An analytical platform's cost has grown faster than its usage. Which controls actually reduce it?

The controls, by effect

• Aggressive auto-suspend on idle. Warehouses left running are the most common single source of waste and the easiest to fix.
• Fix the data layout, since cost is dominated by data scanned and that is a layout property — partitioning against the actual filter predicates, clustering for secondary filters, compaction to a sensible file size, and statistics that are current. Fixing layout reduces the work at no recurring cost, which is the difference from adding compute.
• Result caching, so repeated identical queries do not re-scan. Dashboards refreshing on a schedule are the large beneficiary.
• Statement timeouts, bounding runaway queries rather than leaving them open-ended.
• Prune unused models and tables. A meaningful proportion of scheduled analytical work produces artefacts nothing reads, and finding them is cheaper than making them faster.
• Per-workload separation and budgets, since attribution is what changes behaviour — teams that can see their own spend reduce it, and teams that cannot, do not.
• Cost visibility at the moment a query is written, which changes query design far more effectively than a monthly report to a manager.

The structural cause

Elasticity moves the constraint from capacity to cost, and cost is a softer constraint that fails silently. A fixed cluster refuses work when full; an elastic platform satisfies whatever is asked — including a dashboard refreshing every thirty seconds — and reports it a month later.

The delay between action and feedback is the problem: a month is long enough for a pattern to become normal, be copied, and be embedded in dashboards.

The controls that are added last and should be first

The list above is well known and is typically implemented after the first surprising invoice, because beforehand each item looks like friction on a platform chosen for its ease of use. That is a genuine tension rather than an oversight, and naming it is what allows the controls to be adopted early.

The metric that should be on the dashboard

Cost per unit of business value — per report produced, per active analyst, per model trained — trended over time. Total spend rises with growth and tells you nothing; unit cost falling as usage grows is the only evidence the work is compounding.