Data
Data platforms
Snowflake, Databricks and BigQuery. We trace the bill back to the individual queries, schedules and settings producing it.
What we look at
Five things, each one measured.
- What each query, person and job costsSpend split down to individual queries and scheduled jobs, then ranked. Usually a handful of them account for most of the bill.
- Sized for the worst hourWhether your compute is sized for its normal workload or for one busy afternoon somebody still remembers.
- Paying while nothing runsSettings that keep compute awake when there is nothing to do. On most platforms this is the fastest thing to fix and the one most often missed.
- Running more often than anyone needsJobs refreshing every hour to feed a report somebody opens on Mondays.
- How tables are built, and what is keptLayouts that force a full scan where a smarter one would not, and history settings nobody ever chose on purpose.
What you get
A document, not a dashboard.
- A ranked list of findings, each with what it costs now, what it would cost after the change, roughly how many hours that takes and what could go wrong.
- The evidence behind every finding — the measure, the period, the source.
- A section naming what to leave alone, and why.
- Every read-only role we used, listed, so you can check afterwards.
- An hour with whoever has to do the work, to walk through it.
What is not included
Said up front, not found out later.
- Rewriting your queries or pipelines as part of the review.
- Data modelling or architecture consulting.
- Any ability to change the platform while we review.
If something you need sits outside this, say so when you get in touch. Better to find out now than to read it as a limitation in the report.
The others
Often done alongside this one.
AI and Automation
The jobs your team still does by hand — copying between systems, chasing people, typing the same thing twice. We work out what those hours cost, then build something to do them.
Learn moreSupplier contracts
What your agreements actually give you, set against what is in use — and what renews while nobody is watching the calendar.
Learn moreCloud bills
AWS, Azure and Google. What you are paying for, set against what is actually running and what anyone is actually using.
Learn moreQuestions
About this one specifically.
Most of our bill is one team. Is this still worth doing?
Usually yes, and it makes things easier — when the spend is concentrated, so are the findings, and there are fewer people to get agreement from.
Can you cover more than one platform?
Yes. Organisations often end up with two, and the overlap between them is frequently part of what we find.