Factor 7 of 7 · AUG framework
Performance audit
Does the product feel fast? Performance is a growth multiplier, not a technical concern. Every other factor is amplified or capped by how fast the site loads, responds, and renders.
What this measures
Real-user Core Web Vitals (CWV) measured under realistic conditions: throttled 4G network, mid-range mobile device, cold cache. Not lab metrics. Not synthetic Lighthouse scores. The numbers Google actually uses for Page Experience ranking — and the numbers that determine whether visitors stay or bounce.
The 5 Performance metrics
- LCP (Largest Contentful Paint) — Time until main content visible. Target <1.5s. Failure threshold >4s.
- INP (Interaction to Next Paint) — Time from user input to paint. Target <200ms. Failure >500ms.
- CLS (Cumulative Layout Shift) — Visual instability (0-1 score). Target <0.05. Failure >0.25.
- TTFB (Time to First Byte) — Server response time. Target <400ms. Failure >800ms.
- Page weight — Total transferred KB. Target <100KB. Heavy >500KB.
Why this is a growth multiplier, not a technical concern
Measured AUG impact per Performance improvement:
- LCP 4s → 2s: Bounce -20%, Activation +15%, PV/session +0.4
- LCP 2s → 1s: Bounce -10%, Activation +8%, PV/session +0.2
- INP 500ms → 200ms: Activation +12%, rage-click rate -50%
- CLS 0.15 → 0.02: Bounce -5%, ad viewable rate +15%, pRPM +8%
- TTFB 1s → 300ms: Mobile bounce -18% (mobile disproportionately affected)
- Page weight 500KB → 100KB: Tier-3 geo Activation +35% (data-cost-sensitive users)
Every CWV improvement compounds into Activation, Engagement, Retention, AND Monetization. Performance isn't a separate factor — it's a multiplier on every other factor.
How it's scored (1-10 rubric)
Composite across the 4 CWV sub-scores. Each contributes 0-2.5 points:
- LCP p75 <1.5s: 2.5 · 1.5-2.5s: 1.5 · 2.5-4s: 0.75 · >4s: 0
- INP p75 <200ms: 2.5 · 200-500ms: 1.5 · >500ms: 0
- CLS p75 <0.05: 2.5 · 0.05-0.1: 1.5 · 0.1-0.25: 0.75 · >0.25: 0
- TTFB p75 <400ms: 2.5 · 400-800ms: 1.5 · >800ms: 0
- 1-2: All CWV in poor zone. Site loads >4s on 4G. Mobile experience broken.
- 3-4: Mixed; some CWV need work. LCP ~3s.
- 5-6: All CWV in “needs improvement”. LCP ~2s.
- 7-8: Most CWV good. LCP <1.5s. INP <200ms.
- 9-10: All CWV good zone. LCP <1.2s. Sub-100kB pages.
The 10 quick-win tactics
- Static export (Next.js
output: 'export') — LCP -60%, TTFB -80% - Cloudflare Pages / Workers deployment — TTFB -70% globally
- Critical CSS inlined, rest lazy — LCP -30%
- Font stack without custom fonts (or
font-display: optional) — LCP -20%, CLS -80% - Image dimensions +
decoding="async"+ AVIF — CLS -90%, LCP -25% - Defer all non-critical JS — INP -50%
- Zero above-fold images (CSS gradients instead) — LCP <800ms
- Resource hints (
preconnect,dns-prefetch) — TTFB -100ms - Remove jQuery / lodash / moment.js — Page weight -70%
- content-visibility: auto for below-fold — Paint time -40% on long pages
Performance as moat
A site with LCP 1.2s + INP 150ms + zero CLS + 80KB pages:
- Ranks higher (Page Experience is a Google ranking signal)
- Converts better (every AUG factor lifted)
- Serves Tier-3 geos meaningfully (100k+ potential users/mo that slow sites lose)
- Costs less to run (no CDN bandwidth overage)
- Survives mobile degradation (slow 3G test passes)
Most competitors won't invest here because it's invisible to them. That's the moat.
Performance instrumentation
Wire web-vitals library on Day 1. Send measurements to Plausible (custom events) + Cloudflare Web Analytics (auto-captures CWV). Weekly review: Google Search Console → Page Experience report → “Poor URLs” list → fix one per week.
After 4 weeks, you'll have data to identify the slowest 20% of pages — the ones where 80% of growth friction lives.
The end of the funnel
Performance is Factor 7 — the final factor — because it acts on every prior factor. Strong scores 1-6 with weak Performance 7 = leaking bucket at the floor. Strong Performance with weak 1-6 = great-looking empty store.
Run the full 7-factor audit to see how your scores compound: /audit.
Previous: Factor 6 — Monetization. Or return to methodology overview.