Data-driven is a discipline, not a destination

“Data-driven” is on the values page of almost every company I have worked with or talked to, and I have rarely seen anyone define it. In practice it means “we have dashboards,” or “we hired a data team,” or “the CEO mentions numbers in the all-hands.” None of those is the thing. They are evidence that data exists, not that it is driving anything.

This post does three things. It gives the phrase a definition you can test in a meeting. It offers a ladder for working out honestly where an organization stands. And it connects those habits to the technical investments meant to support them.

A definition you can test

My working definition: an organization uses data consistently, across the levels where decisions are made, to test its assumptions and inform what it does next.

Three conditions. Each one rules out something commonly mistaken for the real thing.

Consistently rules out the quarterly set piece. Data that appears when the board meets and vanishes in between is a presentation habit, not a decision habit.

Across decision-making levels rules out the executive dashboard as proof of culture. If leadership sees numbers and the people running the funnel, the queue or the campaign do not, decisions rest on data at exactly one altitude, and it is the altitude with the least contact with the customer.

To inform decisions includes changing direction, confirming a justified choice, or identifying uncertainty worth investigating. Requiring a reversal every time would reward visible changes over sound judgment. The test is whether evidence could change the decision and whether the team can explain its use.

If the room cannot answer, investigate the decision process before assuming a new platform will fix it.

The ladder

I describe data culture as five rungs. Versions of this ladder appear in every maturity model going, and I make no claim to originality about the shape. The value is in describing each rung honestly enough that people recognize their own. This is a discussion aid, not a validated assessment or a score for the whole company. Functions can occupy different rungs and move backwards under pressure.

1. 1Ad hocData gets mentioned. Leadership brings it up occasionally, usually when a number is good.
2. 2AwareEveryone agrees data is important. It gets airtime, a team, maybe a tool. No consistent use, no direction.
3. 3DevelopingSome teams use data in some decisions. It depends on which individuals are in the room.
4. 4ManagedData informs most decisions. Leadership invests in tools and, more tellingly, in definitions, training and time.
5. 5TransformativeEvidence informs strategy and routine work. Teams know when data is useful, when it is insufficient, and how to learn.

The pattern I have seen repeatedly is that organizations place themselves one rung, sometimes two, above where their meetings say they are. The self-assessment is done by looking at the tooling. The honest assessment is done by looking at what happens when a number and an opinion disagree.

Four questions help make the discussion concrete:

1. When a metric moves, does someone ask what changed before someone proposes what to do?
2. Do people bring data to a meeting unprompted, or only when asked to justify a position they already hold?
3. Does the same metric name mean the same thing in two departments’ decks?
4. When the data contradicts a senior person’s instinct, what happens next, and does it happen the same way each time?

None of these has anything to do with the warehouse.

Built in one order, bought in the other

Culture, strategy, and technology develop together. What matters is that the purchase does not substitute for deciding which behavior should change and why. A small technical improvement can earn support for wider change; a new decision habit can expose a missing technical capability.

Start

1. Decision
2. Evidence needed
3. Small intervention

Review

1. What changed?
2. What failed?
3. What next?

Technology is easier to purchase than a new habit is to establish. That makes it tempting to treat a launch as proof of a capability. A more useful milestone is whether a named team can make a decision better with the capability than without it.

Self-service analytics is the canonical case. For twenty years it was sold as an access problem: give people the tool and they will answer their own questions. Mostly they did not. I have written elsewhere about why the constraint underneath was trust rather than interface. There is a second reason, which is that nobody taught the habits. The seat was provisioned. The discipline was assumed.

The AI wave is repeating this at higher speed. A chat assistant that answers questions in plain language is the most purchasable data capability ever built, and it is landing in organizations that have not settled what their metrics mean. The interface gets solved and the culture gap becomes visible faster, which is at least an improvement on dashboards, which hid it for years.

Four moves that shift a rung

From the leadership seat, the moves that have actually shifted an organization upward are unglamorous, and none of them is a procurement.

1. Ask for the number before the opinion. When leaders open with “what does the data say” rather than closing with it, the meeting reorganizes around evidence. This costs nothing and is the fastest culture change available, because people prepare for the question they expect to be asked.

2. Agree the scope and owner of each shared metric. Two conversion rates may measure different populations or stages. Make those distinctions explicit, and agree which measure applies to the decision in front of the team. Governance should resolve ambiguity while preserving legitimate differences.

3. Make the decision the unit of work. A request for a dashboard is a request for a chart. A request framed as “we need to decide whether to keep this channel” is a request for an answer, and it has a finish line. A data team that insists on the second framing does less work and changes more decisions. I have made the same argument about dashboards, where it applies with more force.

4. Put context and data in the same conversation. The business side knows what changed, when, and why it might matter. The data side knows what the numbers can and cannot support. Neither can do the analysis alone, and most bad conclusions come from one side working without the other. The business brings the half nobody can automate.

What it costs

It needs sustained attention. Changing how meetings work takes repetition, but progress can be visible before an entire culture changes. Track a few recurring decisions, the evidence used, and whether the agreed follow-up happened.

It asks leaders to make assumptions inspectable. Data can be incomplete, biased, or poorly matched to the decision. When judgment departs from the available analysis, state why and what would change the view. When the analysis is persuasive, be willing to revise the original position.

The data team gives something up too. Moving from the third rung to the fourth means the central team stops being the only voice that can speak about the numbers. That is the point, and it still stings.

A mirror, not a badge

The useful question is whether evidence has an honest chance to affect a consequential decision. A company can have a sophisticated platform and fail that test. A small team with modest tooling can pass it.

It is a discipline, and like every discipline it lapses the moment you stop practising it. The ladder is useful as a mirror. It is useless as a badge.