Toby — Product Strategist
Weekly memo · 2026-05-18 · companion to State of Business
2026-05-18 · 5 moves anchored to this week's numbers.
TL;DR — the headline shifted again
Acquisition is no longer a "new headline" — it is the business.
Net MRR is flat at $40,356 and that flatness is now the most misleading number on the dashboard.
Two non-flat realities: 8 new active subs in the last 7 days (down from 10, down from 16, an all-time low),
and 0 retention save offers in 30 days vs 122 cancels. The first means we are not refilling the bucket.
The second means we are not patching the hole. We have 8.5 months until 31% of MRR
($12,447.50 / 2,354 subs) hits its first post-legacy renewal in Feb-27, and the cohort is already leaking
(−17 subs / −$81 MRR WoW with no intervention in sight).
The five moves below are the 2-week-or-less plays that earn the right to even attempt a Feb-27 save campaign.
Week-over-week move continuity
| Theme | Status | Δ from 2026-05-11 |
|---|
| Save flow MVP | URGENT | elevated — SoB report explicitly recommends "this sprint" |
| SW boot regression bisect | URGENT | elevated — 4th sub-800 signup week, YoY −60.8% |
| Feb-27 offer-ladder draft | CARRY | narrowed — design only, no build yet |
| Signup→first-paid funnel instrumentation | NEW | replaces "reverse-engineer 9 saves" — bigger leak |
| Targeted dormant resurrection email | CARRY | unchanged — small, cheap, repeatable |
The 5 moves
Anchor: SoB 2026-05-18: "Retention save flow remains silent (0 logged offers in the last 30d;
only 7 in the 12-week window vs 122 cancels in 30d)." Industry median save-flow recovery for SaaS subscriptions
is materially non-zero — credit-card free-trial flows alone hit 30% conversion in ChartMogul's 2026 dataset.
A zero is plumbing failure, not offer failure.
- Problem
- There is no surface for a churning user to land on. 122 cancels/30d × roughly $4.90/seat avg = ~$600 MRR/mo bleeding through a flow that doesn't exist.
- Hypothesis
- A single-screen interstitial between "Cancel" and confirmation, offering one of {downgrade to monthly, 50% off 3 months, pause for 60 days}, saves 10–20% of cancellations. Even the low end recovers ~$60 MRR/mo immediately; the real prize is being instrumented and battle-tested before Feb-27.
- Smallest test
- Two-week MVP: (a) wire the cancel button to a route that writes to
retention_offers.issued with a single hardcoded offer, (b) accept/decline writes accepted, (c) ship to 100% of churning users — there is no smaller A/B than "does this row exist at all." Verify by re-running the SoB save-flow query next week.
- Success metric
- Week 3:
retention_offers shows ≥80 issued rows in the trailing 30 days (matching cancel volume). Save rate is a secondary KPI; existence is the primary one.
- Why now
- This is the only retention lever we can install before Q4 2026, and Q4 is the latest reasonable runway to be instrumented before the Feb-27 wave.
Anchor: SoB 2026-05-18:
"Signup velocity 720/wk, still at −60.8% YoY — the seven-week downtrend
is now flat-to-down." Four consecutive weeks below 800/wk. New active subs hit
8 / 7d (vs 10 last week,
16 two weeks ago) — same direction, accelerating. Chrome MV3 service-worker lifecycle bugs (blank pages on cold boot, SW becoming inactive
without recovery) are
documented in the Chromium-extensions community
— exactly the failure mode that fits CWS "blank extension page" reviews from late April.
- Problem
- 720 signups → 8 new active paid subs. That's 1.1% sign-to-paid this week, vs an industry median around 8% in ChartMogul's 2026 SaaS conversion report. We are roughly 7× below normal. The most likely single explanation: a chunk of new installs never reach the first usable screen.
- Hypothesis
- A regression in the service-worker boot path (commit landed mid-to-late 2025) causes the new-tab override to silently fail on a non-trivial slice of profiles. The user sees a blank tab, doesn't return, never converts. This is consistent with: 4-week signup collapse, no churn spike (existing users are fine), and the "blank page" CWS quotes.
- Smallest test
- (1) Diff the SW entry-point file between v4.x mid-2025 and current; (2) bisect commits touching
chrome.runtime.onInstalled, onStartup, and the new-tab init path; (3) ship the suspect-fix to the Beta channel; (4) compare 7-day new-active-subs Beta vs main.
- Success metric
- Within 2 weeks: identify a candidate commit. Within 4 weeks: Beta cohort new-active-subs/wk recovers to ≥14 (last non-collapse baseline) while main stays at ~8. If Beta also flatlines, hypothesis is wrong and we redirect to acquisition/funnel (Move 4).
- Why now
- Every week we leave this unbisected is ~12 lost new actives. At 50× monthly LTV that's real money, but the bigger cost is that the Feb-27 save story only works if there's a healthy top-of-funnel behind it.
Anchor: SoB 2026-05-18: "Feb-2027 mega-cliff barely moved: 2,354 legacy subs / $12.45K MRR
due to renew that one month (−17 subs WoW; cohort is leaking, not collapsing)." 31% of all MRR.
The cohort has already begun shedding revenue with no intervention — that is the data point. They are voting
with their feet before the renewal even arrives.
- Problem
- 2,354 subs on the legacy
nBvJx1dq coupon at ~$4.50/seat/yr will hit their first non-legacy renewal in Feb-27. We have no offer ladder, no email cadence, no in-product warning, no migration path. The cohort starting to leak now is a tell that they remember and they're price-sensitive.
- Hypothesis
- A three-tier ladder (loyalty discount → monthly-step-down → pause-and-resume) presented 90 days before renewal, in-product + email, saves 40–55% of the cohort vs the do-nothing baseline of ~25% (industry default annual-to-annual auto-renew survival for re-priced contracts).
- Smallest test
- Two-week deliverable: a 1-page spec + draft email copy, reviewed with 3 customers from the May/Jun 2026 mini-cliffs (153 + 208 subs) — these are our natural beta cohorts because they're hitting renewal next quarter at the same price step. Do not build the flow yet. Do figure out which tier is the right default.
- Success metric
- End of sprint: spec + copy approved. Validated against 3 customer reactions ("would you take this?"). Engineering effort sized so we can commit a Q3 build.
- Why now
- 8.5 months out is the latest moment we can start. Q3 is the latest we can build. Q4 is the latest we can A/B. Slipping any earlier step pushes the campaign into the cliff window itself, which is too late to iterate.
Anchor: SoB 2026-05-18 KPI grid: "Signups (7d) 720 · New active subs (7d) 8."
That ratio (1.1%) is the worst we've recorded. ChartMogul's January 2026 dataset puts the median free-to-paid
at 8%. Without funnel events between signup and paid, we cannot tell Move 2 (SW boot regression) from a paywall
regression from a payments-page regression from an entirely market-driven decline.
- Problem
- We can see the top of the funnel (signup) and the bottom (paid sub). The middle is dark. Last week we conjectured an Axel-style "can't find monthly" UX leak; this week we have no way to confirm or reject it.
- Hypothesis
- Adding 5 events (
signup_complete → first_session_opened → first_tab_saved → upgrade_page_viewed → checkout_started → checkout_completed) will reveal a single dominant step where ≥40% of the loss happens. That step is then a single-owner fix, not a guessing game.
- Smallest test
- Two-week deliverable: instrument client-side via existing Amplitude pipeline (
toby_amplitude.EVENTS_306007), backfill from server logs where possible, segment by acquisition source. Output: one waterfall chart for the strategist memo next week.
- Success metric
- By 2026-06-01: funnel waterfall in next week's SoB. One step identified as the dominant leak. Hypothesis for that step's fix appended to next week's strategist memo.
- Why now
- This is the cheapest move on the list and it unblocks Moves 1, 2, and 5 by telling us where the bleed actually is. Postponing this is postponing every other decision.
Anchor: SoB 2026-05-18: "Dormant (last_active >90d): 305,651 · +56 WoW."
The dormant pool grew by 56 users while signups (720) and new-actives (8) shrank. The pool is filling faster than
the top of the funnel can dilute it. This is the only large user reservoir we own outright.
- Problem
- 305K dormant users represent more potential reactivations than the next 12 months of net-new signups combined. We have done nothing with this pool except watch it grow.
- Hypothesis
- A single short email to a narrow slice (5,000 users dormant 90–180 days, signed-up but not paid) with one CTA — "your saved tab collections are still here" — lifts 7-day reactivation in that cohort by 5–8 percentage points over an unsent control.
- Smallest test
- Two-week deliverable: 5K send, 5K hold-out from the same cohort. Single CTA, single landing URL. No discount, no upsell. Measure
users.last_active in the following 7 days.
- Success metric
- Treatment cohort 7-day-active rate ≥ control + 5pp. If yes, ladder the test to 50K next month and measure paid-conversion downstream.
- Why now
- It is the smallest possible test on the largest possible audience. If it fails, we learn that email is not the right channel and we stop. If it works, it is a 2-week-build reactivation muscle we own forever.
What we are retiring or deferring this week
- Axel upgrade-path fix (last week, NEW) — superseded by Move 4. We were hypothesizing one step in the funnel; now we are going to instrument the whole funnel and let the data point to the right step. Re-promote it next week if Move 4 confirms the upgrade page is the leak.
- Reverse-engineer the 9 reverted cancels (last week, NEW) — superseded by Move 1. Once the save flow exists, we will have instrumented signals on natural saves; we no longer need to do anomaly archaeology on 9 rows.
- Trialing-zombie reconciliation (deferred again) — 8,833 stuck rows, unchanged WoW. Still pure data hygiene, no MRR impact. Re-evaluate when it blocks a measurement we actually need.
- Educator vertical wedge (deferred again) — irrelevant until acquisition isn't bleeding.
- Stripe Sigma LIVE rotation — recommended in the SoB but does not move customer behavior. Engineering hygiene, keep in the SoB action list, not in the strategist memo.
What to watch for in next week's SoB (2026-05-25)
- New active subs / 7d — is the slide continuing past 8, or did it stabilize? This is the single most important number on the dashboard right now.
- retention_offers row count, last 30d — Move 1's binary signal. Either it's still 0 or it isn't.
- Signup velocity — 5th sub-800 week confirms the trend is structural; a recovery toward 900+ would partially un-confirm the SW boot hypothesis.
- Feb-27 cohort MRR — last week $12,528.50, this week $12,447.50 (−$81). If next week is another −$50–$100, the cliff is now a slide and we're behind on Move 3.
- YoY signup delta — widening past −61% or stabilizing around it.
- CWS rating / new "blank page" reviews — direct corroborating signal for Move 2.