with the strongest score on a condition against those with the weakest, on innovation.
Leadership is the widest gap in this category, and the mechanism is not mysterious. Innovation requires proposing something unproven, which requires believing the proposal will be judged on merit rather than held against you.
That belief is a function of observed leadership conduct: whether reasoning is explained, whether standards are applied consistently, whether raised concerns visibly go anywhere. Fairness appears repeatedly in the condition-level data, which suggests innovation is closer to a trust outcome than a creativity one.
What creates trust and what destroys it: Leadership & Trust.
Resilience appearing second in this category is the finding worth pausing on. Innovation is discretionary work — nobody is required to propose an improvement — and discretionary work is the first thing a depleted person stops doing.
A workforce with no reserve does not innovate, however good the conditions are, because there is nothing left over to spend on anything beyond the role. Both halves matter: engagement sets whether people would, capacity sets whether they can.
Whether people can see why the work matters: Purpose & Meaning. Where innovation pressure concentrates by sector: use cases.
Figures come from thrive at work’s own validation study (n=10,000+, 2025–26), fielded across the US and Europe. Each compares employees with the strongest scores on a given condition against those with the weakest, on the outcome named. Each outcome is a score, not an operational record.
These are associations measured within a single cross-sectional study. They show that the two groups score differently; they are not a claim that changing one score produces the other. No score is displayed for any group of fewer than five people.
A 30-minute call with the research team covers what a first measurement would surface, and how your sector compares.