Platform Cost Scorecards for Engineering Teams

Dashboards are passive. A scorecard makes cost health visible and competitive across teams. Here is a simple format that drives action without shaming.

What to measure

• Waste: (requested - used) / requested at p95 per service.
• Guardrail coverage: % of workloads with owner labels, LimitRange, ResourceQuota, PodSecurity, and network policies.
• Unit cost: $ per 1k requests or per customer for the product line.
• Reliability tie-in: SLO compliance; cost optimizations cannot break uptime.

Scorecard layout

• 4–5 metrics max, each with target bands (green/yellow/red).
• One monthly snapshot plus a 90-day trend sparkline.
• Dollar impact for the top two actions that would move a team to green.

Cadence and ownership

• Generate monthly; review in platform + team syncs.
• If a team is red for two months, open a ticket with clear fixes.
• Celebrate teams that move from red to green with shoutouts and budget reallocation.

How to calculate fast

• Automate data pulls from Prometheus (usage), price sheets (cost), and policy scanners (coverage).
• Cache results per namespace/team; avoid slow queries that block reviews.
• Store snapshots so teams can see before/after when they ship fixes.

Keep it constructive

• Frame it as “product performance” not “cost shaming.”
• Tie scorecard moves to roadmap benefits: better unit economics → more room for growth bets.
• Evolve targets quarterly as you improve defaults.

A scorecard keeps cost ownership close to the teams shipping code. Make it visual, actionable, and tied to dollars, and it will change behavior.***