Needs more than a quick edit7 of 12 checks passed
The unsupported claim about 190 referrals would need verification, and the missing per-profile decay calculation means the reasoning on the primary loop's health is incomplete.
Output
Memo: How Tutorly Actually Grows
To: Rachel Dunn and the exec team From: [Staff PM] Re: Growth loops, the paid plan, and squad focus for the next two quarters
The short version
Our growth comes mainly from organic search, and that engine runs on reviews. Reviews only come from completed lessons. Two things are slowing the engine. Most new tutors never get a first booking, and our click-through from search has dropped.
Tripling paid acquisition would buy our lowest-value parents at a marginal cost about 2.7 times what they are worth. It would also get us nowhere near 50,000. I recommend we hold paid flat or trim it, measure what it actually adds, and put all three squads on the review and search engine.
1. Our growth loops
``` LOOP A: Search–review (PRIMARY) Parent books → lesson completed → review → profile reaches 3+ reviews ↑ ↓ new parent ← clicks profile ← ranks/shows in search results
LOOP B: Parent referral Active parent → invites (0.31/mo) → 14% book → new active parent ↺
LOOP C: Tutor referral (supply) Tutor → refers tutor ($50) → new profile → [61% get no booking in 60 days → no reviews]
LINE D: Paid (not a loop) $ → new parent → (some lessons feed Loop A) → spend stops, flow stops ```
Loop A is primary. It is compounding, but it is being throttled. - It brings in 52% of new booking parents, the largest share. - These parents are among our best: 48% are still booking at six months, and their net LTV is $260. - Reviewed profiles capture the traffic. Profiles with three or more reviews get 81% of organic sessions, while profiles with none get 4%. - Organic sessions grew 40%, but booking parents grew only 20%. The loop is producing traffic faster than it converts. - There are two leaks: - Click-through has fallen. On our top 200 keywords it dropped from 6.1% to 5.0% since the March results-page changes, an 18% relative fall. - New tutors are stuck. 61% of new profiles have no booking within 60 days, so they can never earn the reviews that make them visible.
Loop B is contributing. - Referred parents are our best customers: 51% retained at six months, with a net LTV of $275. - It is not viral. Each active parent produces about 0.043 new booking parents a month. Over a typical parent lifetime that is a k-factor of roughly 0.3–0.4, well below 1. - There is a data problem. Applied to 37,000 parents, those invite rates imply about 1,600 referred parents a month. Our attribution says referrals are 9% of new parents, which is closer to 190 a month. One of these figures is wrong, and we need to know which before investing here.
Loop C is compounding the wrong thing. - Tutor referrals brought in 44% of new tutors. - Active tutors grew 62% while booking parents grew 20%. - As a result, bookings per tutor fell 25%, from 9.1 to 6.8 a month. - The loop mass-produces unreviewed profiles that search ignores, and probably creates frustrated tutors who leave. In supply terms it compounds; in value terms it is decaying.
Line D (paid) is decaying. - Paid parents are our weakest: 22% retained at six months, with a net LTV of $120. - Their cost is rising (see below). - Paid does feed Loop A a little, because these parents' lessons generate reviews. But it stops the moment spend stops.
2. The plan to triple paid acquisition
The 2.2 ratio is real but misleading. It divides our blended LTV of $210 by our average CAC of $95. Paid parents are worth $120, not $210, so the true paid ratio at average cost is about 1.3.
The marginal picture is much worse. Last quarter: - At $60k a month and $82 per parent, we bought about 732 parents a month. - At $90k a month and $109 per parent, we bought about 826 parents a month. - So the extra $30k bought about 94 extra parents, at roughly $320 each. - Against an LTV of $120, each marginal dollar returned about $0.38.
Tripling to $270k would push further up that cost curve. Even if the marginal cost stayed at $320, which it will not: - The extra $180k a month would buy about 560 parents a month, or about 3,400 over six months. - Those parents would be worth about $400k in lifetime value, for about $1.08M of spend. That is roughly $680k of value destroyed. - With 22% six-month retention, perhaps 1,500–2,000 of them would still be booking at the raise.
The 50,000 target is the real issue. Reaching it means adding 13,000 net parents in six months, a 35% gain. Last year we added 6,000 in twelve months, a 20% gain. No channel closes that gap by the raise: - Our current pace of about 500 net parents a month gets us to roughly 40,000. - Tripling paid might add 1,500–2,000. - The squad work below might add a similar amount. - My honest range is 41,000–43,000.
I would rather we take a credible story to Series B investors. That story is: a strengthening organic engine, rising retention, and channel-level unit economics we can defend. A burned-through paid spike will show up in their diligence.
My recommendations on paid: - Bring spend back to about $60k a month, where the cost per parent was $82. - Run a geo holdout to measure how many paid parents would have found us organically anyway. - Redirect the $30k saved, and the $180k not spent, into the work below. Some of it could fund first-lesson guarantees for new tutors.
3. Where the three squads should go
Every squad works within one guardrail. No incentive may be tied to leaving a review, directly or indirectly. Legal should review every test that touches the post-lesson flow. Review requests and referral or reward prompts should never appear together.
Squad 1: Search click-through recovery (Loop A, top of funnel)
Test: - Split-test the results-page and profile templates across matched keyword groups within the top 200. - Revert the March changes on half. - On the other half, test titles, meta descriptions, and review structured data.
Threshold: Click-through on the test groups recovers to 5.8% or better within eight weeks, with no loss of rankings. If it does, roll out to all keywords.
Stop condition: No lift of at least 0.3 points after eight weeks, or ranking losses on the test groups.
Why first: This is the cheapest win we have. Recovering the 18% fall is worth roughly a fifth more clicks on our best keywords, from traffic we already rank for.
Squad 2: New-tutor first booking (Loop A, the review bottleneck, plus fixing Loop C)
Test: - Give unreviewed but qualified tutors a "new tutor" slot in matching and in broad-search results. - Back their first lesson with a satisfaction guarantee: a refund or a free rebook if the parent is unhappy. - Gate the $50 tutor referral bonus to subjects and regions where bookings per tutor are above median, and pause it where we are oversupplied.
Threshold: - The share of new profiles booked within 60 days rises from 39% to 55%. - The share reaching three reviews within 90 days rises by 50% relative.
Guardrails: - Parent 30-day rebooking rate on lessons with new tutors stays within 2 points of baseline. - Total bookings do not fall.
Stop condition: No movement after two monthly tutor cohorts, or a breach of the parent rebooking guardrail.
Squad 3: Parent referral (Loop B)
Test: - Weeks 1–2: reconcile the invite and attribution data. This tells us whether the loop is under-attributed (good news) or the invite counts are inflated. - Then test referral prompts at natural moments, such as a lesson milestone or a rebooking. These prompts must be separate from any review request. - Test a two-sided offer: $20 for the referrer and $20 off the referred parent's first lesson.
Threshold: - Invites per active parent rise from 0.31 to 0.45 a month. - Invite conversion holds at 14% or better. - Cost per referred booking parent stays under $60, against an LTV of $275.
Stop condition: - Cost per referred booking parent goes above $90. - Fraud or self-referral exceeds 5% of rewards. - No lift in invite rate after six weeks.
What I am asking the exec team to decide
- Keep paid at about $60k a month and run an incrementality holdout, rather than tripling it.
- Replace the 50,000 headline with a Series B narrative built on growth in organic and referral share, six-month retention, and bookings per tutor. Our honest forecast is 41,000–43,000 parents.
- Assign the three squads as above, with a checkpoint at the offsite plus eight weeks.
I am happy to walk through the arithmetic in detail before the offsite.
Claims the judge couldn’t find in the brief
- Applied to 37,000 parents, those invite rates imply about 1,600 referred parents a month. Our attribution says referrals are 9% of new parents, which is closer to 190 a month.
Grades and run details
Decision model 83 · LLM judge 7 of 13 checks
Decision model checks
- passUses the supplied evidence correctly15%
- passAddresses the actual decision97%
- passRespects explicit constraints31%
- passIdentifies material uncertainty92%
- partialAvoids unsupported claims45%
- passProduces the required deliverable89%
- passFinds the decay behind the growth54%
- passSees the cross-side effect100%
- passAnswers the CEO on paid100%
- partialA closed loop, not a channel36%
- partialThe loop maths holds32%
- partialProposes tests that could fail24%
Run
- Run
- #1
- Time to output
- 71 s
- Submitted
- 2 Oct 2026