Input and Output Metrics

Input and output metrics are Amazon’s discipline for separating controllable operating levers from lagging business results, so teams can manage the actions that create outcomes instead of only watching outcomes after the fact.

Key points

• Carr contrasts this with Amazon’s failed “fitness function” experiment, where several important metrics were weighted into one compound index ^[src-018].
• Compound metrics can become meaningless because they hide which action or reaction is driving revenue, customer growth, or other outcomes ^[src-018].
• Amazon’s lesson was to break metrics out individually and manage each one on its own terms ^[src-018].
• In the single-threaded leadership example, search teams would own controllable metrics such as top-result click-through or page-load latency, not just broad company outputs ^[src-018].
• The concept supports better accountability: a team can be judged by the inputs it improves and the outputs those inputs influence, rather than by a blurry aggregate ^[src-018].
• Statsig’s outcome-obsessed PM article applies the same discipline to product management: shipping roadmap tasks is output, while adoption, activation, funnel conversion, and business impact are the outcomes or leading indicators that should guide prioritization ^[src-034].
• When direct revenue attribution is unclear, Statsig recommends imperfect but useful leading indicators rather than abandoning outcome measurement and retreating to task completion ^[src-034].
• Statsig’s enterprise-scale article introduces Overall Evaluation Criterion as a controlled way to evaluate experiments across revenue, engagement, support, risk, and customer sentiment ^[src-036].
• This creates a useful tension with Amazon’s warning about blurry weighted metrics: an OEC can align long-term evaluation, but only if component metrics and trade-offs stay visible in a transparent catalog ^[src-036].

Related entities

• Amazon
• Bill Carr

Related concepts

• Single-Threaded Leadership
• Working Backwards (Book)
• Agentic Build / Deploy Boundary
• Outcome-Obsessed Product Management
• Roadmap as False Comfort
• Proxy Metrics in Experiments
• Overall Evaluation Criterion
• Enterprise-Scale Experimentation

Source references

• ^[src-018] Lenny’s Podcast — “Unpacking Amazon’s unique ways of working | Bill Carr (author of Working Backwards)” (2023-11-02)
• ^[src-034] Shubham Singhal — “Chasing metrics, not tasks: Why outcome-obsessed PMs win” (2025-05-22)
• ^[src-036] Yuzheng Sun — “Addressing complexity in enterprise-scale experimentation” (2025-04-23)