Skip to content
Case study

A warehouse bill that doubled in a year

Nobody could see which queries cost what. We added cost attribution per team, rewrote the twelve worst models, and set a monthly budget alert.

Monthly platform cost
-46%
Payback
5 mo
Models rewritten
12
Teams with budgets
9

The situation

The warehouse bill had gone from $45k to $84k a month in a year, and nobody could say which team or which workload was responsible. Procurement was preparing to renegotiate the contract.

Constraints

Cutting spend could not slow the morning reporting run, which operations depended on being ready by 6am.

What we did

Cost attribution came first: separate compute per team, query tagging for shared jobs, and a dashboard anyone could open rather than a monthly PDF.

The distribution was as uneven as usual. Twelve scheduled models accounted for most of the spend, and the worst of them rebuilt a table nightly that only needed the last two days.

We rewrote those twelve as incremental models, added a monthly budget alert per team, and left the review process behind.

What happened first

The single largest saving landed in week one, before we changed any code. A team opened the new dashboard, recognised a job they had forgotten about, and switched it off.

Results

MeasureBeforeAfter
Monthly warehouse spend$84k$45k
Longest scheduled job3h 40m38m
Teams with cost visibility09
Budget alerts in placeNoneMonthly
We stopped arguing about the invoice and started managing it. The cost dashboard is open on a screen in the office.
Dana WhitfieldDana WhitfieldOperations Director, Hollis Logistics

What we left behind

  • A project in the client's own repository, with tests and CI
  • Documented definitions with named owners
  • Runbooks for every scheduled job, written for a non-engineer
  • Two weeks of training for the team taking it over

Tell us which number you cannot trust

Send one paragraph about the report that always starts an argument. A data lead, not a salesperson, replies within one business day.