Want smarter insights in your inbox? Sign up for our weekly newsletters to get only what matters to enterprise AI, data, and security leaders. Subscribe Now
In the past decade, companies have spent billions on data infrastructure. Petabyte-scale warehouses. Real-time pipelines. Machine learning (ML) platforms.
And yet — ask your operations lead why churn increased last week, and you’ll likely get three conflicting dashboards. Ask finance to reconcile performance across attribution systems, and you’ll hear, “It depends on who you ask.”
In a world drowning in dashboards, one truth keeps surfacing: Data isn’t the problem — product thinking is.
The quiet collapse of “data-as-a-service”
For years, data teams operated like internal consultancies — reactive, ticket-based, hero-driven. This “data-as-a-service” (DaaS) model was fine when data requests were small and stakes were low. But as companies became “data-driven,” this model fractured under the weight of its own success.
Take Airbnb. Before the launch of its metrics platform, product, finance and ops teams pulled their own versions of metrics like:
- Nights booked
- Active user
- Available listing
Even simple KPIs varied by filters, sources and who was asking. In leadership reviews, different teams presented different numbers — resulting in arguments over whose metric was “correct” rather than what action to take.
These aren’t technology failures. They’re product failures.
The consequences
- Data distrust: Analysts are second-guessed. Dashboards are abandoned.
- Human routers: Data scientists spend more time explaining discrepancies than generating insights.
- Redundant pipelines: Engineers rebuild similar datasets across teams.
- Decision drag: Leaders delay or ignore action due to inconsistent inputs.
Because data trust is a product problem, not a technical one
Most data leaders think they have a data quality issue. But look closer, and you’ll find a data trust issue:
- Your experimentation platform says a feature hurts retention — but product leaders don’t believe it.
- Ops sees a dashboard that contradicts their lived experience.
- Two teams use the same metric name, but different logic.
The pipelines are working. The SQL is sound. But no one trusts the outputs.
This is a product failure, not an engineering one. Because the systems weren’t designed for usability, interpretability or decision-making.
Enter: The data product manager
A new role has emerged across top companies — the data product manager (DPM). Unlike generalist PMs, DPMs operate across brittle, invisible, cross-functional terrain. Their job isn’t to ship dashboards. It’s to ensure the right people have the right insight at the right time to make a decision.
But DPMs don’t stop at piping data into dashboards or curating tables. The best ones go further: They ask, “Is this actually helping someone do their job better?” They define success not in terms of outputs, but outcomes. Not “Was this shipped?” but “Did this materially improve someone’s workflow or decision quality?”
In practice, this means:
- Don’t just define users; observe them. Ask how they believe the product works. Sit beside them. Your job isn’t to ship a dataset — it’s to make your customer more effective. That means deeply understanding how the product fits into the real-world context of their work.
- Own canonical metrics and treat them like APIs — versioned, documented, governed — and ensure they’re tied to consequential decisions like $10 million budget unlocks or go/no-go product launches.
- Build internal interfaces — like feature stores and clean room APIs — not as infrastructure, but as real products with contracts, SLAs, users and feedback loops.
- Say no to projects that feel sophisticated but don’t matter. A data pipeline that no team uses is technical debt, not progress.
- Design for durability. Many data products fail not from bad modeling, but from brittle systems: undocumented logic, flaky pipelines, shadow ownership. Build with the assumption that your future self — or your replacement — will thank you.
- Solve horizontally. Unlike domain-specific PMs, DPMs must constantly zoom out. One team’s lifetime value (LTV) logic is another team’s budget input. A seemingly minor metric update can have second-order consequences across marketing, finance and operations. Stewarding that complexity is the job.
At companies, DPMs are quietly redefining how internal data systems are built, governed and adopted. They aren’t there to clean data. They’re there to make organizations believe in it again.
Why it took so long
For years, we mistook activity for progress. Data engineers built pipelines. Scientists built models. Analysts built dashboards. But no one asked: “Will this insight actually change a business decision?” Or worse: We asked, but no one owned the answer.
Because executive decisions are now data-mediated
In today’s enterprise, nearly every major decision — budget shifts, new launches, org restructures — passes through a data layer first. But these layers are often unowned:
- The metric version used last quarter has changed — but no one knows when or why.
- Experimentation logic differs across teams.
- Attribution models contradict each other, each with plausible logic.
DPMs don’t own the decision — they own the interface that makes the decision legible.
DPMs ensure that metrics are interpretable, assumptions are transparent and tools are aligned to real workflows. Without them, decision paralysis becomes the norm.
Why this role will accelerate in the AI era
AI won’t replace DPMs. It will make them essential:
- 80% of AI project effort still goes to data readiness (Forrester).
- As large language models (LLMs) scale, the cost of garbage inputs compounds. AI doesn’t fix bad data — it amplifies it.
- Regulatory pressure (the EU AI Act, the California Consumer Privacy Act) is pushing orgs to treat internal data systems with product rigor.
DPMs are not traffic coordinators. They’re the architects of trust, interpretability, and responsible AI foundations.
So what now?
If you’re a CPO, CTO or head of data, ask:
- Who owns the data systems that power our biggest decisions?
- Are our internal APIs and metrics versioned, discoverable and governed?
- Do we know which data products are adopted — and which are quietly undermining trust?
If you can’t answer clearly, you don’t need more dashboards.
You need a data product manager.
Seojoon Oh is a data product manager at Uber.
Daily insights on business use cases with VB Daily
If you want to impress your boss, VB Daily has you covered. We give you the inside scoop on what companies are doing with generative AI, from regulatory shifts to practical deployments, so you can share insights for maximum ROI.
Thanks for subscribing. Check out more VB newsletters here.
An error occured.
