Factor 4 of 7 · AUG framework

Retention audit

Do users come back without notification pressure? Retention is the durable compounding engine — the only factor where week N benefits from work done in week N-26.

What this measures

Whether users open the product unprompted in subsequent sessions. Not push notifications. Not email re-engagement. Not paid retargeting ads. Direct return = evidence the product earned a place in the user's memory.

Why this matters more than any other factor

Retention compounds. A site with 30% D30 retention and 1,000 weekly visitors grows to ~5,000 weekly visitors in 6 months through return-visit accumulation alone. A site with 3% D30 retention grows to ~1,100 in the same period despite identical acquisition. The gap is geometric, not linear.

Per the AUG composite formula, a 5% Retention score drags a 9-9-9-9-9-9 product down to a ~22 composite. Conversely, a 9 in Retention with mediocre 5s elsewhere produces a ~25 composite — Retention is the leverage point.

The 8 retention metrics

1. D1 return rate — % users returning next day. Target ≥8% reference, ≥12% utility.
2. D7 return rate — % returning within 7 days. Target ≥15% reference, ≥25% utility.
3. D30 return rate — % returning within 30 days. Target ≥25% reference, ≥40% utility.
4. Cohort retention curve — % cohort retained week-over-week. Week-4 target ≥15%.
5. Return-visit frequency — Median visits per returning user per month. Target ≥2.5.
6. Bookmark rate — % sessions adding bookmark (inferred via return-visit-without-referrer). Target ≥5%.
7. Direct-type-in rate — % sessions arriving via direct (typed URL). Target ≥12%. ≥20% = strong brand.
8. Brand search rate — % sessions arriving via search for brand name. Should grow month-over-month.

How it's scored (1-10 rubric)

• 1-2: D7 <3%. No core loop = no return reason.
• 3-4: D7 ~8%. Some return, but cohort decays fast.
• 5-6: D7 ~15%. Core loop solid. Email digest opt-in working.
• 7-8: D7 ~25%. Direct traffic ≥15%. Brand search growing.
• 9-10: D7 ≥40%. Direct traffic ≥25%. Users name the brand unprompted in feedback.

The core-loop test

Before building anything else, ask: what problem does the user have that recurs?

• Daily recurrence: Extreme retention potential (Wordle, weather, news) — but high-churn risk.
• Weekly: Best for fleet SaaS. Meal planning, market check, fitness log.
• Monthly: Solid. Budget review, tax tracking, billing.
• Quarterly: Good for finance/business. 13F filings, earnings reports, industry benchmarks.
• Annually: Low. Tax prep, tuition calcs.
• One-time: Zero retention. “How many grams in a cup” — user comes once, leaves forever.

Year 1 SaaS rule: build for Weekly or Monthly recurrence. Avoid one-time concepts unless the dataset is vast enough for programmatic long-tail (e.g., tax-treaty lookup has one-time intent per user but 190² pages covers infinite users).

The 10 retention-driving tactics (ranked by D30 lift)

1. Core loop = user's recurring problem — +8% absolute D30 when the product solves a weekly/monthly recurring pain
2. Bookmarkable result URLs — +4% D30. Every result has a stable shareable URL; encourage “save this”
3. Email digest (opt-in, no spam) — +6% of opted-in. Weekly newsletter with new data/features. NEVER auto-subscribe.
4. Save/watchlist feature (no-signup, localStorage) — +5%. Users build state on the site; localStorage means zero friction.
5. Browser notification opt-in (carefully) — +3% (of opted-in); hurt if aggressive. Only ask AFTER user completes second action.
6. Time-sensitive content — +4% for freshness-driven niches. Quarterly 13F updates, seasonal recipes, weekly trend reports.
7. Progressive feature unlock — +3%. “You've calculated 3 times — here's Pro Mode”
8. RSS feed per category — +2%. Feedly/Inoreader subscribers = recurring visits, no email needed.
9. “Your history” personalization — +7%. Session-to-session memory of what user viewed/calculated.
10. Trigger-based return emails — +5% of opted-in. “Your brine was 3 weeks ago — how did it turn out?”

Retention anti-patterns (immutable hard-rejects per AUG framework)

• Growth hacks that spike D1-D7 but collapse D30 (fake notifications, aggressive popups)
• Hidden-cost trials (credit card required upfront for “free” product)
• Confirmshaming unsubscribe (“No thanks, I hate saving money”)
• Auto-email upon visit without explicit opt-in
• Feature removal behind paywall after user was using for free (rug-pull)
• Notification spam (>1/week in most verticals)
• Dark-pattern retention (FOMO manufactured scarcity)

These hurt D90+ retention catastrophically. The AUG floor rule: any ship that drops 7d retention by >10% is auto-flagged as rollback-candidate. A revenue-positive ship that hurts retention is a long-term net negative.

Previous: Factor 3 — Engagement. Next: Factor 5 — Advocacy. Or run the full 7-factor audit.