Organisations talk a lot about becoming “data-driven”.
Strategies are written. Platforms are built. Dashboards appear everywhere.
Yet something strange often happens when organisations try to talk seriously about what the data is showing.
The conversation changes.
A question about two different figures becomes a dispute about methodology.
A decline in performance becomes an argument about context.
A request for greater transparency begins to feel like an accusation.
People who were comfortable discussing data in principle become defensive when the evidence gets close to a real decision, an established position or somebody’s area of responsibility.
I started thinking of this phenomenon as the Data Werewolf.
The metaphor originally emerged as a way of making a complicated organisational problem easier to discuss. But the more I explored it, the more useful it became.
The werewolf is not the data.
It is not the analyst presenting it.
It is not necessarily the person challenging it.
It is the change that takes place when evidence enters an organisation without enough shared language, trust or clarity for people to examine it together.
Most organisations genuinely want to use data better.
Leaders talk about evidence-based decisions. Data teams build platforms, pipelines and reports. Services ask for insight that will help them understand demand, performance and outcomes.
The difficulty begins when the information carries consequences.
Questions such as:
Why are these numbers different?
What explains this pattern?
Why is performance declining here?
can be entirely reasonable.
They can also threaten an accepted account of how well something is working, who is responsible or what should happen next.
At that point, the discussion is no longer only about the number.
It is also about competence, identity, authority, resources and risk.
A report may be technically neutral. The organisational environment into which it arrives rarely is.
That is the moment the werewolf appears.
The academic work behind this article used the 5 Whys as a way of following resistance beyond its most visible form.
The exact chain varies, but it often looks something like this.
Why has the conversation become defensive?
Because the evidence challenges an established understanding of performance.
Why does that feel threatening?
Because accepting the evidence may require an explanation, a decision or a change in direction.
Why does that feel personal?
Because responsibility, professional identity and service performance have become closely connected.
Why can the group not examine the evidence collectively?
Because definitions, ownership and confidence are uneven. People are not necessarily working from the same understanding of what the measure means or how it was produced.
Why does that situation persist?
Because the organisation has invested more consistently in systems and reporting than in shared literacy, governance, trust and the behaviours needed to learn from what the systems reveal.
The first reaction may therefore be about a particular number.
The underlying cause may sit much deeper in the organisation.
This is why replacing one dashboard with a better dashboard rarely solves the problem.
The reaction tends to take several recognisable forms.
Sometimes the easiest response is not to engage.
The data is described as incomplete, premature or insufficiently understood. The discussion moves elsewhere. A further piece of analysis is requested, followed by another.
Caution may be justified. Data can be incomplete.
But an endless requirement for greater certainty can also prevent the organisation from acting upon what is already visible.
Avoidance is not always evidence of bad faith. It may reflect low confidence. Somebody who does not feel equipped to interpret a measure may be reluctant to expose that uncertainty in front of colleagues.
Silence can be a literacy problem disguised as indifference.
The second reaction is to challenge definitions, sources or methodology.
Again, this can be legitimate.
Two reports may use different populations. A performance indicator may have changed. A figure may be technically correct while still creating a misleading impression.
Healthy data cultures need people who question evidence rather than accepting whatever appears on a dashboard.
The difficulty comes when methodological scrutiny is applied selectively: welcomed when the evidence supports the preferred narrative, but treated as disqualifying when it does not.
A dispute about quality may also reveal a deeper problem. Nobody may be clearly responsible for agreeing definitions, maintaining shared measures or explaining how the figures were produced.
The argument is then not simply about trust in the number.
It is about trust in the system around it.
The third reaction is to reshape the interpretation.
Caveats multiply. Comparators are changed. The time period becomes unusually important. The measure is reframed until the original signal becomes less uncomfortable.
Context matters. Numbers do not interpret themselves.
But storytelling can be used either to illuminate evidence or to domesticate it.
When the story is allowed to change but the underlying assumption is not, data becomes decoration around a decision that has already been made.
These responses are human. Most of us have probably used some version of them.
The point of the metaphor is not to identify villains.
It is to recognise the conditions under which reasonable people repeatedly produce defensive conversations.
The original research compared public-sector approaches and found the same broad lesson appearing in different forms: the difficulty is rarely technical alone.
The UK Government’s Future Councils pilot later described similar systemic barriers to digital transformation across local government. Skills, governance, organisational silos and culture do not operate as separate problems. They reinforce one another.
A fragmented organisation is likely to hold fragmented data.
Fragmented data produces competing measures and repeated reconciliation work.
Competing measures weaken trust.
Weak trust encourages more control, more caveats and greater dependence on a small number of specialists.
That dependence then makes data feel even less like a shared organisational asset.
This is how a technical problem becomes cultural, and a cultural problem becomes structural.
The contrast between ambition and practice matters here. An organisation can genuinely believe that data is important while still behaving as though it belongs to somebody else.
Strategies may describe data as a corporate asset. Daily conversations can continue to treat it as an IT product, a performance-management weapon or a compliance obligation.
Those are not merely different descriptions.
They lead to different behaviour.
Governance is essential.
Public organisations hold sensitive information. Decisions about access, quality, ownership and appropriate use cannot be casual.
But governance also teaches people what the organisation believes data to be.
If every conversation begins with restriction, approval and risk, people learn that data is hazardous.
If ownership is fragmented, they learn that data belongs to whichever specialist team appears to control it.
If governance is visible only when somebody wants to stop something, it becomes associated with compliance rather than value.
Good governance does not remove control. It makes responsible use easier.
It should help people answer practical questions:
Who owns this information?
Who agrees the definition?
Where did the figure come from?
Who can explain its limitations?
How can it be used safely?
What happens when two services need to work across an organisational boundary?
Without those answers, governance may exist on paper while uncertainty continues in practice.
That uncertainty has a cultural effect. People either retreat from the data or become dependent upon a small group of specialists who are expected to mediate every conversation.
Neither response creates a data-informed organisation.
Essex County Council’s strategy, Data is Everyone’s Business, captured an important alternative in its title. Data can be protected and governed without being treated as the private territory of a technical function.
Data literacy is often framed as a programme for teaching non-technical people how to read charts or use analytical tools.
That matters, but it is only half the problem.
Leaders and service professionals need enough confidence to question evidence, understand uncertainty and recognise the difference between a useful challenge and a convenient objection.
Data professionals also need sufficient understanding of the organisation, the service and the decisions being made.
A technically correct answer can still be organisationally useless.
The people closest to the data may not understand the operational context. The people closest to the service may not understand how the measure was constructed.
Both groups can then leave the same conversation believing the other has failed to understand something obvious.
Literacy is therefore relational.
It is the development of enough shared understanding for different forms of expertise to meet.
That takes more than a training course. It develops through repeated conversations, practical examples, curiosity and the freedom to say:
I don’t understand how this figure was produced.
This result does not match my experience.
What would we need to examine to understand the difference?
Those sentences can begin a defensive exchange.
They can also begin organisational learning.
The difference is often the way leadership receives them.
Leaders shape data culture through their behaviour long before they approve a strategy.
When leaders use data to assign blame, people learn to hide from it.
When they demand certainty from incomplete evidence but make intuitive decisions without equivalent scrutiny, people notice.
When they challenge an inconvenient figure aggressively but accept a favourable one without question, they teach the organisation what the conversation is really for.
The alternative is not passive acceptance of every number.
It is disciplined curiosity.
A leader can ask difficult questions without turning the conversation into an interrogation.
They can acknowledge uncertainty without using it as an excuse for inaction.
They can change their mind publicly when the evidence justifies it.
They can separate understanding what happened from deciding who should be blamed.
These behaviours create psychological safety, but psychological safety does not mean comfort or the absence of accountability.
It means people can surface problems, question assumptions and admit uncertainty without calculating first whether doing so will damage them.
That is essential because the most valuable evidence is often the evidence that interrupts the organisation’s preferred account of itself.
The Data Werewolf never disappears completely.
Evidence has consequences. Resources are limited. Services are complex. Measures are imperfect. People care about their work and do not experience questions about it as entirely abstract.
The aim is not to remove every difficult reaction.
It is to create an organisation capable of remaining in the conversation after the reaction begins.
That requires:
- shared definitions rather than competing versions of the truth;
- visible ownership rather than responsibility scattered across teams;
- governance that enables responsible use rather than merely restricting access;
- data literacy built through practice and conversation;
- data professionals who understand the business as well as the tools;
- leaders who model curiosity when the evidence is uncomfortable;
- enough trust for disagreement to become enquiry rather than defence.
The organisations that succeed with data will not necessarily be those with the most sophisticated platforms.
They will be those able to turn information into shared understanding, and shared understanding into action.
The promise of data is often described as clarity.
But clarity is not produced by the dashboard alone.
It emerges through the quality of the conversation around it.
The werewolf appears when evidence gets too close to something the organisation is not yet ready to examine.
Taming it begins by asking why.
And then being willing to remain present long enough to hear the answer.
This essay is an accessible adaptation of academic work exploring why conversations about data become difficult. The research drew on organisational literature, a 5 Whys analysis and comparative public-sector case studies, and received a high distinction.
A later research project developed related questions through primary research inside a local authority. Those internal findings are not reproduced here.
The analysis and views expressed are my own and do not represent my employer or any other organisation.