Percentile (p95, p99)

// Definition

A statistic that reports the value below which a given percentage of measurements fall. p95 means 95% of requests were faster than this number — and 5% were slower. Performance teams report tail percentiles because averages hide the slow long tail.

// Why it matters

Averages lie under load — a 200ms mean can hide a 4s p99 that's making your worst-affected users churn. Percentiles tell you the experience at the tail: p95 = 95% of requests are faster than this. QA reports percentiles, not means, because SLAs and real pain live in the tail.

// How to test

// Capture response times and assert the tail, not the average
const times: number[] = []
Cypress._.times(100, () => {
  const t0 = performance.now()
  cy.request('/api/dashboard').then(() => times.push(performance.now() - t0))
})
cy.then(() => {
  const sorted = times.sort((a, b) => a - b)
  const p95 = sorted[Math.floor(sorted.length * 0.95)]
  expect(p95, 'p95 latency').to.be.lessThan(800) // SLA threshold
})

// Related terms

• Latency

  The time between sending a request and receiving the first byte of response. Lower is better. Often reported at percentiles (p50, p95, p99) since averages hide tail behaviour.

• Response Time

  Total time from request initiation to response completion. Includes latency plus transfer and processing time. The end-user-perceived performance metric.

• Throughput

  The amount of work completed per unit of time — requests per second, transactions per second. Higher is better.