The first question most companies ask about a data warehouse is how much it will cost. The honest answer is that it depends on a handful of specific things, and you control more of them than you might think. Knowing what drives the cost makes proposals easier to compare and easier to challenge.
The number and type of sources
Each source system, whether an ERP, a CRM, a time-clock system or a spreadsheet, needs a connection, a load process and a mapping. Ten sources cost much more than three, and an old system with no clean interface costs more than one with a documented API. List your sources before you ask anyone for a quote.
The state of the data
Clean data loads easily. Messy data, with duplicate customers, inconsistent codes and free-text fields where there should be categories, needs cleaning and rules. This is often the largest hidden cost, and it is not caused by the warehouse. It is already there. We cover the pattern in why your ERP data is wrong.
How much history you need
Loading three years of history is a larger job than loading from today forward, and old data is usually in worse shape. Decide how much history reports genuinely need.
How fresh the data must be
A nightly refresh is straightforward. Hourly is more work. Near real time is a different architecture altogether. Be honest about what decisions need data that fresh. Many do not, and our note on getting real-time data without replacing your systems covers the cases that do.
The modeling effort
Turning raw tables into definitions people agree on, such as what counts as revenue or what counts as an active customer, takes meetings as well as code. It is time well spent, but it is time, and it depends on how many numbers you need defined.
Security and access rules
If everyone sees everything, it is simple. If different people must see different slices, the design and testing take longer.
Where it runs
Cloud, on-premises or a mix changes the cost profile, as discussed in Azure SQL versus on-premises SQL Server. The cheapest option on paper is not always the cheapest once you count administration.
What sits on top
A warehouse with no reports delivers nothing. The dashboards, exports and scheduled reports that people actually use are part of the cost, even if they are quoted separately.
Running it afterwards
A warehouse needs monitoring, fixes when a source changes, and new requests. Ask what the ongoing cost is, and who does it.
How to keep the cost under control
Start with one subject area that matters, such as sales or production, instead of everything at once. Define the handful of reports that justify it. Use refresh frequency that matches real decisions. Reuse infrastructure you already have where it is sensible. Avoid building a data lake you do not need, as covered in data warehouse versus data lake. And phase the work so each stage delivers something people use.
For a walk through the basics, see building a data warehouse from scratch.
If you are weighing a warehouse project and want help scoping it so the quotes are comparable, reach out.
Have a question about this topic? We work with businesses worldwide.
Get in Touch