Step 2 — Bali / Indonesia TAA Landscape

Subject: SatuSatu (satusatu.com), PT Tiptip Network Indonesia. 253-SKU Bali tours/activities/attractions platform. Inbound international traveller (USD, English-first), domestic Indonesians as secondary tier. Prepared: 2026-07-26. All web sources accessed 2026-07-26. Evidence tiers: A = official statistics agencies / filings / live product. B = credible trade press and analysts. C = blogs, forums, vendor marketing, market-research SEO pages. Labelling: every number is [VERIFIED] (stated in a collected source) or [INFERENCE] (derived here, with band).

Scope warning. This synthesis draws on the 19 raw search dumps assigned to Step 2, plus 14 additional M&A stub files in research/raw/ opened in a follow-up pass to rewrite Section 8. Where those files are thin or off-target, that is stated explicitly rather than papered over. Section 5 (Bali supply counts) retains a hard hole — no count of Bali TAA operators exists in any collected source. Section 8's remaining limitation is different in kind: the transactions are identified and dated, but six of the most relevant have undisclosed prices, so the margin conclusion is bounded by that (see §8.4).

1. Bali / Indonesia inbound arrivals

1.1 Volume and trend

Bali passed its 2019 foreign-arrival peak in 2024–25 and grew ~10% in 2025. Indonesia nationally did not pass 2019 in 2025 (15.39 m vs 16.11 m). Bali is carrying the national recovery.

1.2 2026 monthly run-rate — the trend has broken

Meanwhile Indonesia nationally, Jan–May 2026: 6,067,014 arrivals, +7.7% y/y [VERIFIED, A — BPS via Indonesia Investments].

This is the single most decision-relevant arrivals fact in the file: Bali is flat-to-down ~1.8% YTD 2026 while the country it sits in is up 7.7%. Bali is losing share of Indonesia's inbound growth. National growth is being driven by ASEAN short-haul into Jakarta/Batam, not by Bali.

1.3 Source-market mix

FY2025, Bali, top 10 [VERIFIED, A/B — BPS Bali via Bali Hotels Association and ANTARA]:

March 2026, English-first core [VERIFIED, A/B — BHA top-20 table from BPS/DISPARDA]: Australia 119,777 + UK 24,206 + USA 23,003 + New Zealand 10,017 + Canada 7,000 = 184,003 = 39.0% of the 472,070 monthly total [INFERENCE, arithmetic on verified components].

Direct read for SatuSatu: an English-first, USD-denominated storefront addresses roughly 39% of monthly inbound volume as native-English speakers, and Australia alone is a quarter of it. Australia is not a "long-haul Western tourist" market — it is a 6-hour, high-frequency, repeat-visit, price-sensitive market. India (8.2%) and China (7.7%) are the next two and are neither English-first nor USD-comfortable.

1.4 2026 full-year trajectory

Method: Jan–May was 38.06% of the 2025 year; Jan–May 2026 actual = 2,598,143. Base = 2,598,143 / 0.3806.

Conflicting forecast: an ARIMA model published on LinkedIn (Ross Woods, Tier C) projected 7.33 m for Bali 2026. That is above even my high case and was built before the Jan–Apr 2026 decline was known. I do not trust it. Government target for Indonesia is 16–17.6 m foreign arrivals in 2026 [VERIFIED, B — ANTARA, 13 Jan 2026]; Indonesia Investments projects 16.5 m [VERIFIED, B]. Both are national, not Bali.

2. Seasonality amplitude

Full 2025 monthly series. Eight months are directly reported; four (Aug, Sep, Nov, plus Jan/Feb cross-check) are reconstructed from BPS-reported month-on-month percentages. The reconstruction sums to 6,948,675 against the official 6,948,754 — a 79-visit gap, i.e. it is essentially exact.

Amplitude measures

Read for concierge staffing. A 1.55× peak-to-trough on arrivals is moderate by resort-destination standards — Bali is a year-round destination, not a summer one. Practical consequence: a human-concierge model sized for July is ~35% over-staffed in February, and one sized for February will fail in July. The swing is large enough to matter for a headcount-based cost line but small enough that a single flexed team (core + seasonal contractors for Jun–Sep) covers it. Any capacity-based business case should model Jun–Sep at ~38% of annual demand in four months, not a smooth twelfth-per-month.

Caveat that cuts the other way. Arrivals ≠ activity demand. Domestic Bali arrivals peak on a completely different calendar — Nyepi and Eid al-Fitr, which fell together in March 2026 and drove inter-province domestic arrivals to 1,060,798, +67.5% y/y [VERIFIED, A/B — BHA from BPS/DISPARDA]. A dual-tier product has two uncorrelated peaks, which flattens combined staffing need. That is a genuine argument for the domestic tier that has nothing to do with revenue.

Directional 2026 change: the 2026 shape is flatter and lower than 2025 — Jan–Mar 2026 all sat between 472k and 502k, a 6% band, against a 529k–470k range in 2025. Early 2026 has less amplitude and less volume.

3. TAA online penetration — Indonesia vs global

3.1 Global (the structural headroom)

Source: Phocuswright / Arival, "Travel Experiences 2026: Market Sizing, Operator and Consumer Behavior Highlights," published July 2026 (Tier B, sponsored by Civitatis; methodology is a bottom-up operator model from the Arival Global Operator Landscape 4th Ed., 5,664 qualified operator responses, plus a 4Q25 consumer survey of 800 US and 1,750 UK/FR/DE/ES travellers).

The headroom statement, stated precisely: experiences are at 33% online against 64% for travel overall — 2× less digitised. But the closing rate is only ~1.5–2 percentage points per year. Inside a 24-month window the online share moves from ~33% to ~36%. Offline does not collapse. Anyone building a business case on "offline is about to flip online" is wrong on the timeline by roughly a decade.

3.2 Indonesia specifically

No credible Indonesia-specific TAA online penetration figure exists in the collected sources. This is a real gap.

What the files contain instead is a set of mutually inconsistent, un-methodologised market-research SEO pages:

Three vendors, three numbers, ~2× spread, no shared definition. Do not use any of them. Indonesia TAA online penetration is unverified — not found in collected sources.

The defensible proxy is APAC-level: APAC is the fastest-growing region at 8%/yr and reaches 30% of the global experiences market by 2029 [VERIFIED, Tier C secondary of a Tier B report]. Indonesia's internet penetration (>212 m users, >75% smartphone penetration by 2025, per thereportcubes, Tier C) is not the binding constraint — supply-side digitisation is (Section 5).

4. 🔴 Where inbound Bali travellers actually book activities

Headline: there is no direct, Bali-wide measurement of tours-and-activities booking channel split in the collected sources. None. The most decision-relevant question in this step is the one with the weakest data. What follows is the structural evidence, labelled honestly.

4.1 What is actually measured

(a) Global walk-up and offline share — the single hardest number available. Phocuswright/Arival 4Q25 consumer survey (US + UK/FR/DE/ES), Tier B, [VERIFIED]:

• Fewer than half of US and European travellers said they booked a recent tour, activity or attraction online.
• Nearly 1 in 5 travellers under 55, and 24% of those 55+, booked as a walk-up at a provider or ticket office.
• Only 12–30% of experiences are booked same-day; older travellers do this far more.
• Rates of online booking are "especially low in the 55+ age group."

This is the cleanest evidence in the entire file that walk-up is a first-class channel in experiences and nowhere else in travel. It is US/EU, not Bali — but Bali's inbound mix is 39% Anglo (Section 1.3), so it transfers better than most global figures would.

(b) ASITA Bali's travel-agent claim — large, load-bearing, and internally broken. The Bali chapter of ASITA (Association of the Indonesian Tours and Travel Agencies) states, via chairman I Putu Winastra [VERIFIED as a claim, Tier B — The Bali Times]:

• 6.9 m tourists used travel agents to visit Bali in 2025 (from 6.3 m in 2024, 5.3 m in 2023, 2.1 m in 2022).
• "Travel agencies accounted for around 65 percent of international tourist arrivals to Bali in 2025."
• ~2 m arrived via travel agents Jan–Apr 2026.

These two claims cannot both be true. Bali's total direct foreign arrivals in 2025 were 6,948,754. 6.9 m is 99.3% of that, not 65%. Either the 6.9 m includes domestic arrivals, or it counts trip components rather than tourists, or the 65% is the accurate figure and the 6.9 m is inflated. It is a trade-association self-report with no published methodology.

My reading: treat the 65% figure as the usable one and even then as an upper bound with a wide error bar, and treat it as describing trip/arrival booking (flights, packages, transfers), not excursion booking. A traveller whose flight-and-hotel came through an agent may still buy their Nusa Penida day trip from a street desk. ASITA has an obvious institutional interest in a high number.

(c) Bali hotel-reservation channel studies — the only Bali-specific quantification, and it is about rooms, not activities. Two academic studies of individual Bali properties, Tier C (peer-reviewed but single-property and 2016–2019 vintage):

• The Westin Resort Nusa Dua: OTAs ~37% of room reservations (2018), ~39% (2019); offline travel agents ~24–25% [VERIFIED as reported].
• A second Bali hotel dataset (33,338 reservations 2016; 32,337 in 2017; 37,864 in 2018): concluded offline travel agents supplied ~60% of tourists at the studied property [VERIFIED as reported].

Those two findings disagree with each other by a factor of ~2.4 on offline share. They are hotel rooms. They pre-date COVID. They are two buildings.

(d) Driver commission mechanics. "If your driver books tickets for attractions… they quite likely get a commission on ticket sales, usually around 20% but sometimes more" — with Elephant Park and white-water rafting named as examples [VERIFIED as a claim, Tier C — baliholidaysecrets.com]. Corroborating soft signal: TripAdvisor Bali forum, "Drivers take you to places that most likely give them good commission" (Tier C). Private driver day rate: IDR 500,000–800,000 (~USD 30–55) for a full day including fuel [VERIFIED, Tier C — balibelin.com, ohana-agency.com, both 2026].

No source in the collected files quantifies what share of Bali activity bookings originate from driver referral. Not one. The 20% commission rate is documented; the volume is not.

4.2 Synthesised channel estimate

The booking-channels.md deep-research file already produced a proxy synthesis. I reproduce its bands because they are the only Bali-scoped estimate available, and I attach my own confidence rating — which is lower than the file's own.

All figures in this table are [INFERENCE] with low confidence. None is [VERIFIED].

4.3 What I actually believe, and why

Stripping out the false precision, four things are structurally supported:

1. Offline is the majority channel, or close to it. Globally, experiences are 33% online [VERIFIED, Tier B]. Bali's supply is more fragmented and less digitised than the US/EU average that figure is weighted toward. There is no mechanism by which Bali would be more digitised than the global mean. Offline share in Bali is therefore ≥ 60%, and plausibly 65–75% [INFERENCE].
2. The driver is a distribution channel with a 20% take rate and zero customer-acquisition cost. He is already in the car, already trusted, already going that direction, and he is paid the same rate a wholesaler gets (Section 6). He is not a nuisance to be routed around — he is the incumbent last-mile reseller.
3. Walk-up is real and specific to this vertical. ~19–24% of US/EU travellers walked up to a ticket office for their most recent experience [VERIFIED, Tier B]. That behaviour does not exist at this magnitude in air or lodging.
4. The competitor set is not Klook. For 60–75% of the market, the competitor is a man with a Toyota Avanza, a hotel front desk, and a laminated A4 sign on Jalan Legian.

Unverified and needed: Bali-specific booking-channel audit; pre-trip vs in-destination split; channel by traveller segment (Australian repeat visitor vs first-time European); channel by Bali sub-region (Canggu vs Ubud vs Nusa Dua); actual driver-referral volume. The booking-channels.md file lists the same gaps and recommends primary collection (OTA booking exports, operator POS receipts, structured inbound surveys, driver/concierge interviews). I concur — this is the highest-value primary research SatuSatu could commission, and it is cheap.

5. Supply fragmentation

5.1 Global operator structure

Arival Global Operator Landscape (3rd Ed. Oct 2024, 7,000+ operators; 4th Ed. Jan 2026, 5,664–7,000+ operators) and Phocuswright/Arival 2026. Tier B; several figures reach me through automate.travel (Tier C secondary of a Tier B primary) — flagged where so.

Phocuswright/Arival's own methodology note is the honest framing: "This market is highly fragmented with hundreds of thousands of mostly small, private operators… Across many categories, no definitive registry of businesses exists."

5.2 Bali-specific operator counts — hard gap

No count of Bali tours/activities/attractions operators exists in the collected sources, and no measure of what share of them run any booking system.

What is in the files:

• ASITA Bali publishes a full-member directory at asitabali.org, listing licensed inbound/outbound travel agencies with DPP registration numbers, contacts and target markets [VERIFIED as existing, Tier A — primary association source]. The search result captured only the first ~10 alphabetical entries; the total member count was never retrieved. This is directly obtainable and should be in Step 3.
• ASITA Bali is travel agencies, not activity operators. It does not cover Mount Batur trekking guides, Nusa Penida boat operators, Ubud cooking classes, or waterfall entrance concessions — which is exactly SatuSatu's Pool B target.
• Klook's platform-level figures give scale context but not Bali counts: ~310,000 offerings across ~4,200 destinations as of 30 Sep 2025 [VERIFIED, Tier A — SEC F-1].
• Global platform listing growth: 385,000 (2020) → 986,000 (2026), +156% [VERIFIED as a claim, Tier C — marketintelo, no methodology; treat as directional only].

Inference from structure, not from a Bali count: if >70% of global operators are small/micro and 58% of small operators have no booking system, and Bali's long tail is at or below the global digitisation mean, then the majority of SatuSatu's Pool B addressable supply has no booking system of any kind — not a bad one, none. This is not a temporary condition that a channel-manager integration fixes. It is the market. Pool B's absent real-time availability is a property of the supply base, not a SatuSatu implementation gap.

6. 🔴 Where the margin sits, and who captures it

6.1 The chain, layer by layer

Two things are hard here. Everything else is trade-press convention or vendor marketing.

Hard datum #1 — the OTA layer, from an SEC filing. Klook Technology Limited, Form F-1, filed 10 Nov 2025 [VERIFIED, Tier A]:

Also [VERIFIED]: Klook revenue $417.1 m FY2024 (+24.4%); net loss $99.3 m FY2024; net loss $141.5 m on $407.4 m revenue for 9M 2025 (Fortune, 15 Jul 2026, citing the Nov 2025 SEC filing); 65 m experiences booked LTM Sep-2025.

Conflict flagged: mostlymetrics.com (Tier C) reports LTM-Sep-2025 net loss of −$17 m; Fortune (Tier B) reports −$141.5 m for 9M 2025 from the same filing. These are different periods and possibly different measures (the −$17 m may be an adjusted figure). I trust the Fortune/SEC 9M figure because it names the filing and the period. Do not cite the −$17 m.

Hard datum #2 — the operator net-rate convention. Arival's own guidance to operators: "If you are working with inbound tour operators, receptives or DMCs, then you would take approximately 25%–30% from your retail rate" [VERIFIED, Tier B — arival.travel]. Corroborated by Rezdy and Trekksoft (Tier C trade blogs, consistent with each other): retail travel agents 10–20% of retail; tour wholesalers 25–30%; inbound tour operators 25–30%.

6.2 The waterfall

On a USD 100 retail experience in Bali:

6.3 Where SatuSatu actually sits

Pool A (GlobalTix-sourced, non-exclusive). SatuSatu is buying at a third party's net price, which is itself already downstream of the operator and, in most cases, of a DMC. SatuSatu is layer 3 or 4 on a chain with 25–30 total points of spread between operator net and consumer retail.

Constraint that closes the argument: SatuSatu's Pool A inventory is non-exclusive, meaning the same SKU is retailed by Klook, Viator and GetYourGuide. SatuSatu therefore cannot price above them. It buys at GlobalTix net and must sell at market retail.

Anchor for the ceiling: Klook — 310,000 SKUs, $3.04 bn GTV, direct merchant contracting, its own payments rails, 13 m reviews, and a decade of scale — realises 11.2% gross profit on GTV [VERIFIED, Tier A]. A 253-SKU reseller buying at another aggregator's net price has no mechanism to beat that number. 11.2% is not SatuSatu's target; it is the asymptote SatuSatu cannot reach on Pool A.

Pool B (direct-contracted, exclusive, no real-time availability).

Basis: direct contracting at operator net rate, i.e. the full 25–30 points that Arival documents as the wholesaler/receptive convention, plus exclusivity supporting the upper end. Against this sits the manual-fulfilment cost of no real-time availability (concierge labour per booking, cancellation exposure, overbooking risk) — not quantified in the collected sources.

The structural fact for strategy: Pool B carries roughly 2.5–4× the gross margin of Pool A. SatuSatu's margin problem is not a pricing problem or a conversion problem. It is a sourcing problem. Every incremental point of GBV mix shifted from Pool A to Pool B is worth ~17–20 points of gross margin on that GBV.

And the offsetting fact: the 20% driver commission and the 15–32% agent commissions are drawn from the same 25–30-point pool. SatuSatu, a driver, and a street tour desk are all competing for the identical spread. The driver's cost to serve is zero and his conversion rate — captive audience, in-vehicle, at the moment of intent — is far above any website's. On Pool A economics SatuSatu cannot outbid a driver. On Pool B economics it can.

7. Domestic demand hedge

7.1 What the data supports

7.2 Conflict, and how to read it

Bali's two domestic series disagree in direction. Inter-province arrivals were −8.02% in 2025; national domestic trips were +18.95%. Both are BPS. They are not the same thing: "inter-province arrivals" counts entries into Bali Province from another province; "wisnus trips" counts all qualifying trips including intra-province day trips, which inflates the Bali figure to 26.6 m. Use the 5.70 m inter-province arrivals figure for any addressable-market calculation involving a traveller who needs an activity booked. The 26.6 m number is trip volume including Balinese residents moving around Bali and is not an addressable audience.

The direction conflict is the important part: Indonesians travelled far more in 2025, and fewer of them came to Bali. Bali lost domestic share while national domestic volume hit a record.

7.3 Sizing the hedge

Spending, and the reason volume overstates the opportunity.

• International tourist average spend in Indonesia, Q1 2026: US$1,345.61 per visit, average stay 10.84 nights [VERIFIED, A/B — BPS via Indonesia Expat, 5 May 2026]. April 2025 comparable: $1,286.84 over 9.74 nights [VERIFIED, Tier C — BPS via YouTube].
• Domestic spend structure: transport is the largest component, then accommodation, then food & beverage — Bank Indonesia's Indonesia Tourism Outlook 2025/2026 explicitly notes that the dominance of transport in the domestic spend structure "berimplikasi pada terbatasnya ruang belanja" — implies limited remaining discretionary spend [VERIFIED, Tier A — Bank Indonesia PDF].
• No per-head domestic activity/attraction spend figure for Bali exists in the collected sources. Unverified.

Verdict on the hedge. Domestic is 45% of Bali arrivals by head count and its peaks (Nyepi, Eid, school holidays) are uncorrelated with the Anglo Jun–Sep peak — which is genuinely valuable for concierge utilisation. But: BI's own data says domestic wallet is transport-dominated with little left over; Bali's domestic arrivals fell 8% in 2025 while the national market boomed; and domestic will not pay USD prices for a multi-day pass. Size it as a load-balancing and utilisation play, not a revenue hedge. Without a domestic activity-spend-per-head figure, any revenue projection for the domestic tier is unfounded.

8. Funding and M&A in TAA / travel-AI, last 12 months, SEA-weighted

Question this section answers: who is buying, who is being bought, at what multiples where disclosed — and whether the capital flow is consistent with this study's Section 6 finding that distribution is structurally thin (Klook: 11.2% gross profit on GTV) while curated or exclusive supply is where margin survives.

Verdict up front: the thesis holds on the distribution half and is directionally supported but not proven on the supply half. Connectivity middleware is demonstrably consolidating and being priced as infrastructure rather than as a margin pool — there is a hard multiple for this. Curated supply is consistently what gets bought and is consistently described as "curated" by its own acquirers — but prices are undisclosed in five of the most relevant transactions, so the margin claim rests on acquirer behaviour and language, not on disclosed economics. Section 8.4 states the counter-case plainly.

8.1 Transactions in the last 12 months (Jul 2025 – Jul 2026)

⚠️ Window note: the two Fever events (2025-06-04, 2025-06-05) fall ~13 months back and are therefore marginally outside a strict 12-month window. Retained because Fever is the only asset in the entire set that is both EBITDA-profitable and an owner of original supply, which makes it load-bearing for the thesis. Dated explicitly so the reader can discount it.

8.2 The one disclosed multiple in the whole set

Only one transaction in this file permits a multiple to be computed, and it is the one that matters most for the thesis.

Read this carefully, because it is the single sharpest datum in the section. A connectivity platform with "hundreds of integrations across both supply and demand," including exclusive relationships with leading live-entertainment ticketing systems and global theme parks, was folded into a combined group valued at under 6% of the annual GMV flowing through it.

Three caveats, all real: the $40 m is the combined entity's valuation, not Redeam's standalone price (which is undisclosed — PhocusWire says explicitly "for an undisclosed amount"); the $700 m is a forward projection by the acquirer, not audited volume; and EV/volume is not EV/revenue. This is a volume multiple, not a revenue multiple, and it is directional only. But the direction is unambiguous: the market did not pay for that GMV, because the middleware layer does not touch it.

Compare against Section 6: Klook, a demand aggregator, converts 11.2% of GTV into gross profit. A connectivity layer converts so little of the volume it carries into enterprise value that the whole company is worth 5.7% of one year's flow. Middleware is thinner than distribution, and distribution is already thin.

Corroborating, from the structural record: when Tripadvisor bought Bókun in April 2018 it announced it would move Bókun from "€100 per month, versus the industry standard of 5%-6% taken on online bookings" to "a fraction of a percent per booking, far under the industry standard" [VERIFIED, Tier A — Tripadvisor IR, 2018-04-20]. A demand aggregator bought a middleware business and immediately priced its take rate toward zero. That is not a company monetising an acquisition; that is a company buying a toll booth in order to demolish it. Booking Holdings announced FareHarbor one day earlier, on 2018-04-19 [VERIFIED, Tier A].

8.3 The pattern, by layer

(a) Demand aggregation — consolidating into the big OTAs, and the independents cannot get out.

• Tiqets went to Expedia Group at $279 m. Expedia's own framing is not consumer-facing at all: "a key step toward our vision to build the most comprehensive, global travel solution powered by our expanded APIs, across activities, air, car and insurance" — Alfonso Paredes, President of B2B [VERIFIED, Tier A]. Expedia bought an experiences platform to feed its B2B API, not its consumer brands.
• Klook filed 2025-11-10 and still has not priced 8+ months later [VERIFIED]. GetYourGuide is doing a share sale, not an IPO [VERIFIED, Tier B — TravelAge West]. Vitruvian's incremental $50 m into Civitatis in June 2024 was secondary shares bought from other shareholders — liquidity for existing holders, not primary growth capital [VERIFIED, Tier B — Skift].
• Three independent scaled experiences OTAs, three routes to liquidity, none of them a clean IPO.

(b) Connectivity middleware — rolling up, explicitly commoditising itself.

• Palisis + Prioticket → Experience Technology Group, 2026-05-12: combined 50,000+ experiences, 250+ API-integrated reseller partners, 60+ reservation systems integrated. Terms undisclosed. Both brands continue; contracts, integrations and contacts unchanged. Crucially, the companies "stress that they will maintain a neutral stance towards all distribution partners" [VERIFIED, Tier B — Travolution].
• Neutrality is the product, and neutrality is the opposite of pricing power. A layer whose entire pitch is that it will not favour anyone cannot extract from anyone.
• The standard is going open: LeisureKing connected to Prioticket via the OCTO specification, 14 July 2026, "giving its network of operators a single integration" [VERIFIED, Tier A — Prioticket company news]. Open connectivity standards are how a layer becomes a commodity.
• Expedition Software (Vertica Capital Partners) is the same roll-up pattern one layer over: Rezdy + Regiondo combined, then Checkfront acquired. Vertica's stated playbook is FareHarbor, which "grew from less than $5 million in funding to 30× revenue growth and over 70% market share before its acquisition by Booking.com" [VERIFIED, Tier A/B — verticacp.com]. Note the exit: FareHarbor's terminal buyer was a demand aggregator.
• ⚠️ Date conflict, unresolved. Three sources give three dates for the Rezdy/Checkfront/Regiondo combination: Bailador (ASX-listed investor, Tier A) says June 2023; Rezdy's own blog says 8 Aug 2023 and describes all three merging at once; Vertica says Rezdy+Regiondo in November 2023 with Checkfront as a separate add-on dated both "March 2024" (page header) and "January 2024" (body text, same page). I trust Bailador's June 2023 for the initial combination — it is a listed investor's portfolio disclosure with regulatory exposure — and treat Checkfront as a later add-on of uncertain date. Vertica's own page contradicting itself by two months is a caution on that source.
• GlobalTix — directly relevant, as SatuSatu's Pool A supplier. Series B of S$6.5 m, led by Tin Men Capital (first invested 2018), with Seeds Capital, ORZON Ventures (managed by 500 Thailand) and a US family office, announced 2024-10-14 [VERIFIED, Tier A — globaltix.com]. Company scale at the time: 150,000+ experiences, 12,000+ travel agents, 12 m+ tickets issued annually, 10 offices across Asia, and Tin Men states GlobalTix had become cash-flow positive [VERIFIED, Tier A].
  • ⚠️ Do not double-count. A PRNewswire release dated 2024-12-19 headlined "GlobalTix Secures $5 Million Series B Funding" is the same round restated in USD — identical lead investor, identical participants, identical round designation. S$6.5 m ≈ US$5 m. One round, two currencies, two press dates.
  • The signal: a business issuing 12 million tickets a year across ten countries raised roughly US$5 million at Series B. That is a very small round for that footprint. It is consistent with a capital-efficient, cash-generative, structurally low-margin middleware business — exactly what Section 6 predicts, and exactly why SatuSatu's Pool A cost base cannot improve.

(c) Curated / exclusive supply — this is what gets bought, and the acquirers say why.

The word curated is not mine. It appears in the acquirer's or investor's own framing in five separate transactions:

• Fever is the strongest single data point and also the most caveated. It raised >$100 m from L Catterton and Point72 on 2025-06-04 having achieved "over 20× revenue growth from pre-pandemic levels, while maintaining full-year EBITDA profitability" in 2024 [VERIFIED, Tier A]. It is the only asset in this entire set disclosed as profitable. It owns and produces original experiences rather than reselling third-party inventory. The day after the raise it acquired DICE (>10 m monthly active fans, ticket sales doubled in two years) [VERIFIED, Tier A]. Capital went to the owner of supply, and the owner of supply then bought the distributor — the exact inversion of the middleware pattern.
• Dusit Thani ← Elite Havens (full acquisition), ~September 2018 [VERIFIED as event, Tier A — dusit-international.com; date is [INFERENCE] from the release's own CDN asset path /storage/newsroom/news-release/2018/09/, as no date appears in the page text]. Bali-relevant and instructive: a Thai hotel group bought outright a concierge-led managed-inventory business running 200+ fully staffed luxury villas across Indonesia, Thailand, Sri Lanka and the Maldives. It did not buy a booking engine. It bought contracted supply plus the human layer that services it.
• Holibob raised a $12 m Series A led by Ryan Howsam (Staysure founder, an investor since 2020), plus an extendable credit facility, on the back of 600% revenue growth over six months and the acquisition of TourismSolved [VERIFIED as event, Tier A — holibob.tech]. Date unverified — the captured page carries no article date, only unrelated blog-sidebar dates. Positioning is explicitly the curation-and-context thesis: moving the industry "from being majority offline to majority online."

(d) Where the AI money actually went.

Every AI-labelled dollar in this set went to supply-side content and connectivity automation, not to consumer-facing agents:

• Headout ← Dabble: an acqui-hire of three computer-vision founders, no product transferred [VERIFIED, Tier A].
• Headout "Magic Listing": AI that generates operator listings from data already in their booking system — "No re-entering content, media, pricing, or policies" [VERIFIED, Tier A — Prioticket].
• GlobalTix: Series B proceeds directed at "AI and predictive analytics to optimise pricing, identify industry trends" [VERIFIED, Tier A].
• SPUN: AI agents for visa processing — B2B infrastructure, and explicitly priced at market rather than subsidised [VERIFIED, Tier B].

This independently corroborates Force 2 in Section 9: nobody with capital is funding an AI that books tours. They are funding AI that fixes the supply side's content and connectivity problem.

8.4 Does the consolidation thesis hold? — and where it does not

Holds, strongly:

1. Middleware is priced as infrastructure, not as a margin pool. One disclosed multiple: ~5.7% of processed volume [INFERENCE]. One documented instance of an acquirer buying a middleware business and immediately cutting its take rate to "a fraction of a percent" [VERIFIED, 2018]. One merger whose headline promise is neutrality [VERIFIED, 2026]. One open connectivity standard (OCTO) actively displacing bilateral integrations [VERIFIED, 2026].
2. Independent demand aggregators cannot exit cleanly. Klook unpriced after 8 months; GetYourGuide doing a secondary; Civitatis's last raise was a secondary purchase; Tiqets sold to a strategic. Consistent with 11.2% gross-profit-on-GTV being a hard, unattractive ceiling.
3. Direction of travel is uniform: curated supply is the acquired object. Tiqets, Civitatis, Elite Havens, DICE, Holibob's TourismSolved.

Does not hold, or is not yet demonstrated:

1. There are almost no prices. Palisis/Prioticket, Fever/DICE, Headout/Dabble, Bókun, FareHarbor, Redeam standalone — six transactions, all undisclosed. The margin claim for curated supply cannot be priced from this evidence. Tiqets at $279 m has no accompanying revenue or GBV disclosure in these files, so no multiple is computable — and $279 m for a platform in 60+ countries and 1,000+ cities is not self-evidently a premium.
2. Curated supply is bought; it never buys. Every acquirer in this set is a distributor, a middleware roll-up, a PE firm or a hotel group. Fever is the sole exception, and Fever is live events, not destination TAA — its economics derive from owning original IP and media reach, which is not the same business as contracting a Balinese waterfall operator. "Curated supply gets acquired" is equally consistent with curated supply being cheap and sub-scale as with it being high-margin. I cannot separate those two readings on this evidence.
3. Two of the five "curated" data points are stale or undated — Civitatis is June 2024; Holibob's Series A date is unverified.

Net: the evidence firmly supports "distribution and connectivity are thin and consolidating." It supports "curated supply is where margin survives" only as the most plausible reading of acquirer behaviour and language, not as a demonstrated fact. Anyone who tells you the M&A record proves curated supply carries the margin is over-reading six undisclosed prices.

8.5 What it means for SatuSatu

• Pool A's supplier is not going to get cheaper. GlobalTix is cash-flow positive on a ~US$5 m Series B while issuing 12 m tickets a year across 10 countries [VERIFIED]. It is capital-efficient, not capital-hungry, and therefore has no motive to buy volume with discounts. SatuSatu's Pool A cost base is set by a rational, profitable operator with no reason to concede margin to a 253-SKU reseller.
• Do not build to be the connectivity layer. That layer is consolidating into Experience Technology Group, Expedition Software and Travel Curious Group, is standardising on OCTO, sells neutrality, and has one observed valuation of ~5.7% of throughput. D3 (reseller API) must therefore be justified as a proof of exclusive supply, not as an infrastructure business — which is what Section 10 already concludes on independent grounds.
• The acquirer set for a business like SatuSatu is strategics, not the public market. Expedia bought Tiqets for its B2B API. Dusit bought Elite Havens for contracted Bali-region inventory plus a concierge layer. Both are precedents for what SatuSatu's Pool B plus human concierge could be worth to a buyer — and neither buyer wanted a catalogue of non-exclusive, third-party-sourced SKUs, which every one of them can already obtain from GlobalTix directly.
• Elite Havens is the closest structural analogue in the file: Bali-centred, concierge-led, exclusively managed inventory, bought whole by a hospitality strategic. Price undisclosed [VERIFIED as event; date [INFERENCE] ~Sep 2018]. It is a single, eight-year-old precedent and should be treated as a directional analogue, not a comparable.

9. Five structural forces most likely to reshape this economics inside 24 months

Ranked by expected impact on SatuSatu's unit economics by mid-2028.

Force 1 — Supply-side digitisation does not arrive. Pool B's availability gap is permanent.

Evidence: 39% of global operators run no booking system; 58% of small operators (<1,000 guests/yr) have none; 54% of operators founded after 2022 have none — the newest cohort is the least digitised; 300+ fragmented booking platforms with no dominant standard; average operator uses 5 non-integrated tools [all VERIFIED, Tier B via C secondary — Arival GOL 4th Ed.].

Why it matters: the post-2022 cohort figure is the tell. Digitisation is not diffusing downward over time — new small operators enter without systems and stay that way. Real-time availability for Bali's long tail will not exist by 2028. Any Pool B roadmap premised on "operators will get on channel managers" is premised on a trend that the data says is not happening. Plan for permanent manual fulfilment, and price it.

Direction: neutral-to-negative for cost; strongly positive for defensibility — an incumbent-proof moat, because Klook and Viator will not do manual fulfilment for a 30-guest-a-month waterfall operator either.

Force 2 — AI disintermediates discovery, not booking. A distribution/SEO problem on a 12–24 month clock.

The established finding, held precisely: as of 2026-07-26 no general AI assistant books tours end-to-end on its own rails.

Evidence [all VERIFIED]:

• OpenAI withdrew Instant Checkout in March 2026. Reported 5 Mar 2026; PhocusWire 10 Mar 2026: OpenAI "is shifting its Instant Checkout offering," refocusing on search, discovery and merchant connections and "ceding the transaction itself to external platforms." Expedia stock +12%, Booking Holdings +8% on the news (Tier B/C — uof.digital, PhocusWire, altexsoft).
• Klook is live on ChatGPT — and the flow ends at the door. Klook's own post: "just tap 'View on Klook' to open the full listing and complete your booking." Invoked via @klook handle after linking in the Apps/GPT Store (Tier A — Klook's own channels).
• Skyscanner launched a dedicated ChatGPT app for flight search, 27 Feb 2026, UK/US (Tier B — OAG).
• Sabre + PayPal + MindTrip built what OAG calls "the first end-to-end agentic booking pipeline," March 2026 — with OAG's own caveat: "The real question is whether consumers will trust an AI to select, book, and pay on their behalf (which they currently don't)." (Tier B).
• Stripe's hospitality lead: "Agentic commerce is very new, and we'll likely see several iterations over time as it matures." (Tier B — PhocusWire).
• Trust and accuracy remain unresolved: CNBC, 11 Mar 2026, documents hallucination and trust gaps; an agency example of ChatGPT routing a client into a 45-minute transfer that should have been 10 minutes, with the practitioner noting "They seem like they're edge cases, but they're actually very common" (Tier B).
• Klook's own 11,000-user survey claims 91% of global travellers use AI travel planners (Tier C — vendor survey, self-serving, treat as directional evidence of discovery behaviour only, not booking).

The correct framing. This is not an existential threat to the booking layer inside 24 months. It is a discovery-channel migration that is already underway. The consequence is concrete and boring: the referral traffic that used to arrive via Google organic will increasingly arrive — or fail to arrive — via LLM surfaces. Skift's Megatrends 2026 framing (via Klook): visibility now depends on being mentioned by ChatGPT, Gemini and Perplexity; LLMs are becoming the gatekeepers. Klook is a first-party ChatGPT app. SatuSatu, with 253 SKUs and no structured feed, is not in any model's retrieval set.

Action horizon: 12–24 months, and it is an AEO/structured-data/feed-distribution problem, not a product-architecture problem. Do not rebuild the booking flow for agents. Do make the catalogue machine-readable and get it into the surfaces that answer "what should I do in Bali."

Force 3 — Online penetration grinds up ~1.5–2 pts/year. Offline stays the majority through 2029.

Evidence: experiences 17% online (2019) → 33% (2025) → 42–43% (2029F), against 64% for travel overall [VERIFIED, Tier B — Phocuswright/Arival]. Walk-up remains ~19–24% of recent bookings among US/EU travellers [VERIFIED, Tier B].

Inside 24 months the online share of experiences moves from ~33% to ~36% [INFERENCE]. The offline majority is not going away on this planning horizon. This force is a reason for patience, not a reason for urgency — and it is the strongest argument against betting the company on a pure online-catalogue play (D1).

Force 4 — Channel consolidation compresses net rates. The layer SatuSatu sits in is being squeezed from both ends.

Evidence: OTA share of operator bookings 24% (2019) → 37% (2025); direct-website share 29% → 25% [VERIFIED]. Airbnb, Booking and Expedia expanding experiences divisions and acquiring specialist OTAs; Klook and MyRealTrip preparing IPOs; GetYourGuide planning a share sale [VERIFIED, Tier B]. 46% of operators are already unhappy with OTA commissions, and 31% cite "commission too high" as their reason for avoiding OTAs entirely — yet OTA dependency keeps growing [VERIFIED].

Consequence. As the big three fold experiences into their core funnels, aggregator net rates get renegotiated toward the buyer with volume. SatuSatu's Pool A cost base is set by GlobalTix, whose own cost base is set against Klook and Viator's volumes. A 253-SKU reseller has no negotiating position in that chain and will be the last to receive any improvement and the first to absorb any compression.

Force 5 — Bali's volume tailwind has stopped; the price tailwind is FX, not demand.

Evidence: Bali foreign arrivals −1.11% Jan–Apr 2026 and −1.77% Jan–May [VERIFIED / INFERENCE], while Indonesia nationally is +7.7% [VERIFIED]. Bali inter-province domestic arrivals −8.02% in 2025 while national domestic trips hit a record +18.95% [VERIFIED]. Bali star-hotel occupancy 60.88% in Dec 2025 vs 63.71% in Dec 2024 [VERIFIED as reported, Tier C — topbalihotels citing BPS]. April 2026 saw sea arrivals collapse −94.61% m/m through Benoa/Padang Bai/Celukan Bawang as cruise calls ended [VERIFIED, Tier B — Bali Discovery citing BPS Bali]. Europe is contracting: −5.9% y/y into Indonesia in May 2026, driven by Middle East tensions and jet fuel costs suppressing long-haul via Gulf hubs, with the specific note that European travellers are the high-spending, long-staying segment [VERIFIED, Tier A/B — BPS via Indonesia Investments].

The one offsetting force is currency: the rupiah's exchange-rate drop "has increased Indonesia's relative affordability for visitors from several of Bali's largest markets" [VERIFIED, Tier B — Migrant Times, 4 Jun 2026]; IDR ~17,966/USD [VERIFIED, Tier C — Britannica 2026].

Consequence. Growth must now come from share and attach rate, not from more arrivals. A weak rupiah widens the USD gross margin on IDR-denominated Pool B cost — which is a real, quantifiable tailwind for direct contracting and no help at all to Pool A, whose net prices are quoted by an aggregator.

10. SO WHAT FOR SATUSATU

1. Pool A is structurally uninvestable as a margin engine, and the ceiling is now a number rather than an opinion. Klook realises 11.2% gross profit on GTV with 310k SKUs, direct contracting and its own rails [VERIFIED, SEC]. SatuSatu's Pool A base case is ~10% gross / ~7% net of payment and FX [INFERENCE]. Consequence for D1 (horizontal catalog): D1 scales the lowest-margin pool. Adding 2,000 GlobalTix SKUs multiplies revenue at 7% net contribution while adding catalogue, content, and support cost linearly. D1 should be explicitly repriced as a traffic/SEO asset that feeds D0 and D2, not as a P&L line — or deferred.

2. The margin fix is a sourcing programme, not a product programme. Pool B carries 2.5–4× Pool A's gross margin (27% vs 10% base) [INFERENCE, anchored on Arival's 25–30% net-rate convention]. Consequence for D0: every roadmap item should be scored on whether it increases Pool B's share of GBV. A 10-point mix shift from A to B is worth more than doubling total volume on Pool A.

3. Stop treating Pool B's missing real-time availability as a defect to be engineered away — it is the moat. 58% of small operators have no booking system and the post-2022 cohort is worse at 54% [VERIFIED]. This will not resolve by 2028. Consequence for D0: the human concierge is not a temporary crutch pending an API; it is the fulfilment layer that Klook and Viator structurally will not build for a 30-guest-a-month operator. Budget it permanently, instrument its cost-per-booking, and market the confirmation SLA as the product.

4. The real competitor is the driver, and on Pool A economics SatuSatu cannot outbid him. Drivers take ~20% of ticket sales [VERIFIED as claim, Tier C] — the same slice a wholesaler gets — at zero CAC, with in-vehicle conversion at the moment of intent. Consequence for D2 (B2B partner dashboard): the highest-value D2 partner is not a hotel. It is the driver-and-guide network. A Pool B SKU at 27% gross margin can pay a driver 15–18% and still clear more than a Pool A SKU sold direct at 10%. That arithmetic only works on Pool B — which makes D2 and Pool B contracting a single joint programme, not two initiatives.

5. Commission the Bali booking-channel study before committing capital to any of D1/D2/D3. No Bali-specific TAA channel split exists anywhere in the collected evidence — the best available is a 40–70% offline band at low confidence, built from two pre-COVID single-hotel room-reservation studies that disagree by 2.4× [documented in Section 4]. Consequence: D2's sizing and D3's partner economics both depend on a number nobody has measured. Structured inbound-traveller intercepts, driver interviews, and operator POS pulls are cheap relative to building the wrong channel.

6. AI is a 12–24 month distribution deadline, not a product threat — and SatuSatu is currently invisible to it. Discovery is migrating to LLM surfaces; booking is not (OpenAI pulled Instant Checkout in March 2026; Klook's ChatGPT flow ends at "tap View on Klook") [VERIFIED]. Consequence: do not rebuild the booking flow for agents. Do make the 253 SKUs machine-readable — structured schema, a clean feed, entity-level content depth on the Pass and on each Pool B exclusive — and pursue placement in the assistant surfaces. A 253-SKU catalogue can plausibly rank in an LLM retrieval set for narrow Bali intents; it can never win generic search against Klook. Curation is the asset that survives this force; breadth is the asset that dies to it.

7. The volume tailwind is gone, which changes what the Pass has to do. Bali foreign arrivals are −1.8% YTD 2026 while Indonesia is +7.7%; domestic Bali arrivals fell 8% in 2025 [VERIFIED]. Consequence for D0: the Bali All-Access Pass must be justified on attach rate and basket size within a flat market, not on riding arrival growth. Its real advantage is now margin architecture, not demand capture — a bundle lets SatuSatu blend high-margin Pool B into a single USD price where the traveller cannot line-item-compare against Klook. Bundling is the only mechanism in the model that defeats the non-exclusivity price ceiling on Pool A. That deserves to be the central design constraint of the Pass, and the 90-day activation window should be evaluated against it.

8. Target the 39%, and size D3 against a consolidating buyer set. English-first markets are 39.0% of March 2026 arrivals, with Australia alone at 25.4% [VERIFIED/INFERENCE] — the USD, English-first positioning fits, but it is an Australian short-haul repeat market, not an American long-haul one, and India (8.2%) and China (7.7%) are addressed by neither the language nor the currency. Consequence for D3 (reseller API): the sector's capital events run through Airbnb/Booking/Expedia acquiring specialist OTAs while Klook's IPO has sat unpriced for eight months [VERIFIED]. D3's value is not as a revenue line at these margins — it is as proof of a distributable, exclusive Pool B inventory set, which is the only asset in the model that a consolidator cannot buy from GlobalTix directly. D3 should be built to be demonstrated, at Pool B depth, not to be scaled on Pool A breadth.

Open gaps (ranked for Step 3)