Usable with a quick edit8 of 12 checks passed
The strategy is actionable, evidence-based, and framed for the board; a PM could add a numeric threshold to the Month 3 gate and a brief moat argument with a quick edit without redoing any reasoning.
Output
Tutorly: 12-Month Strategy Board paper, for decision
The problem Bookings have grown 8% while tutor supply grew 60%. We are not short of tutors or students. We are failing to connect them. Students say their top problem is finding the right tutor. Tutors say they lack students, and 70% get no bookings in a typical month. Once a student books twice, 64% keep booking, so the product works when a match happens. The leak is in the first match.
Why not paid student acquisition The board's instinct is to buy more students. I recommend against it for now: - Paid students would arrive at the same broken search and matching flow that already frustrates them. We would be paying to acquire customers we are likely to lose before their second booking. - Our tutor base is already larger than our demand. More students would mainly bring more tutors into the same competition for the few bookings available. - Acquisition spend scales with cost. Matching improvements scale with use.
The bet: fix first-match conversion (£1.2m) Over 12 months we will build a guided matching system and test it against the current search:
- Structured intake (Months 1–4). Students give subject, exam board, year group, goals, budget, and availability. Required before search.
- Curated shortlists (Months 2–6). Within one hour, students receive three ranked tutors with a free 15-minute intro call. A small concierge team handles hard cases such as niche subjects and last-minute exam prep.
- Ranking that rewards fit and responsiveness (Months 3–8). Ranking weighs subject expertise, response time, and student outcomes. New tutors get guaranteed early exposure so that supply does not stall at the bottom.
- Tutor activation (Months 4–12). Tutors with zero bookings in 30 days receive coaching on profile quality and availability, plus targeted introductions.
Budget (£1.2m) - Product and engineering: £600k - Concierge matching team: £300k - Tutor activation and targeted student tests: £200k - Measurement and contingency: £100k
Targets (12 months) - Bookings: 4,000 → 5,500 per month - Students with a first booking within 7 days of intake: baseline in Month 1, target +50% - Share of tutors with at least one booking per month: ~30% → 45% - Repeat rate (second booking): hold at or above 64%
Decision gates - Month 3: Intake and shortlist live for a test cohort. Confirm the conversion baseline and the uplift. - Month 6: If first-match conversion has not improved by at least 20% against control, stop further build and move remaining funds to a small, controlled paid-acquisition test. If it has, scale the concierge model and begin targeted acquisition in segments where our supply is deepest. - Month 9: Review unit economics. Each booking should contribute enough commission to cover concierge cost.
Runway This plan spends £1.2m over 12 months and leaves about six months of runway for the gates above. If the Month 6 gate fails, we retain the option to pivot without running out of cash.
What we are asking the board to approve 1. Fund the £1.2m matching-led plan. 2. Approve the Month 6 gate and the reallocation rule if it fails. 3. Defer paid student acquisition until conversion data supports it.
The board's instinct is right that growth needs more students. But we already have more tutors than we can match, and the evidence points to the match as the bottleneck. Fixing that first makes every future pound of acquisition work harder.
Grades and run details
Decision model 71 · LLM judge 85
Decision model checks
- failUses the supplied evidence correctly18%
- passAddresses the actual decision100%
- passRespects explicit constraints35%
- passIdentifies material uncertainty100%
- failAvoids unsupported claims34%
- passProduces the required deliverable89%
- passSurfaces the central contradiction100%
- passMakes a clear choice100%
- passDiagnosis before prescription100%
- partialProposes tests that could fail48%
- failNames an advantage that's hard to copy46%
- passSays what it gives up100%
Run
- Run
- #1
- API response time
- 16 s
- Submitted
- 8 Oct 2026