The problem
The food bank runs on numbers: pounds distributed, meals served, agencies supported, grants reported. Those numbers came from several systems, and each system had its own quiet definition of what a “meal” or a “distribution” was. When two reports disagreed, no one could say which was right, because there was no agreed source of truth to check against.
For a nonprofit that answers to funders and a board, an unexplained discrepancy is not a rounding error. It is a credibility problem.
The approach
This engagement did not start with a model. It started with a data audit, because the problem was governance, not intelligence. We inventoried every system that produced a reportable number and, working with the food bank’s team, wrote down a single owned definition for each metric that mattered: what it counts, where it comes from, and who is accountable for it.
With the dictionary in place, we built a validation layer that reconciled the source systems against those definitions and flagged where they diverged. Then we built reporting views that drew exclusively from governed numbers, so a grant report and a board report told the same story because they were reading from the same source.
We resisted the temptation to add analytics on top before the foundation was trustworthy. A model built on numbers no one trusts only launders the confusion.
The outcome
The food bank now reports from one definition of every figure it publishes. When a number moves, the team can trace why, and when a funder asks how a metric is calculated, there is a documented answer. The dictionary is a living asset, maintained as new programs and data sources come online.
It is unglamorous work. It is also the work that made everything downstream possible.
