# Driving Cultural Change to Data-DrivenData-DrivenAn approach where decisions are systematically informed by data analysis rather than intuition alone.View full definition →
In 2016, GE poured billions into Predix and a 30,000-person "digital industrial" transformation. The platform worked. The dashboards were beautiful. The data lakedata lakeA data lake is a centralized repository that stores large volumes of raw data in its native format, from structured tables to unstructured files, until needed.View full definition → filled. And yet, five years later, most of the analytics never changed a single operational decision that a plant manager made under pressure at 2 a.m. The tooling shipped. The culture didn't move. GE eventually sold off the digital unit's ambitions at a fraction of the investment.
This lesson is about the levers you actually pull to close that gap.
Being data-drivendata-drivenAn approach where decisions are systematically informed by data analysis rather than intuition alone.View full definition → is a competition between two decision-making operating systems. The incumbent OS — intuition, seniority, precedent, political capital — is fast, socially rewarded, and deeply embedded. The challenger OS — evidence, experimentation, willingness to be proven wrong — is slower to feel comfortable and threatens the people who built careers on being right by instinct.
When you frame the shift as "rolling out Tableau" or "standing up a feature storefeature storeA centralised repository managing ML features, ensuring consistency between training and serving environments.View full definition →," you are answering a question nobody's incumbent OS was asking. Adoption stalls not because the tool is bad but because using it imposes a personal cost — cognitive effort, loss of status, exposure to being wrong — on individuals who see no offsetting reward.
This is why the classic capability vs. behavior distinction matters more than any maturity model. Capability is what the organization *can* do. Behavior is what it *actually does* under time pressure. Most CDOs over-invest in capability and under-invest in the behavioral system that determines whether capability gets used. You can measure the imbalance directly:
If your capability metrics are climbing and your behavior metrics are flat, you have a change-management problem wearing a technology costume. The John Kotter insight your fundamentals covered applies here with a sharp twist: the "urgency" you must create is not urgency to adopt tools, it's urgency to change how decisions get made. Those are not the same message, and conflating them is the single most common CDO failure.
There is also a structural reason culture resists. Analytics redistributes *interpretive authority*. Before, the VPVPA clear statement of the benefits your product delivers, the problems it solves and why customers should choose you over alternatives.View full definition → of Sales owned the narrative of why the quarter went the way it did. A clean cohort analysiscohort analysisCohort analysis groups users by a shared starting trait or time (such as signup month) and tracks their behavior over time to reveal retention and lifecycle patterns.View full definition → can now contradict him in front of his boss. You are not introducing a dashboard; you are relocating power. Treat every data-drivendata-drivenAn approach where decisions are systematically informed by data analysis rather than intuition alone.View full definition → initiative as a political act, because to the people inside it, that is exactly what it is.
The most expensive mistake a CDO makes is believing the CEO's *sponsorship* is the same as the CEO's *behavior*. Sponsorship buys budget. Behavior changes culture. An executive who funds the data program but still opens every review with "my sense is…" has just told 400 people what actually gets rewarded here.
Your job is to engineer specific, visible, repeatable executive behaviors — and to treat those behaviors as the deliverable, not a soft "by the way."
Concretely, negotiate these into existence:
1. The evidence-first meeting ritual. Rewire the standard business review so the *data* is presented before the *interpretation*, and the senior person speaks *last*, not first. This is the single highest-leverage behavioral change available to you. Amazon's six-page narrative memo, read in silence at the start of a meeting, is famous precisely because it forces the argument to stand on evidence before hierarchy weighs in. You don't need the memo format; you need the principle: seniority does not get to anchor the room before the numbers do.
2. Public mind-changing. Ask your sponsor to change a real decision, out loud, because of data — and to name that that's what happened. "I came in convinced we should kill this SKU. The retention data changed my mind." One such moment from a respected leader does more than a year of your evangelism, because it makes being-wrong-then-corrected *high status* instead of embarrassing.
3. The disconfirmation question. Coach leaders to ask, in reviews, "What would have to be true for this to be wrong, and did we look?" This normalizes the challenger OS at the top of the org, where behavior propagates downward fastest.
The CDO's role here is closer to a chief of staff than a technologist. You script the rituals, you prep the sponsor, you supply the disconfirming evidence, and you make the executive look smart for using it. If you cannot get *behavioral* commitments — not just verbal support — from the top two layers, do not launch a broad culture push. Concentrate your energy on winning those commitments first, because everything downstream depends on them.
Culture is the aggregate of what people believe will help or hurt their careers. If your incentive structure rewards confident assertion and punishes visible uncertainty, no amount of training will produce evidence-based behavior. People are rational; they optimize for the reward they can see.
Start by auditing the implicit incentives, which are far stronger than the ones in the HR handbook:
The fix is to make the challenger OS the profitable one. Three practical moves:
Reward the decision quality, not the outcome. Tie recognition to whether a decision was made with sound evidence and clear reasoning, separately from whether it worked out. This is critical because if you only reward wins, people stop taking evidence-based bets that carry visible risk and retreat to defensible gut calls. Google's postmortem culture — blameless, focused on the system not the person — is the mechanism that makes it safe to surface what the data actually showed.
Instrument the behavior you want and put it in the review. You can only reward what you can see. Lightweight decision logs turn invisible behavior into a managed metric:
decision_record:
decision: "Cut paid-search budget in EMEA by 30%"
owner: "VP Marketing"
date: 2024-11-04
evidence_used: ["incrementality_test_Q3", "MMM_v4"]
assumption_to_revisit: "channel saturation holds below current spend"
review_date: 2025-02-04
outcome: null # filled at review; drives learning, not blameThis is not bureaucracy for its own sake — it is the artifact that lets you *see* whether decisions are getting more evidence-based over time, and it makes the behavior legible to the people writing performance reviews.
Move the analysts to the decision, not the dashboard. Reward your data team for decisions influenced, not assets produced. A data scientist whose bonus depends on model accuracy will build accurate models nobody uses. One whose recognition depends on a business decision that changed will fight to get into the room where it's made. Change the incentive, change the posture.
Culture change dies in the gap between vision and proof. You announced the transformation; twelve months later the org is looking for evidence it was real. Quick wins are how you buy the political capital and belief to keep going — but most CDOs pick them badly.
The instinct is to pick the *easiest* win. Wrong criterion. Pick the win that is visible, credible, and attributable — a decision that a respected skeptic will acknowledge got better *because of the data*, in an area they care about. A backend data-quality improvement that saves the pipelinepipelineAll active sales opportunities across the stages of the sales process, together with their combined potential value and probability of closing.View full definition → team three hours is real value and utterly useless as a culture lever, because no one who matters will feel it.
Use a simple selection filter for your first three initiatives:
Then *manufacture the narrative*. A quick win that no one hears about did not happen, culturally. When a regional manager beats forecast using a churn model, that becomes the story you retell in every town hall, every board deck, every onboarding. You are not exaggerating; you are ensuring the win does the propaganda work it earned. Behavioral contagion in organizations runs on stories, not statistics — one vivid, named example beats ten aggregate charts.
Sequence deliberately: win small and visible → publicize → win adjacent → connect the wins into a trend line. The goal is to reachreachThe number of unique people exposed to your message in a given period. Unlike impressions, reach counts each person once, no matter how often they see it.View full definition → the point where using data is the *default expected* behavior and the person still running on pure gut feels the social pressure, rather than the analyst feeling it. That inversion — when the burden of proof shifts to the person *not* using evidence — is the moment culture has actually turned.
Knowledge check
1. The GE Predix example is used in the lesson primarily to illustrate which principle?
2. According to the lesson, why does framing the shift as 'rolling out Tableau' or 'standing up a feature store' cause adoption to stall?
3. How does the lesson distinguish 'capability' from 'behavior', and why does the distinction matter to a CDO?
4. Select ALL correct answers. According to the lesson, what makes the 'incumbent OS' of intuition, seniority, and precedent so resistant to displacement?
Select all the correct answers.
5. Select ALL correct answers. Which statements accurately reflect the lesson's argument about the gap between 'we have data' and 'we decide with data'?
Select all the correct answers.
The three levers are not a menu; they are a system, and pulling them in the wrong order wastes them. Leadership behavior comes first because it sets the reward gradient that makes everything else rational. Incentives come second because they sustain the behavior after your personal attention moves elsewhere. Quick wins come third — or rather, run continuously underneath — because they supply the belief that keeps leaders and incentive-owners committed.
A field diagnostic you can run this week: pick your three most recent significant decisions. For each, trace who spoke first, whether evidence was present before opinion, whether anyone asked what would disconfirm the call, and whether the person who used data was rewarded or exposed. You will learn more from that trace than from any culture survey, because it measures behavior under real conditions rather than stated values.
Watch for the two failure modes. The first is theater — dashboards on every wall, "data-drivendata-drivenAn approach where decisions are systematically informed by data analysis rather than intuition alone.View full definition →" in every deck, and decisions still made in the hallway afterward. The tell is a widening gap between capability metrics and behavior metrics. The second is backlash — you moved too fast, threatened too much interpretive authority too publicly, and the incumbent OS organized against you. The tell is quiet non-adoption from a powerful cohort. The counter to both is the same: return to leadership behavior, secure genuine commitment, and re-sequence your wins around the skeptics you most need.