Baseline performance metrics established

Severity: Recommended

You can't know if performance degraded if you don't know what "normal" looks like. Baselines are the reference point for all performance work.

Check automatically:

1. Look for performance/baseline documentation:

# Look for performance documentation
grep -riE "baseline|benchmark|performance.*metrics|p50|p95|p99|latency|throughput" docs/ README.md CLAUDE.md --include="*.md" 2>/dev/null

# Check for performance test results stored
find . -maxdepth 3 -type d -name "*performance*" -o -name "*benchmark*" -o -name "*loadtest*" 2>/dev/null | grep -v node_modules

# Look for APM dashboards referenced
grep -riE "datadog|newrelic|dynatrace|grafana.*dashboard" docs/ README.md --include="*.md" 2>/dev/null

# Check for benchmark scripts
find . -name "*benchmark*" -type f 2>/dev/null | grep -v node_modules

Ask user:

• "What's your API's p95 response time under normal load?"
• "How many requests per second can your main endpoints handle?"
• "Where are these baselines documented?"

Key metrics to have documented:

Metric	What It Measures
p50/p95/p99 response time	Latency distribution
Throughput (RPS)	Requests per second at steady state
Error rate	Baseline error percentage
Resource utilization	CPU/memory at normal load

Cross-reference with:

• LST-001 (tool needed to measure baselines)
• LST-005 (capacity limits are the upper bound of baselines)
• Section 17 (performance monitoring) - APM tools provide ongoing baselines

Pass criteria:

• Key metrics documented (response times, throughput, error rates)
• Baselines established from actual measurements (not guesses)
• Team knows where to find baseline data (APM tool, docs, etc.)

Fail criteria:

• No idea what "normal" performance looks like
• Baselines exist but are outdated (pre-major changes)
• Only anecdotal ("it feels fast")

Evidence to capture:

• Location of baseline documentation
• Key metrics tracked (p50, p95, p99, throughput, error rate)
• When baselines were last updated

---