From Spreadsheets to a Warehouse People Trust

Summary

A food distributor ran the business on spreadsheets: sales kept one, finance kept another, operations a third — five versions of every number and a monthly argument about which was right. We built one warehouse with nightly pipelines and lineage anyone could follow, working alongside their analyst so the knowledge stayed in the building.

The Challenge

The Problem

Every planning meeting started with reconciling numbers instead of using them:

• Five versions of truth: Each department's spreadsheet told a different story about the same month.
• Manual everything: Copy, paste, email, repeat — hours every week, errors every month.
• AI plans on hold: Leadership wanted forecasting, but no model survives data nobody trusts.

Root Cause Analysis

There was no owned, automatic path from source systems to decisions. Each team built its own because nobody owned the shared one. The fix was ownership and automation, not a bigger spreadsheet.

Our Approach

Phase 1: Profile the Sources (Week 1)

• Shadowed the analyst through one full monthly close.
• Documented every source system: row counts, freshness, null rates, quirks.
• Named an owner for each source table — a person, not a department.

Phase 2: Build the Warehouse (Weeks 2–5)

• Stood up a warehouse in the client's own cloud account.
• Built nightly pipelines from order, inventory, and accounting systems.
• Published one sales dashboard matching the reconciled numbers both teams signed off.

Phase 3: Handover (Week 6)

• Documented lineage end to end: every dashboard number traceable to its rows.
• Trained the analyst to extend pipelines and onboard new sources.
• Archived the competing spreadsheets in a small ceremony nobody mourned.

Key Achievements

Quantitative Results

• One dashboard both teams now plan from — the monthly reconciliation meeting no longer exists.
• Hours returned weekly to analysis instead of copy-paste.
• Six weeks from kickoff to handover.

Qualitative Benefits

• Arguments about numbers replaced by arguments about decisions — the upgrade that matters.
• Lineage documentation means new hires trust the data in days, not quarters.
• The forecasting project they wanted is now possible; the data is ready when they are.

Lessons Learned

1. Shadow before schema: Watching one full cycle beats a month of requirements meetings.
2. Owners, not committees: Every table needs one person who cares if it is right.
3. Build beside, not around: Their analyst co-built everything, so nothing left with us.
4. Boring wins: Nightly pipelines and documentation beat clever architecture.

Conclusion

Nobody writes poems about a warehouse, but everything the business wants next — forecasting, assistants, real planning — stands on it. Six weeks of unglamorous work bought them years of trustworthy numbers. That is the trade, and it is always worth making.

---

Client: A food distributor Duration: 6 weeks Team Size: 2 consultants Outcome: One trusted warehouse, owned and run by the client