SatuSatu — AI & Platform Opportunity: Executive Decision Memo

Date: 2026-07-27 · For: CEO, CFO · Source: research steps 00–05, red-teamed at 04-opportunity-register.md §6 · Decision date: 2026-08-14 (GATE 3)

1. The thesis, in one sentence

We have built a precise financial case on a business nobody has counted, so the only defensible commitment this quarter is one person-day of measurement, one contracting conversation that avoids 8–29 engineering-weeks, and about two to three engineering-weeks of deterministic fixes that are right whatever the measurement says.

2. The three bets, and the money at stake

Every dollar below is R × parameter, where R = 1,000 bookings/month is an assumption adopted for comparability, not an observation. The plausible range for the real business is 200–400 bookings/month. The gap between those columns is the memo.

CURRENT basis — cash on the business that exists today. Recurring, annual.

TARGET basis — B2B. Zero partners today. $0 until the exclusivity check passes and a first partner transacts. Never added to the table above.

Bet 1 — Measure, then decide. Cost: hours.

Two measurements gate the whole file. Count the actual monthly bookings, by 2026-08-03. Cross-check whether Pool B operators already self-list on GlobalTix or Klook — one day of catalogue work, by 2026-08-07. The second reaches 17 of 24 register rows, including ~76% of the "safe" zero-engineering lane, because exclusivity is a term inside the 27% Pool B margin, not merely a line in the pitch. GATE 3 on 2026-08-14 routes to one of three outcomes: fund the conditional B2B lane; fund only the unconditional lane; or rebuild around D0/D1 with no B2B at all.

Bet 2 — Contract the second aggregator on non-overlapping territory. Cost: zero engineering.

One clause beats twelve fixes. Integrating a second feed "properly" — dedup, canonical model, review queue, taxonomy crosswalk — costs 8–29 engineering-weeks we do not have. Contracting it for non-overlapping territory (non-Bali Indonesia, or Singapore) avoids all of that for one negotiation and ~0.5 person-weeks. It is independent of booking volume and of the exclusivity check, so it holds in every scenario. At 200 bookings/month its low case ($5,432, being 8 engineering-weeks at $679) exceeds the entire zero-engineering lane's floor ($3,968). If the negotiation fails, the correct answer is one feed — not a 20-week integration project.

Bet 3 — The unconditional deterministic slate. ~2.0–2.9 engineering-weeks.

Split the inherited supplier sold-count from our own booking count (a live advertising-accuracy exposure, not a margin argument); a nightly rate-integrity check; the sublicensable image-and-content clause in the Pool B contract template, which gets more expensive with every contract signed; refund terms rendered from the supplier field; query logging; a chargeback census; AI listing extraction bought, never built. None of it is an AI bet, and all of it survives either measurement outcome.

What it costs, and the one hire. Unconditional cash-equivalent is ~$2,700–$3,700 plus counsel. The engineering ask across both lanes is ~8–9 engineering-weeks against an assumed budget of 6 — and nobody has confirmed the 6. The real constraint is not engineering. It is business development. The plan carries ~17.7–19.4 non-engineering person-weeks, of which BD is ~12.8–13.5, and roughly 11 of those are one row: outreach to 20–40 named Indonesian agents to test whether Pool B sells by rate sheet at all. No file in this study records that SatuSatu has a BD function. The ask is plain: hire or borrow one BD FTE for a quarter, or drop the D2 chain. There is no cheaper substitute — only the decision to run the test or not.

One structural limit. Real-time availability at Bali's long-tail operators is not solved by anything in this roadmap, and the research concludes it will not exist by 2028. D2 and D3 are therefore capped at request-to-book for the entire horizon of this plan. Design for that, not around it.

3. What we are explicitly NOT doing

• A reseller API. Pool B cannot pass an industry-standard certification — Viator gates production on real-time availability, booking hold and a calendar view. Ours would be request-to-book: an API that adds latency to a workflow whose only defect is latency. Break-even needs ~20–40 partners each clearing ≥480 Pool B bookings/year.
• An AI concierge as a headline bet. Fully-loaded Indonesian concierge labour is $0.079/minute. Break-even handle time against a $0.99-per-resolution vendor is 12.5 minutes; safely automatable work is 2–3 minutes. It survives only as a conditional capacity play: if Pass volume is limited by concierge roster, one FTE serves ~1,068 passes/year and the case becomes GMV lift (~$7,760), not cost saving (~$1,422). That is unverified. It is a measurement in the 90-day plan, not a bet.
• Charging for the partner dashboard. The only precedent — KKday's rezio — ceilings at ~US$6.0M ARR on 5,000+ operators against a US$70M round. D2 is a channel: COGS, not a revenue line.
• Building connectivity. Middleware prices at ~5.7% of processed volume. Tripadvisor acquired Bókun on 2018-04-20 (acquirer, target and date confirmed against 02-competitor-matrix.md §4a; the "priced toward a fraction of a percent" claim is flagged unconfirmed in that same section and is not relied on here). The layer sells neutrality and is standardising on OCTO. Buy it; do not become it.
• Reusing TipTip's forecasting model. Concert ticketing is a different forecasting problem, and the +50% margin claim attached to it is correlational and single-source. Borrow the people and the process. Never the model.

4. The single biggest risk

A precisely-reasoned answer about a business nobody has measured.

At 200–400 bookings/month — a range consistent with 253 SKUs, one supplier feed, keyword-only search, no AI in the product and a three-month-old hero product — 13 of 24 rows cost less to ignore than to convene a meeting about. At 300/month exactly one row clears a $3,000/year materiality floor. At 150, none does. The failure mode is not that a row is wrong; it is that the machine is well built and bolted to an unmeasured base, then presented in dollars.

The market risk compounds it. Bali arrivals turned negative in 2026: −1.11% y/y Jan–Apr while Indonesia nationally ran +7.7% (BPS, verified). Europe — the high-spending, long-staying cohort our English-first USD storefront addresses — is −5.9% y/y into Indonesia in May 2026. Several CURRENT-basis rows sit on a shrinking base, and both terms of volume × average ticket value are under pressure at once.

Two withdrawals, stated so they are not quoted back at us: the $140,500 headline (it summed incompatible bases and double-counted a row) and the "103% of Pool A gross profit" cross-check (circular).

Appendix follows on a separate page. It is a summary wrapper only — SWOT was explicitly rejected as this study's analytical engine and no conclusion in this memo derives from it.

Appendix — SWOT

Summary of §1–§4 only. Nothing here is new analysis, and nothing above was derived from this frame.