Toby — Weekly Strategist Memo

Run date: 2026-05-11 · Companion to state-of-business-2026-05-11.html · Anchored to fresh numbers, not vibes.

Read of the week

The headline shifted. For weeks the story has been churn-and-cliff; this week it's acquisition collapse. New active subs / 7d printed 10 — the lowest ever — while cancellations held at 32. Net contraction is now ~−22 paid subs/wk, quiet but real. Top of funnel is louder (+3.6% WoW signups) but softer (−61.2% YoY, worse than 5/6's −58.3%). Combined with the 4th consecutive week of recurring CWS "blank page" complaints, the SW boot-path regression hypothesis is no longer speculative — it's the most likely root cause and it's costing us dozens of paid conversions per week.

Three small positive signals worth chasing: 9 queued cancels reversed with no save flow shipped (something organic is working — find it); May-26 cliff dropped $4,040 → $3,135 as a chunk of renewals rolled forward (replicate the conditions for Feb-27); and WAU/MAU inched to 85.3% — the habit-loop cohort is intact, the leak is in the broader 30d tail.

Five moves below. Ordered by leverage × ship-ability in <2 weeks.

The five moves

NEW

1 · Unblock the free-to-paid upgrade path (the "Axel fix")

Anchored to: "Axel Streichardt · Apr 27 · 3★: couldn't locate monthly plan option despite wanting to upgrade from free tier" + headline metric: 10 new active subs / 7d (all-time low) against 722 signups. Conversion of intent-to-pay free users is leaking visibly.
Problem
Free-tier users who want to pay can't find a monthly option from inside the product. The CWS comment isn't a one-off — it's a recurring theme on review pages and it shows up in our paid-sub print: 10 new actives is acquisition death even before churn enters the equation.
Hypothesis
A single visible "Upgrade" entry point on the new-tab, routing to a pricing page that shows monthly and yearly side-by-side (not one-or-the-other), recovers 1.5–2× the current new-paid rate within 4 weeks.
Smallest test
Ship a Chrome update that adds a single "Upgrade" pill to the new-tab top bar (no behavioural change for already-paying users). Route to /pricing?ref=newtab. Make sure both $5/mo and $48/yr are visible above the fold. Instrument click → checkout funnel.
Success metric
Weekly new active subs (baseline 10). Target: ≥25/wk within 4 weeks of ship. Secondary: /pricing?ref=newtab click-through rate > 1.5% of new-tab opens.
Cost
S — UI pill + pricing page tweak + a CWS update. Ship in one sprint.
Market scan: Workona, Cluster, OneTab all default-feature an in-product upgrade CTA on the most-trafficked surface. We're the outlier hiding it. This isn't a category innovation — it's table stakes we never re-shipped after the freemium pivot.
ELEVATED

2 · Service Worker boot-path forensics (and a Beta-channel hotfix)

Anchored to: YoY signups −61.2% (vs −58.3% on 5/6 — the gap is widening) + recurring CWS quotes: "Killdeer · Mar 6 · 2★: extension tab stopped opening, displaying blank page on infinite load". 4th consecutive week the hypothesis has appeared in the report.
Problem
If new installs are landing on a blank new-tab because of a Manifest V3 service-worker registration/lifecycle failure, organic signups don't convert and CWS rating slowly bleeds. The YoY gap widening (not narrowing) is the smoking gun.
Hypothesis
A specific SW lifecycle change between our last "healthy" install cohort (mid-2024) and today drops the new-tab override on cold installs in some Chrome builds. Fixing it recovers 20–40% of organic signup→activation, which alone would 2–3× the new-paid rate.
Smallest test
(a) Bisect service-worker.js commits between mid-2024 and today on a clean Chrome profile — reproduce the blank-page condition. (b) Wire a minimal telemetry beacon (sw_boot_ok) to the new-tab so we can measure cold-install activation rate going forward. (c) Ship the fix to Beta channel first; measure activation delta vs Stable for 7 days.
Success metric
Beta vs Stable cold-install activation rate (proxy: sw_boot_ok beacon / install). Target: ≥15 pt activation gap. Secondary: weekly signups recover toward the 8-week median (~830) once shipped to Stable.
Cost
M — eng diagnostic time + telemetry wire-up. Bisect should be 1–2 days; fix scope unknown until reproduced.
Market scan: Chromium's own issue tracker and chromium-extensions Google Group both have active threads on intermittent MV3 SW registration failures into 2026 — it's a known class of bug. We're not alone, but we're not protected either. Adding a boot beacon is the only defence.
CARRY · REFRAMED

3 · Save-flow MVP, instrumented from day one, pre-staged for Feb-27

Anchored to: 605 cancels vs 8 logged retention-offer accepts in trailing 12w (1.3% accept rate, mechanical not behavioural) + Feb-2027 cliff: 2,371 subs / $12,528.50 MRR / 86.5% legacy — ~31% of net MRR rolling at once.
Problem
We don't have a working save funnel. When the Feb-27 mega-cliff hits in 9 months, we will catch zero. Industry benchmark for a basic cancel flow is 15–30% save; a sophisticated one hits 30–42%. We're catching 1.3% — which is more likely "the flow isn't wired" than "we suck at retention".
Hypothesis
A minimum-viable cancel interstitial offering pause for 1 month OR 50% off for 3 months, presented at the moment of cancel click and logged to retention_offers on every impression, captures 15–20% even with no personalization.
Smallest test
Build a two-card interstitial that intercepts the cancel button. Log impressions + accepts + reasons. A/B half the cancel population against the current straight-cancel path for 4 weeks. Don't optimize copy — just measure baseline.
Success metric
Retention-offer accept rate ≥ 10% within 30 days of ship (a 10× over current 1.3%). Secondary: 12w cancels trending below 600.
Cost
M — billing-page UI + Stripe API for pause/discount + 2 new retention_offers insert paths.
Market scan: ChurnWard, ProsperStack, and Chargebee Retention all converge on 20–30% as a normal SaaS save rate. Top-decile (30–42%) requires reason-matched offers — we don't need that for v1, we need anything wired.
NEW

4 · Reverse-engineer the 9 saves we got for free

Anchored to: "Queued to cancel: 302 → 293 (−9 WoW)" with no save flow shipped + "May-2026 cliff bucket dropped from $4,040.50 → $3,135" — some users decided not to cancel. We have no idea why.
Problem
Nine paying subs flipped from queued-to-cancel back to active this week, and a chunk of May renewals rolled forward without intervention. That's free retention signal — if we can identify the behavioural pattern of "reverters" vs "completers", we can engineer it deliberately for the Feb-27 cohort.
Hypothesis
Reverters re-engaged with a specific feature (most likely: opened/edited a saved collection) within the 2-week pre-renewal window. The save isn't random — it's a re-discovered habit moment.
Smallest test
One-time analysis query: pull the 9 user_ids that exited the queued-to-cancel pool in the last 14 days; pull a matched cohort of 50 who churned in the same window; compare last-30d session counts, collection-open events, tab-saves, and tenure. Look for the variable with the cleanest separation.
Success metric
Identify ≥ 1 behavioural variable (e.g. "≥3 collection opens in last 14d") that distinguishes reverters from churners at p < 0.1. Feeds directly into idea 3's targeting and a future "we noticed you opened X yesterday — pause instead?" save-flow variant.
Cost
S — one analyst-day in Postgres + Amplitude.
Market scan: skipped — this is internal forensics, no external comp.
CARRY · NARROWED

5 · One targeted reactivation email to "real" dormant users

Anchored to: 305,595 dormant 90d+ users (+991 WoW) — pool grew slightly faster than weekly signups again. With new-active-subs at 10/wk, even a 0.5% paid-conversion off reactivated returners would 5× the current rate.
Problem
305K dormant accounts is 437× our weekly organic signups. Last week's memo flagged the pool; this week the pool grew. We've never sent a single deliberate reactivation message.
Hypothesis
A single, plain-text email to dormant users who were real users at their peak (≥5 collections, ≥50 tabs saved lifetime) — subject: "Your collections are still here" — gets a 2–10% 14-day return rate, top decile ~17%. Even 3% return on a 50K-cap send is 1,500 returners, vastly more than weekly organic activation.
Smallest test
Segment: last_active_at < now() - 90 days AND collection_count ≥ 5 AND tabs_saved_lifetime ≥ 50. Cap first send at 5K. Hold out 1K matched control. Plain-text, no marketing copy, no promo — just "your data is still here, here's the link". Measure 14-day return.
Success metric
14-day return rate ≥ 2 pts above control. Secondary: paid-tier conversions among returners (any > 0 is informative given our 10/wk baseline).
Cost
S — one SQL segment, one email send, two-week wait.
Market scan: Zeta, Netcore, Churnkey all cite 2–10% reactivation as the realistic band; top-decile segmentation reaches 17%. Filtering on prior collection activity puts us in that top-decile cohort by construction.

Retired or deferred

What to watch next week

New active subs / 7dbaseline 10 (all-time low). Any drift back toward the 20–25 range is meaningful.
YoY signup gapcurrently −61.2%. Watch for stabilization vs further widening (i.e. SW issue accelerating).
Queued-to-cancelcurrently 293. Does the reverter trickle continue? Is it 282 next week?
retention_offers row countstill 16 lifetime / 8 in 12w. Any increase means somebody started instrumenting.
Trialing zombiesstill 8,833 (4 weeks straight). Any movement = sync was touched.
CWS blank-page complaintstrack frequency in new reviews — the canary for SW health.
Feb-27 cliff $currently $12,528.50. Renewals or downgrades will move this; we want it dropping.
Net MRRcurrently $40,390.50 (−$112.50 WoW). Watch for the −22 subs/wk contraction compounding into MRR drift.