As a mid-level Air Force officer at the Air Force Personnel Center, I was leading one of the service’s bi-annual logistics career field talent management conferences—responsible for two dozen senior leaders reviewing 500 to 700 officer records for future leadership and professional development opportunities. The data retrieval process was cumbersome at best. The three “authoritative” systems didn’t talk to each other, couldn’t be easily corrected and weren’t transparent to the people relying on them resulting in a Picasso-like mosaic of information.
The conference progressed, however my confidence unraveled as senior leaders poked holes in the data. The records were inconsistent, outdated and inaccurate in places. My control over the situation was slipping as the group’s patience waned and the temperature in the room rose. The conference was devolving into a tense argument over whose data was right, instead of doing what we were actually there to do: identify the best logistics leaders for advancement.
The problem wasn’t the effort my team was putting in; the system we were working within had set us up for failure. Data discipline means treating reliable information as an operational asset, not a reporting by-product, so leaders can make decisions from a shared version of reality. A CEO doesn’t need to worry about officer records, but they absolutely need to worry about the equivalent—the ERP migration where finance, sales and operations each define “revenue” slightly differently; the acquisition where two companies’ customer databases never quite reconcile; the board deck built on numbers three departments would each explain differently if asked.
Here’s the four-step operating model we eventually built to fix it; one any leader can apply well before their own version of that room gets tense.
1. Assign ownership.
Don’t start with technology, start with a name. For every critical data set (customer records, revenue definitions, inventory counts), name one person accountable for its accuracy, not just its custody.
If two departments claim ownership, force a decision.
2. Define what “right” means.
Get the people who touch a metric in a room and force them to agree, in writing, on one definition. A board deck built on multiple definitions, is that Picasso-like mosaic waiting to happen.
Schedule a 60-minute working session to finalize definitions; don’t let the session devolve into “admiring” the problem.
3. Measure data quality and post the score.
Track error rates, missing fields and how often a number gets corrected after it’s published. Then put that scorecard in front of leadership, not just the data team.
Pick three metrics and publish them monthly. Leaders won’t change what they can’t see.
4. Connect the systems that matter to the decision.
Don’t chase full integration. It’s slow, expensive and usually unnecessary. Map the two or three systems that feed the decision in front of you, and connect those first.
Which two systems drive the decision in front of us? Connect only those.
Revisiting my talent conference failure—the exact situation I vowed never to repeat—we put this model in place prior to our next conference. An agreed upon single database owned by the Air Force Personnel Center (Step 1). Standardized definitions for each performance measure (Step 2). Appointed a single point of contact for data quality (Step 3). Synchronized information across only the stakeholders and systems needed (Step 4). The result: data preparation time fell from two days to two hours, transcription errors dropped by 90 percent and the team’s focus returned to identifying future leaders.





