The setup will sound familiar: every month, three people spent the better part of a week pulling exports, reconciling mismatched columns and pasting charts into a deck. By the time leadership saw the numbers, they were two weeks old.
What we built
- Automated extracts from the three source systems, on a schedule.
- A reconciliation layer that flags mismatches instead of hiding them.
- A generated report — same template the team already used, now assembled in minutes.
How the five weeks went
- Weeks one and two: scoped pilot on a single report, real data, kill switch included.
- Weeks three and four: reconciliation layer, scheduling and alerting.
- Week five: handover, documentation and a run-it-yourself session with the team.
What changed
The cycle went from days to minutes, but the more interesting effect was trust: when numbers are fresh and reproducible, people stop debating the data and start debating the decision. The team now runs the report weekly — something that was unthinkable when it cost three days.
Fresh numbers ended the debate about the data — and started the one about the decision.
