Most campus data projects begin with a platform decision: which warehouse, which lakehouse, which BI tool. A year later the platform is live, and the same arguments about the numbers are still happening in cabinet meetings.
The missing step is older than any of these tools. Peter Drucker asked five questions of every organization that wanted to be effective. They apply just as well to a data program, and we start every engagement with them.
1. What is the mission?
A college does not exist to run reports. It exists to deliver public value: completion, transfer, workforce readiness, economic mobility, and equitable outcomes, using public money responsibly. A data program should be able to say which of those it helps the institution measure and improve. If it can't, it is an IT project, not a mission investment.
2. Who is the customer?
Usually not one group. IT owns the systems, institutional research owns the definitions, and the operational offices (financial aid, the registrar, HR, enrollment, finance) own the business rules. Leadership and the board are customers too, and students are the ultimate beneficiaries. Name the primary customers explicitly, because their needs will compete.
3. What does the customer value?
Ask them. In our experience the answers are consistent: numbers that are accurate and repeatable, definitions everyone shares, operational reports that arrive on time (most of the demand), a transition that doesn't break daily work, and the skills to serve themselves. Advanced analytics and AI come up too, but rarely first.
4. What are the results?
"Results are on the outside." A deployed platform is a cost, not a result. Agree before you build what would be visibly different: fewer disputes over numbers in meetings, less time reconciling data, a smaller report inventory, external reporting that can be reproduced, leaders asking for insights instead of extracts. Define indicators at the strategic, operational, technical, governance, and cultural levels.
5. What is the plan?
Only now choose the architecture, and choose it to fit the strengths you already have: your staff's skills, your existing systems, your governance capacity. Then phase it. Start where the value is highest, often one operational office at a time, keep reporting running throughout, and build skills as you go.
Why this matters for AI
Drucker wrote that information is data endowed with relevance and purpose. AI makes that sentence urgent. A model trained or prompted on data without shared definitions and context will produce confident answers to the wrong question. The five questions are how an institution supplies the relevance and purpose first.
Where to start
Put IT, research and planning, and the operational offices at the same table and answer the five questions together. If the answers are hard to agree on, that is the most valuable finding of all.
