Agencies GAO reviewed had stopped using OPM’s dashboard, including OPM.
How stale its data got before anyone flagged it.
Every federal program manager has a dashboard like this somewhere in their portfolio: expensive, well-intentioned, and at some point people, stopped being opened. In March 2026, a Government Accountability Office review put a number on just how common that is. This time, the agency that stopped using its own tool was the one that built it.
The tool was OPM’s Cyber Workforce Dashboard, launched government-wide in 2023 to help agencies plan cyber staffing. As FedScoop reported following GAO’s review, agencies cited recurring problems with access, functionality, and trust in the underlying data. When asked why it had stopped relying on its own dashboard, OPM did not have a clear answer.
I share this story to see this pattern constantly in federal data and analytics work, and it almost always traces back to the same root cause: the dashboard was built as a report instead of being managed as a product.
Report Thinking vs. Product Thinking
A report answers one question, gets delivered, and the team moves on to the next request. A data product is different. It has an owner, a defined set of users, and a lifecycle that continues well past launch day. It gets monitored, measured, and improved the same way any other mission system does.
| One-Off Report | VS | Data Product |
| Built to answer one question | Built to serve ongoing decisions | |
| Designed by the builder | Designed around user personas | |
| No defined owner after delivery | Has a designated product owner | |
| Metrics chosen by intuition | Metrics chosen via methodology | |
| Launched once, rarely updated | Follows a continuous lifecycle | |
| No data quality checks | Data quality is core requirement | |
| Buried in a folder no one finds | Published to a discoverable marketplace | |
| Success = delivery date | Success = decision frequency & user trust |
That distinction sounds simple, but it’s the difference between a tool people are still opening a year later and one that quietly gets bookmarked and forgotten. In our work, we look for a clear operating model before we’re willing to call something a data product.
The Four Pillars of Data Products
User Focused
Designed for and with the intended audience. Built around real decisions, not available data.
Data Quality
Accurate, consistent, and trusted data with visible quality ratings and ownership.
Scalability
Architecture that grows with usage and volume without losing performance or reliability.
Iterative
Monitored, measured, and continuously improved. Never ‘done’ — always getting better.
Bypassing that operating model, you will end up with a version of the OPM story: a capable team, real investment, and a tool nobody trusts enough to keep using.
Is Your Existing Dashboard a Report or a Product?
For program managers, the fastest way to diagnose the issue is to ask whether the dashboard has the basic operating model of a product. If these answers are undefined, usage problems are usually a symptom, not the root issue.
- Does the dashboard have a named product owner?
- Are user personas documented and tied to specific decisions?
- Is every major metric connected to an action, threshold, or operational question?
- Are data owners and refresh cycles visible to users?
- Is usage tracked after launch?
- Is there a backlog for improvements?
- Is there a governance cadence for data quality, access, and adoption?
- Can users tell what changed, why it matters, and what to do next?
Personas Drive the Architecture, Not Just the Layout
The fastest way to end up with a report instead of a product is to design for stakeholders in the abstract, rather than for the specific people who will open the tool, and the decisions they need to make. A cyber workforce dashboard serving 24 agencies, OMB, and multiple oversight bodies through one generic view was always going to struggle. Different users needed different things from it, and the design did not distinguish between them.
We start every engagement by defining the personas a dashboard serves. Usually that’s an executive who needs a 30-second read, a manager tracking exceptions daily, an analyst who needs full drill-down access, and an operator who needs one number and a threshold alert. Those personas don’t just shape what the screen looks like. They define the architecture itself, which is why we build almost everything around a consistent three-level model.
| 1 |
Scorecard View
The 30-second read
North Star metric, KPI tiles, traffic-light status. Designed so the most critical number for any user is visible in under 5 seconds.
|
| 1 |
Trend & Comparison View
The 5-minute analysis
Period-over-period trends, team breakdowns, benchmark comparisons. Available when Level 1 raises a flag that needs.
|
| 1 |
Drill-Down & Exploration
Deep investigation
Full filters, raw tables, multi-dimensional views, exports. Never the default — always on demand for root-cause investigation.
|
Every persona enters at the scorecard and stops as soon as their question is answered. Nobody gets handed a wall of filters just to find one number, and nobody has to leave the tool to understand why a metric turned red. Underneath that architecture sits a small set of governance habits we treat as non-negotiable: named data ownership, visible freshness indicators, and row-level access control. It is worth noting that GAO’s finding on OPM was as much a freshness and governance failure as a design failure. Workforce data sitting eleven months stale is exactly what happens when nobody owns the job of noticing.
That discipline matters even more in federal environments, where dashboard teams often depend on upstream systems they do not control, data ownership can be split across offices or agencies, and security rules determine who can see what. Reporting requirements, budget cycles, and compliance obligations can also reward launch over continuous improvement. Product thinking gives program managers a way to manage those realities instead of hoping the dashboard will stay useful on its own.
From Insight to Action: Why Storytelling Still Matters
Getting the architecture right solves half the problem. The other half is what happens on the screen itself. A well-governed, persona-driven dashboard can still fail to change anyone’s behavior if it just displays numbers and leaves the user to figure out what they mean. The dashboards that truly drive mission outcomes are the ones designed to tell a story, not just report a status.
We build every important chart or callout around a simple four-part arc: what’s normal, what changed, why it matters to this specific user, and what they should do about it. Skip the last two steps and you’ve built an expensive way to display a number. Include them, and that same chart becomes the reason someone reassigns a caseload, escalates a finding, or catches a compliance issue a week earlier than they would have otherwise.
Data alone doesn’t move anyone. A story does.
-
Setup
What’s the normal, expected state?
-
Insight
What changed, and by how much?
-
So What
Why does this matter to this persona?
-
Call to Action
What decision or action follows?
Rule of thumb: if you can’t finish the sentence “so the user should…,” the chart isn’t telling a story yet.
This is also where the real return on a data and analytics investment shows up. Agencies do not get mission value from a dashboard being technically accurate. They get it when the dashboard leads a real person to a real decision faster than they would have gotten to on their own. That is the bar to which we hold every product, and it is a different bar than most legacy reporting tools were ever designed to clear.
We Apply This Across the Federal Government
Across agencies such as IRS, FHFA, and CBP, we have used this approach to help program teams move from static reporting to governed, decision-oriented analytics products. That work often includes defining user personas, mapping dashboard views to operational decisions, improving data trust, and creating governance routines that keep products useful after launch.
That approach becomes most valuable when agencies are trying to move beyond recurring reports and create analytics products that change how leaders manage programs.
The missions and users are different every time, but the underlying discipline does not change: know who the dashboard is for, know what decision it needs to support, and make sure someone is accountable for whether it is still working six months from now.
What Program Managers Should Ask Before Funding the Next Dashboard
Before investing in a new dashboard or trying to rescue one that is underused, program managers should ask a few product-level questions:
- Who owns this dashboard after launch?
- Which decisions will it improve?
- Which users does it serve, and how do their needs differ?
- What data quality controls make the outputs trustworthy?
- How will we know if people are using it?
- What improvement cycle keeps it relevant?
Why This Matters More as AI Enters the Picture
Every agency we work with is now layering AI and advanced analytics on top of its existing reporting: automated case triage, predictive workforce and program planning, and natural-language access to agency data. None of that changes the underlying requirement. It only raises the stakes. Before agencies can scale AI, they need trusted data products that people already use, govern, and improve. Dashboards are often the proving ground for that discipline. An AI model or automated pipeline built on top of an unowned, ungoverned, or poorly adopted dashboard inherits those weaknesses, only faster and with less visibility into why. The agencies that get real mission value out of AI over the next few years will be the ones that already treat their data and reporting as products, not projects.
Justin Bell | 11/6/2026