Having dashboards is not the same as making decisions with data. Every organisation I've walked into in the last three years has told me it's data-driven. Every one of them has Looker, Tableau, or Power BI deployed at non-trivial cost. Maybe 20% of them actually change decisions when the data contradicts the instinct. The rest have built an expensive theatre of rationality.
This is the core confusion: data-driven culture is not about producing data. It's about letting data override the assumption when they conflict. That's a governance and incentive problem, not a tooling problem. And it's one that no dashboard vendor will solve for you.
The Dashboard Trap
Here is the pattern I see, reliably, across sectors: an organisation realises it lacks data visibility. It invests in a modern data stack — dbt, BigQuery or Snowflake, a BI layer. Reports are built. Dashboards are shared. Leadership adds the dashboard links to the weekly update template. Everyone nods in meetings when the charts go up.
And then nothing changes.
The product team still ships the feature the CPO always wanted to ship. The sales team still chases the segment the CCO personally believes in. The marketing team still runs the campaign format they've always run. The data sits on the screen, decorating decisions that were already made.
I call this dashboard theatre: the performance of data-drivenness without the substance. It is extremely common and expensive. The BI stack costs money. The data engineers who build and maintain it cost money. The analyst time spent producing reports that nobody acts on costs money. And none of it produces the actual value that data-driven decision-making can produce — better predictions, faster learning, compound advantage over time.
What Data-Driven Actually Means
Let me be concrete about the definition I use when assessing whether an organisation is actually data-driven:
A decision is data-driven when the data had the power to produce a different outcome. If the data could only confirm the pre-existing view, the decision was instinct-driven with data decoration.
This means three things must be true before any decision can legitimately be called data-driven:
- A clear question was posed before the data was pulled. Not "show me the numbers" — "what would we need to see in the data to change the roadmap priority?" If the question isn't posed first, confirmation bias fills the gap.
- The decision-maker committed to a threshold in advance. "If retention is below 40% at the 30-day mark, we kill this feature." Not "if the data is concerning." Concerning is subjective. 40% is not.
- There is a recorded instance of the data producing a different decision than the instinct pointed toward. If no such record exists, the data has never actually driven anything.
Most organisations fail all three. Dashboards answer the question "what happened" without any prior commitment to what "what happened" should change. That's reporting, not decision intelligence.
The Incentive Structure That Preserves the Theatre
The data-driven theatre persists because the incentives around it are perverse in a specific way.
Leaders who push back on data they dislike — who say "I don't believe that number" or "our situation is different" — almost never face consequences for that pushback. The organisation doesn't track whether their gut override was correct. The data team builds another report. The cycle repeats.
Meanwhile, the leaders who do act on data that contradicts their instinct face a significant career risk if the data turns out to be wrong or incomplete. "You made this decision based on a metric that turned out to be misleading" is a much more visible failure than "you ignored the data and the gut call didn't work either."
This asymmetry — high visible accountability for data-driven failures, low accountability for data-ignoring failures — is what kills data culture more reliably than any technical deficit.
The accountability asymmetry
Until a leader is visibly held accountable for overriding data with instinct — and that override turns out to be wrong — the rational response is to use data as decoration rather than driver. This is not a character flaw. It is a predictable response to the incentive structure. Changing the culture means changing the accountability structure, not deploying another dashboard.
Dashboard-Driven vs. Decision-Driven: A Comparison
| Dimension | Dashboard-Driven (Theatre) | Decision-Driven (Actual) |
|---|---|---|
| Question sequence | Data pulled, then question formed | Question posed, then data pulled |
| Decision threshold | Qualitative ("if it looks concerning") | Quantitative, committed in advance |
| Data override rate | Near zero — data confirms instinct | Measurable — some decisions reversed by data |
| Metric ownership | Collective (everyone, therefore no one) | Named owner per KPI with explicit accountability |
| Report cadence | Weekly/monthly, regardless of decision cycle | Triggered by decisions, not by calendar |
| Data quality investment | Minimal — dashboards built fast, trusted loosely | High — decisions depend on it, trust is earned |
| Cultural signal | "We have great dashboards" | "The data changed our minds last quarter" |
If your organisation sounds more like the left column, you're paying for theatre. The cost is not just the BI stack — it's the opportunity cost of decisions made by instinct that could have been made better.
What I Actually Do in Engagements
When I join a transformation engagement and data culture is on the diagnostic list, I do three things before recommending any tooling change.
First, I audit the decision log. I ask for the last 20 significant decisions made by the leadership team — product, commercial, operational — and I ask for the data artefact that accompanied each one. Then I ask whether the data could have produced a different outcome, and whether it did. In most organisations, the data was consulted after the decision was directionally committed. In some, there was no data artefact at all. In a few, the data genuinely changed the direction. Those few are the reference point.
Second, I identify the highest-stakes recurring decision. Every organisation has one: a weekly or monthly decision that significantly affects resource allocation, customer outcomes, or commercial performance. I focus the data culture intervention on that one decision first. Not the whole business. One decision. Make it genuinely data-driven — prior question, committed threshold, recorded outcome. Then propagate the practice.
Third, I establish metric ownership. Every metric that matters gets a named owner. That person is responsible for data quality, for surfacing the metric when decisions require it, and for recording when the metric influenced a decision. Without named ownership, data quality degrades and accountability diffuses.
The one-decision intervention
Start with the highest-stakes recurring decision in your business and make it genuinely data-driven — not theatre. That means a written question posed before pulling data, a numerical threshold committed in advance, and a documented record of the outcome. Once one decision is genuinely data-driven, the practice has a reference point. Propagation is much easier than cold-start.
The Data Quality Problem That Nobody Wants to Talk About
There is a reason that data-driven culture is hard beyond incentives: the data is often not trustworthy enough to make confident decisions.
I've seen organisations with beautiful dashboards built on top of data pipelines that have never been tested for correctness. Numbers that conflict between systems — the CRM says one revenue figure, the data warehouse says another, finance says a third. Metrics that are defined differently in different reports. Dimensions that changed meaning when the product changed without the data model catching up.
In this environment, the rational response to "trust the data" is actually "don't trust the data." The leadership team has often been burned — they acted on a metric, the metric turned out to be computed incorrectly, the outcome was bad. The lesson they drew was "data is unreliable." That lesson is wrong only in direction: the data is unreliable, but the solution is to make it reliable, not to return to instinct.
Data quality is a prerequisite for data culture. This seems obvious. It is routinely ignored. Organisations invest in BI tooling on top of unvalidated pipelines and then wonder why nobody trusts the output.
| Data Quality Failure Mode | Symptom | Fix |
|---|---|---|
| Conflicting numbers across systems | Leaders cite different figures in the same meeting | Canonical source definition, single source of truth per metric |
| Undefined metrics | "Conversion" means different things in product, marketing, and sales | Metric dictionary, enforced in BI layer |
| Pipeline failures undiscovered | Dashboard shows stale data silently | Data observability tooling (Monte Carlo, Elementary) |
| Model drift | Metrics no longer reflect business reality after product changes | ADR-style data model change process |
| No data tests | Transformations produce incorrect output silently | dbt tests on every critical model |
What Genuine Data Culture Looks Like
I've seen it work in two organisations at material scale: a German insurtech and a European energy trading operation. Both had specific properties in common.
Leaders in both organisations could cite — unprompted — a recent decision they made that the data reversed. Not a minor one. A meaningful commercial or product commitment that they personally believed in, that the data said was wrong, and that they changed based on the data. And they were right to change it. They had the outcome record to prove it.
The practice of committing to a threshold before pulling data was enforced at both. Not universally — humans are human — but consistently enough that it shaped the culture. When a leader started a meeting with "the data shows," someone would ask "what did you expect to see before you pulled it?" That question changed the conversation.
Both had invested heavily in data quality before investing in BI tooling. Not the other way around. They had data contracts between producing and consuming systems. They had tests on their transformation layer. They had a defined owner for every metric in the business.
Neither of them described their culture as "data-driven." Both described it as "evidence-based." That distinction is not semantic — evidence implies something that can challenge a prior belief. Data is just information. Evidence is information with epistemic power.
If your organisation has the dashboards but not the decisions — and you want to close that gap without another round of tooling investment — let's talk. Book a 30-minute discovery call and I'll tell you directly where the intervention needs to happen.