Credits, DBUs, and comparing the incomparable

Teams running more than one platform eventually try to compare them, usually during a renewal. The comparison is almost always wrong, and it is wrong in a specific way: it compares the unit each vendor bills in as though the units measured the same thing.

The units are not units of work

A credit is a unit of allocated compute for a period. A DBU is a unit of processing on a cluster whose price varies by workload type and runtime. Neither is a unit of query, a unit of data or a unit of result. Converting between them with a single multiplier produces a number with a decimal point and no meaning.

What can be compared

• Cost to produce a specific table at a specific freshness, measured on both, on the same input.
• Cost to answer a representative query set, run enough times to see the cache behaviour.
• Cost of idle: what each platform bills for an hour in which nothing is asked of it.
• Cost of the second environment, which is where the pricing models diverge most sharply.

Note that every item on that list is an experiment rather than a calculation. That is the point. The only honest cross-platform number is one you measured on your own workload, and the only honest cross-platform statement is one that says which workload it was measured on.

Attribution differs too, and that matters more

On one platform a dedicated warehouse means every credit it burns is attributable to one workload. On another, billing is per cluster and several workloads share it, so a tool that wants to report what it itself costs you cannot separate its own queries from yours. This is not a detail for the finance team — it decides whether you will ever be able to answer "what did this cost" about anything running on that platform. Ask the question before you choose, not after.