"Central data warehouse" sounds like a corporate matter. What it is is thoroughly ordinary: one place that collects data from several systems, makes it consistent, and serves reports from there instead of from each system directly. A 30-person business can need one; a large group may not, if it runs on a single system.
Three situations where it is genuinely needed
1. Data lives in three or more systems
Sales in one package, accounting in another, attendance in a third app, marketplace orders in a fourth. Working out the true profit on one order means opening four places. This is the most common case and the clearest.
2. You need history the source system overwrites
The sales system stores an item’s current price. Change it and the old price is gone. Analysing margin over time requires somewhere that keeps the state at each point — which is exactly what a warehouse does.
3. Heavy reports slow the live system
Running a two-year aggregate report against the operational database makes the till stutter. Moving analysis elsewhere solves both problems at once.
When you do NOT need one yet
- You have one system and its reporting is adequate — another layer is just another thing to break.
- The source data is still dirty. Collecting dirty data from three places into one produces centralised dirty data.
- Nobody has defined the indicators yet. Building a warehouse before knowing what to measure is building it for its own sake.
A minimal architecture for a small business
No expensive technology required. Three layers is enough, each with one clear job:
- 1Raw layer — copy the source data verbatim, changing nothing. Keep the original so that when something looks wrong you can trace it back.
- 2Conformed layer — unify customer codes, product codes, units and time zones; resolve duplicates. This is the heaviest layer and it decides the quality.
- 3Serving layer — tables pre-aggregated by day, by customer, by product, so the dashboard queries quickly.
How often to sync
The default answer should be "nightly", not "real time". Real-time syncing costs several times more and most management reporting does not need it. Only the indicators genuinely watched intraday — revenue, late orders — should read directly from the source.
Three common mistakes
- 1Build the warehouse first, ask what for later. The result is a warehouse full of tables no report uses. Start from 8–10 specific business questions and load only the data that answers them.
- 2No monitoring. A source stops sending for three days, nobody notices, and reports keep displaying normally on incomplete data. Every pipeline needs an alert for lateness and for abnormal record counts.
- 3Nobody owns the definitions. When accounting and sales mean different things by "revenue", the warehouse cannot fix it — it only makes the disagreement visible in more places.
If your business is in situation 1 or 3 above, the data and reporting solution area is where to start — usually one collection pipeline and a few aggregate tables cover most of the need.
Want to talk specifics?
SealCore surveys at your premises and sends a fixed quote after the first session — including when the conclusion is that you do not need custom software.
