Why
A unit cost metric without a target is an observation, not a KPI. Trending reveals efficiency degradation before absolute spend alerts fire — a 15% increase in cost-per-transaction is a signal even if total spend is within budget (it means you’re scaling inefficiently). Targets set expectations; trending detects drift; investigation finds the root cause.
What
Set unit cost targets per workload, track actuals against targets over time, and feed rising unit costs into architecture review and forecasting processes.
How
Establish Baselines and Targets
For each workload with a unit cost metric (from S9-01), set a target based on current performance and expected scaling behaviour:
| Workload | Current Unit Cost | Target | Rationale |
|---|---|---|---|
| Checkout API | $0.0023/txn | Flat QoQ | Scaling should be linear |
| Data Pipeline | $0.15/GB | -10% QoQ | Optimisation efforts underway |
| ML Inference | $0.012/prediction | Flat QoQ | New model, establish baseline first |
Target types: “Flat QoQ” (efficiency keeps pace with growth), “Declining” (active optimisation expected), or “Absolute” (e.g., stay below $0.003/txn).
Implement Trend Monitoring
Add to the unit economics dashboard:
- Target line on the trend chart (visual comparison of actual vs target)
- Alert when unit cost exceeds target by >10% for two consecutive months
- QoQ comparison table showing direction of travel
Connect to Governance Processes
| Trigger | Action |
|---|---|
| Unit cost rising while volume is flat | Investigate: likely resource waste or config drift |
| Unit cost rising proportional to volume | Expected — but validate commitment coverage |
| Unit cost rising faster than volume | Architecture review — scaling is non-linear. Trigger S7-01. |
| Unit cost declining | Success — feed into FinOps success story / newsletter |
Feed unit cost trends into the forecasting model (S3-05). If cost-per-transaction is rising 5% per month, the forecast should reflect this rather than assuming linear scaling.
Deliverable Checklist
- Unit cost targets set per workload (flat, declining, or absolute)
- Target lines added to unit economics dashboard
- Alert configured for target breaches (>10% for 2 months)
- QoQ comparison view operational
- Rising unit cost trigger linked to architecture review process
- Unit cost trends feeding into forecast model