When we were engaged by NOAA's National Environmental Satellite, Data, and Information Service in 2021, the agency had already spent five years trying to transform itself. Four sets of governance guidelines. Multiple consulting engagements. New directives from the top. None of it had changed anything. The strategic plan calling for innovation and new partnerships had been written in 2016. By the time we arrived, the words were still there. The behavior they described was not.
The pattern NESDIS had been stuck in is one I recognize immediately, because I see it in nearly every organization that calls us after a failed strategy cycle. The decisions were made at the top, communicated downward, and quietly ignored — not out of malice, but because the people expected to act on them had no ownership of the decisions. They had been told what to do. They had not been asked to build anything. Every meeting had ended with apparent consensus. None of them had produced real commitment.
Dr. Steve Volz, NESDIS's Director, made a deliberate choice for our session that I think was the most important design decision in the entire engagement: he removed himself from the deliberation. Not from the room — he participated with the specific role of asking challenging, provocative questions — but from the driver's seat. He created the space for 24 change agents to build a strategy without him at the centre of it. When those 24 people finally stopped waiting for direction from above and started resolving things themselves, the room changed. They weren't performing consensus. They were building something.
Why rooms agree before they're ready to
Lovallo and Sibony's research on cognitive bias in strategic decision-making identifies authority deference as one of the five families of social bias that operate in every unstructured group deliberation. When participants can infer the views of the most senior person in the room — from how the question is framed, from the tone of the first contribution, from the body language around the table — they rapidly update their stated positions toward the authority-consistent view. This is not cynical calculation. It is a deeply adaptive human response. The cost of being publicly wrong in front of a dominant voice is high. The cost of quiet agreement is zero.
The result is a room that looks aligned while the genuine distribution of views has never been surfaced. The risks that participants privately understood don't disappear — they become execution failures, surfacing in corridor conversations after the session, in the bilateral calls where people compare what they didn't say. The person with the least organizational power is often the one who knows exactly why the strategy won't work. They are also the least likely to say so in an unstructured room.
Structure does what culture cannot
Cultures of psychological safety help interrupt premature consensus — but they take years to build and cannot be manufactured for a single session. What can be designed for a single session is a structural mechanism that makes dissent a process requirement rather than a social choice.
The role we used with Dr. Volz is one version of this. A formal devil's advocate function is another. Both depersonalize the challenge: it becomes a process requirement, not a personal confrontation with authority. Simpler still: written individual assessments captured before group discussion begins reveal the actual distribution of views before social pressure has organized them into apparent agreement.
When participants record their positions before anyone knows what the group will converge on, you can see the real picture. A narrow, consistent distribution usually means the consensus that follows is genuine. A wide distribution followed by rapid convergence in discussion means the room agreed — but the room wasn't ready to.
NESDIS left with 18 recommendations and a mandate to present them to 50 senior leaders at the agency's Fall Strategy Meeting. More than that, they left having built something rather than been handed it. That distinction — between a room that performs consensus and one that earns it — is what determines whether the strategy survives contact with implementation.
Frequently Asked Questions
Why is fast consensus in a strategy session a warning sign rather than a success signal?
Fast consensus in a group deliberation about a genuinely complex strategic question almost always indicates that social dynamics have produced convergence before the genuine distribution of views has been surfaced. Authority deference causes participants to update their stated positions toward the most senior person's apparent view, regardless of the quality of the underlying reasoning. The risks that were privately understood but not raised do not disappear from the system — they become execution failures.
What structural mechanisms interrupt premature convergence in strategy sessions?
The research points consistently toward structural rather than cultural solutions. A formal red team or devil's advocate role depersonalizes dissent — it becomes a process requirement rather than a social choice. Written independent assessments captured before group discussion reveals the actual distribution of views. When participants record their positions before anyone knows what the group will converge on, you can distinguish genuine alignment from social bias producing rapid convergence.
How does authority deference produce poor strategic decisions?
Lovallo and Sibony's research identifies authority deference as one of the five families of social bias operating in every unstructured group deliberation. When participants can infer the views of the most senior person in the room, they rapidly update their stated positions toward the authority-consistent view. The result is decisions that reflect the views of the most powerful person rather than the collective intelligence of the group — and the stakeholder with the least organizational power is simultaneously most likely to hold the critical insight about why the strategy won't work.