liveBook · Manning

Performance requirement

As explained in Fast by Default

A criterion specifying the required speed, responsiveness, or resource behavior of a software system.

A performance requirement is a specific, measurable commitment for how fast or resource-efficient a system must be in a user-relevant situation. It replaces vague requests such as “make search faster” with a testable target such as “95% of search queries return within 200ms.” The requirement should state the flow, threshold, population or context, and measurement condition. section 3.0 section 3.4

Treat the user journey as the unit of performance. Users experience searching, checkout, startup, exports, and editing—not isolated database queries, bundles, or endpoints. A requirement for a complete flow prevents teams from improving one component while an unaddressed bottleneck still delays the outcome. section 3.1.3 section 3.4.1

Write an effective performance requirement by making it:

  • Specific: name the operation, threshold, and measurement, such as filtering within 500ms for 95% of users. section 3.1
  • User-centered: tie the target to a visible frustration or job rather than an internal metric with no observable effect. section 3.1.3
  • Context-aware: distinguish relevant devices, networks, file sizes, project sizes, or user segments. section 3.4 section 3.4.1
  • Realistic and jointly owned: account for engineering constraints and set the target before optimization begins. section 3.0 section 3.6
  • Maintained: review and tighten targets as product maturity, infrastructure, and user expectations change. section 3.4.3

From component tuning to a flow requirement

A team reports smaller bundles, faster queries, and lower memory use but cannot tell whether filtering is good enough. A better requirement is: “Filtering completes within 500ms for 95% of users, with no individual step exceeding 200ms.” Profiling can then reveal sequential requests as the limiting step, directing work toward parallel requests and caching rather than unrelated optimizations. section 3.1

Requirements also make trade-offs explicit. Pair a feature commitment with its performance constraint, such as delivering real-time commenting without page load exceeding three seconds. If a prototype temporarily regresses performance, record the accepted limit and the planned target rather than letting performance debt become permanent. section 3.4.4

Table 3.7 Performance targets by product maturity
Platform
Early Stage
Growth Stage
Maturity Stage
Web
LCP < 4s, no failures
Meet Core Web Vitals
Exceed Core Web Vitals
Backend
P95 < 1s, 99% uptime
P95 < 500ms, 99.5% uptime
P95 < 200ms, 99.9% uptime
Mobile
Cold start < 5s, no crashes
Cold start < 3s, 60fps
Cold start < 2s, smooth animations
Desktop
Launch < 10s, responsive
Launch < 5s, efficient resources
Sub-second launch, minimal resources

Compare example target levels for web, backend, mobile, and desktop products at early, growth, and maturity stages. section 3.4.3

Put requirements into everyday delivery: choose one business-critical flow, establish a measurable baseline and target, make it part of the definition of done, and use measurement before and after changes to validate user impact. Performance requirements can affect adoption as well as satisfaction: services that fail buyer load tests or respond much slower than alternatives may be rejected before features are evaluated. section 3.5 section 2.2.3

How Fast by Default covers it

How deep it goes
Develops requirement writing through flow examples, contextual targets, trade-offs, maturity reviews, and measurement guidance.
Examples
Uses filtering, checkout, document export, app startup, feed refresh, code-editor workflows, and API evaluation under load.
What it assumes
Assumes comfort with production code, HTTP, databases, and deployment pipelines; does not require performance specialization.
What you can do after
Apply it by defining, reviewing, and validating measurable flow-based targets for critical product work.
How it connects
Connects performance requirements to user journeys, percentiles, performance budgets, CI/CD, field data, and lab data.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Fast by Default. The section links point to the parts of the book it draws on.