Six weeks into her tenure, the newly appointed CDO of a European insurer presented a 47-slide data strategy to the executive committee. It was excellent. It mapped a target architecture, a governance operating model, three AI use cases, and a data-literacy program. The CFO nodded, approved a budget, and asked one question: *"What will be different by the time we close the books next quarter?"* She didn't have an answer. Eighteen months later, that CDO was gone—not because the strategy was wrong, but because it never metabolized into visible value fast enough to earn the political capital required to finish it.
This is the central tension of your first operating cycle. You already know *what* good data strategy contains. What separates CDOs who survive their third year from those who don't is the ability to compress vision into a roadmap that produces proof before patience runs out. This lesson is about that compression.
A data vision is not an aspiration ("become a data-drivendata-drivenAn approach where decisions are systematically informed by data analysis rather than intuition alone.Voir la définition complète → organization"). That phrase is dead on arrival—it's unfalsifiable, undifferentiated, and unownable. A usable vision does three things a slogan can't: it names the *business outcome* data unlocks, it implies what you will *not* do, and it survives repetition by a line manager who wasn't in the room.
The discipline here is subtraction. Your vision must pass the one-sentence test: if your Head of Claims can't restate it accurately to her team without your deck, it isn't a vision—it's decoration.
Compare two framings for the same insurer:
The second one tells you where the money is (underwriting, claims), what "better" means (speed, accuracy), and—crucially—who benefits (the decision-makers, not the data team). Everything you *don't* mention becomes a deliberate non-priority. That's the point.
Your vision must attach to a number the CEO is already accountable for. Before you write a word, obtain the current strategic plan and identify the two or three metrics the executive team is personally graded on—combined ratio, net revenue retentionnet revenue retentionNet Revenue Retention measures the percentage of recurring revenue retained and grown from existing customers over a period, including upsell and expansion, net of downgrades and churn.Voir la définition complète →, cost-to-serve, time-to-market. Your vision borrows the credibility of those metrics rather than inventing its own.
This is where most technically-strong CDOs fail. They anchor to *data* metrics (data qualitydata qualityThe degree to which data is fit for purpose: accurate, complete, consistent, timely, valid and unique. Poor quality data undermines analytics, reporting and AI.Voir la définition complète → scores, catalog coverage, platform uptime). Those are inputs. Nobody on the board lies awake over catalog coverage. When you present, the causal chain must run in one direction only: data capability → business decision → business metric the CEO already owns. If you can't draw that arrow, cut the initiative.
A roadmap is where vision meets the brutal reality of scarce engineers, dirty data, and organizational antibodies. The instinct to "boil the ocean"—to fix governance, rebuild architecture, launch AI, and evangelize culture simultaneously—is fatal because it produces no completed thing anyone can point to.
The standard 2x2 of *value vs. effort* is where you start, but senior practice demands a sharper instrument. Rank each candidate initiative on four axes:
1. Value — magnitude of impact on a CEO-scoreboard metric (not a data metric).
2. Feasibility — data readiness, technical maturity, and dependency depth.
3. Time-to-proof — how many weeks until *someone outside your team* sees a result.
4. Sponsorship — is there a named executive who will fight for this in a budget cut?
The fourth axis is the one that separates operators from architects. An initiative with high value and high feasibility but *no sponsor* is a trap: you'll build it, and it will sit unused because no business owner is on the hook for adoption. Kill or defer it, however painful.
Sequence the surviving initiatives into three horizons, but resist the temptation to make them equal-sized. Weight the near term heavily.
The sequencing logic is political as much as technical: Horizon 1 funds Horizon 2's permission. You are not building in order of architectural elegance; you are building in order of trust accumulation.
Here's a compact way to make the prioritization defensible in a room full of skeptics. Score, don't argue:
priority_score = (value × sponsorship) / (effort × time_to_proof_weeks)
# Example: automated regulatory report
# value=8, sponsorship=9, effort=3, time_to_proof=6
# score = (8 × 9) / (3 × 6) = 72 / 18 = 4.0
# Example: enterprise-wide data catalog rollout
# value=6, sponsorship=4, effort=9, time_to_proof=40
# score = (6 × 4) / (9 × 40) = 24 / 360 = 0.07The point of the formula isn't false precision—it's forcing every stakeholder who wants their pet project prioritized to argue in the same currency. When someone insists their initiative jump the queue, you don't debate opinions; you ask which input number they're contesting and why. The conversation shifts from politics to evidence.
Prioritization is 20% choosing what to do and 80% surviving the fallout of what you declined. When you defer a senior leader's request, never say "no." Say: *"That's Horizon 2. Here's the specific Horizon 1 win that has to land first to make yours succeed—and here's how you can help it land faster."* This reframes rejection as sequencing and converts a would-be opponent into a stakeholder in your near-term win. Every "not yet" should come with a condition under which it becomes "yes."
Keep a visible parking lot—a maintained list of deferred initiatives with their scores. It signals that ideas weren't ignored, they were assessed, and it gives you a ready backlog when Horizon 1 delivers and appetite grows.
Vérification des acquis
1. According to the lesson, why did the European insurer's CDO ultimately lose her role despite presenting an excellent 47-slide strategy?
2. The lesson describes the CDO's first operating cycle as a 'central tension.' What best captures that tension?
3. Why does the lesson reject 'become a data-driven organization' as a valid data vision?
4. Select ALL correct answers. According to the lesson, what does a usable data vision accomplish that a slogan cannot?
Sélectionnez toutes les réponses correctes.
5. Select ALL correct answers. What makes the 'strong' insurer vision superior to the 'weak' one, per the lesson's reasoning?
Sélectionnez toutes les réponses correctes.
A roadmap that isn't wired into an operating rhythm is a wish list. The transition from a plan on a slide to a plan in motion happens through three mechanisms: a delivery cadence, value contracts, and a stop-doing discipline.
For every Horizon 1 and Horizon 2 initiative, write a one-page value contract co-signed by you and the business sponsor before a single engineer is assigned. It states, in the sponsor's language:
The signature matters more than the content. It converts a data initiative from "IT's project" into a *joint* commitment where the business owns adoption. The most common cause of stranded data investment isn't bad technology—it's a built capability nobody changed their behavior to use. The value contract makes that behavior change an explicit, owned deliverable, not an afterthought.
Establish a layered review rhythm so problems surface in weeks, not quarters:
The quarterly re-scoring is what keeps the roadmap alive. A roadmap fixed for twelve months is a liability; the business will shift, a use case will prove unexpectedly hard, a new regulation will land. Re-scoring turns the roadmap from a promise you'll be judged against into an instrument you actively steer. Announce this cadence up front so that changing the plan reads as *disciplined adaptation*, not failure.
Every quarter, name at least one thing the data function will *stop* doing—a legacy report nobody reads, a shadow data pull you've been quietly maintaining, an initiative that lost its sponsor. Two things happen. First, you free capacity, which is almost always your true constraint. Second, you model the ruthlessness you're demanding of the organization. A CDO who only ever *adds* to the workload loses the moral authority to prioritize. Publish the stop-doing list alongside the roadmap; it's often more credible evidence of judgment than the start-doing list.
You will be tempted to report on outputs—pipelines built, models deployed, datasets governed. Resist. Build a single, standing value dashboard that tracks each initiative's contribution to its target business metric, refreshed monthly, in front of the executive committee. When the CFO asks "what's different this quarter," the answer is a line on a chart tied to a number they already care about, with the sponsor's name next to it. That dashboard is the single most important political artifact you will maintain—it is the running proof that data is a value center, not a cost center, and it renews your license to fund Horizons 2 and 3.