Is a cloud warehouse cheaper than a server?
It is cheaper where load varies and more expensive where it is flat and predictable. Metered compute rewards bursts and punishes constant heavy querying.
Where the money in a data platform actually goes.
Warehouse pricing pages are easy to read and easy to misread. The storage line is rarely what decides the bill, and the largest cost usually is not software at all.
Four costs, and they are not equal
A data platform costs money in four places: storage, compute, the pipelines that load it, and the people who define and maintain the model. Storage is almost always the smallest and the one people focus on.
Compute is metered on most modern warehouses, which means the bill responds to how often things run and how carelessly the queries are written. A dashboard set to refresh every five minutes because somebody clicked the default can cost more than the storage for the entire history.
The choice that moves the number most
Not which product. How often data is loaded and how often it is queried. Hourly loading instead of nightly can multiply compute cost without changing a single business decision, because nobody looks at the figure until the morning.
Ask what decision each refresh supports. Most reporting is read once a day. Some operational data genuinely needs to be current within minutes, and that part can be built separately rather than paying real-time prices across the whole platform.
Products, and when each stops making sense
Postgres carries far more businesses than the market implies, and for tens of millions of rows with sensible indexing it is often the cheapest correct answer. It stops making sense when queries scan the whole history repeatedly, or when loading and reporting start competing for the same machine.
BigQuery, Snowflake and Databricks each price compute separately from storage and each suit particular shapes of workload. They earn their cost at scale and at variable load. Below that they are a subscription to capacity you are not using.
The cost nobody budgets
Definitions. Agreeing what revenue means, what a customer is, and which date a report measures against takes senior time, and it is the work that makes the platform trustworthy.
Skipping it produces a technically working warehouse that nobody believes, which is the most expensive outcome available, because the money is spent and the spreadsheets stay.
How to size it before committing
Take the largest three tables you would load, count the rows and estimate the growth. Take the reports you would build and count how often each genuinely needs to be current. Those two numbers produce a usable estimate on any of the platforms above.
Then run the first slice, one business area rather than the whole estate, and measure the real bill for a month before extending it.
What to establish before choosing a platform
Checklist
- Row counts and growth rate for the three largest tables
- How current each report genuinely needs to be
- Which reporting tool the business already knows
- Whether the spend needs to be predictable or can vary with use
- Who owns the metric definitions after launch
Questions we get asked.
How much does a first platform cost to build?
It depends almost entirely on how many sources are loaded and how clean they are. Scoping one business area first gives a real figure rather than a range, and it is the approach we recommend.
Can we start on Postgres and move later?
Yes, and it is a reasonable path if the pipelines and the model are built to be portable. What makes a move painful is business logic embedded in product-specific features.
Start with the problem.
Describe what is slowing the operation down. We will come back with the approach, the scope, and what it costs.
Get in touch