Audit of the aggregator and shared-support research

Independent reviewer, live-verified against primary sources. Arithmetic is
sound throughout — every spot-check passed. The errors are attribution
and comparability errors, which are harder to spot and more dangerous.

3.1 Confirmed correct

FareHarbor 2% API fee (operator-borne) · Airbnb Experiences 20%, host-side ·
GetYourGuide ≤1.0% cancellation / ≤0.2% no-show, 90-day rolling · TourRadar fee
table including per-booking currency fees · BikeTours.com 20%, US$500 deposit,
US$30 booking fee · RACV Bike Assist tiers ($47 down to $37.60, unlimited callouts,
$50 taxi cap, Victoria only) · Correos Paq Mochila operational spec (from 08:00,
before 14:30, 15 kg, 60 km/stage, 20:00 booking deadline, nine languages).

3.2 CRITICAL — pull or rewrite before modelling

C-1. FareHarbor 6%: the conclusion is the reverse of the cited source.
It concludes "a new Australian operator would pay it, and would grow out of
it… Australia is in scope." FareHarbor's article states three cumulative
conditions; only two are quoted, and the third is dropped:

"The business has been using FareHarbor for at least 365 days as of October 3, 2024"

That is a fixed historical cut-off — an operator signing up in 2026 cannot
satisfy it and so would not pay the 6%. Separately, nothing in the source
says Australia is in scope; the only geographies named are US/Puerto Rico
(billing) and Guam/French Polynesia (API carve-out). "Australia is in scope" is an
unsourced assertion inside a [VERIFIED] block.
Net effect is inverted: direct bookings are free; OTA bookings attract the 2%
API fee, which is operator-borne.

C-2. The Camino "€6 vs €12 proves load factor" finding does not survive.
Caminofácil's own page:

"el precio por mochila y etapa puede estar entre los 9 euros (1 etapa del Camino
Francés) hasta los 6 euros, si la cantidad de mochilas y etapas es larga."

Every headline figure is a "Desde" (from) price. The €6 is a volume-discount
floor for one customer booking many bags across many stages — not the Francés
rate. Like-for-like single stage is €9 against the Norte's €12: ~1.33×, not
2×. And the discount mechanism is one customer's basket size, not strangers
sharing a van — which is the opposite of the thesis it was used to support. The
table already contradicts itself (price column "€6–9/stage", notes
column "Norte €12").

C-3. The Escapegoat price pair attributes a guide and a road transfer to "the
vehicle". A$75–275pp is called "the value of the vehicle". The two products
are not the same thing minus a van:

The delta is a full-day private guide plus an Adelaide transfer plus extra
tastings plus 2–5 extra hours, in a different location. The follow-on sentence
"validates the A$189pp from both sides" inherits the error.

C-4. The Wine Australia cellar-door grant does not apply to the target partners,
and is suspended. Eligibility requires ≥A$1,207,000 (GST-excl.) in rebatable
wine sales, and only sales above that threshold count — which excludes
essentially every small Barossa cellar door. It is not open for applications in
2026 (Round 7 closed 31 Oct 2025), and is paused for 2028-29 and 2029-30,
inside any realistic plan horizon. The pitch "every case we carry qualifies for
your cellar-door grant" would be wrong for most partners and wrong for two of the
next four years.

C-5. Two of the ten stated Gaps are falsified by the same research.

• "No response-time SLA exists anywhere in the world" — the same research records
  ADFC Pannenhilfe: mobile repair within a maximum of 120 minutes, plus Correos
  "before 14:30", Trail Journeys "11:00–16:00", Travel-Lite "collected by 9:00am".
  The gap is used to justify "publishing one would make this the first operator
  anywhere" — that claim is false.
• "Nobody publishes vehicle capacity… exactly four disclosures" — at least
  seven are quoted: Michigan C2C (55 riders), MTB Rotorua (5/9/24),
  Adrenalin (16-seat), Buller (5), Tastes of the Hunter (11/20/24), Share Shuttle
  (5/13), Escapegoat (8/11).

3.3 The load-factor inference — a leap, and the research rebuts itself

Even before C-2 removes half the swing, at least six alternative explanations
survive, several evidenced on the operators' own sites:

1. Basket-size discount, not route density. Caminofácil discounts on "la
   cantidad de mochilas y etapas" in one order; Xacotrans offers "discounts for
   10+ bags same origin/destination". Both are volume ladders inside one
   customer's booking — the opposite of pooling strangers.
2. Stage distance is priced explicitly — Correos caps at 60 km/day, Xacotrans at
   30 km with surcharges. Norte stages are longer.
3. Road access and terrain — Brigantes quotes "may vary based on route,
   mileage, and road access"; Xacotrans prices O Cebreiro→Sarria by
   consultation, i.e. the hardest stage on the busiest route costs more.
   Density alone cannot explain that.
4. The counter-example is already here, and ignored. Sherpa Van
   prices the Pennine Way "by remoteness rather than distance" — £40 vs £20
   on adjacent segments of a single trail with one demand profile. A 2× swing
   with zero load-factor variation.
5. Competition, not cost. The Francés has 6+ operators; the Norte a handful.
   €12 may simply be uncontested. Commercially this inverts the reading: filling
   the van does not automatically let you charge less — competitive pressure
   does.
6. Deadheading on thin routes is route geometry, not per-bag sharing.

And the quantitative objection points the other way. The Francés carries
roughly an order of magnitude more pilgrims than the Norte. If load factor
dominated, the gap would be far larger than ~1.33×. A shallow price response to
a steep density difference is evidence that most of the cost is distance, driver
hours and fuel — costs that do not shrink as the van fills. For a business whose
thesis is "sharing the van collapses unit cost", this is the finding that should
have been drawn, and it is materially less encouraging than the one written.

Related — do not model on the Jacotrans €6 → €50 step. Source is a Gronze forum
complaint thread; the €6 baseline is unverifiable (the Jacotrans row itself says
the operator publishes no prices); neither end is confirmed. It currently carries
the label "the single most direct warning".

And Brigantes "£35 / £18" does not show the second customer halving unit cost —
Brigantes prices 1 bag vs 2+ bags within one booking. A couple with two
bags, not two unrelated parties. Same per-order volume discount as Caminofácil.

3.4 Numbers that should carry no figure at all

• Viator 20% / 20–40% — 2018 third-party, 8 years old, nothing verifiable on
  Viator property. Should appear with no number.
• GetYourGuide 20–30% — TrekkSoft 2019, seven years old, from a
  competitor's marketing blog, and unfalsifiable because GYG refuses to publish.
  The model's entire downside case (contribution $124→$107, break-even 6.1→7.1
  riders) is built on this figure.
• Correos €4/stage and €5.50–7/stage — third-party aggregators; €4 undated.
  Correos blog comments (user-generated) are presented alongside verified data.
• Hunter Valley A$150–300pp — unattributed consumer blog.
• Airbnb's 20% is live-confirmed, but its corroboration (Checkfront 2020,
  Tourpreneur 2020) is six years old and predates Airbnb rebuilding Experiences —
  it adds false confidence, not support.

3.5 Cross-chapter collisions

• X-1 TourRadar. The Distribution chapter says "not published;
  per-contract"; the aggregator research says "15.25% marketplace"
  [VERIFIED]. The latter is right — but it
  contradicts itself within four lines, noting 15.25% is "not necessarily
  universal" (TourRadar: "check your contract") while the table grades it flat
  [VERIFIED] and the pull-quote asserts "roughly 18.7% all-in" unhedged.
• X-2 TrailHopper — the most dangerous item in the audit. The conclusion drawn is
  that the customer is "being asked to pay roughly double", calling TrailHopper "a
  sharper threat… for half the price" — comparing TrailHopper's transport-only
  tier (A$89.95–99.95) against the full product. The same table lists TrailHopper's
  all-inclusive at A$189.95 (4 wineries, lunch, pre-booked tastings) against a
  A$189 product. Like-for-like it is the same price, not half — a sharper
  competitive threat on features, and not cheaper at all.
• X-3 "Exactly four" capacity disclosures is stated in both the Van
  Economics chapter and the aggregator research — but over two disjoint
  datasets with completely different members. The number was copied, not
  recounted, and both counts are wrong (see C-5).
• X-4 Break-even 6.1→7.1 is a legitimate derivation (checks out against
  the $755 fixed and $189−$25−$57=$107), but is presented as a relayed finding
  rather than a new calculation — making a guess from a 2019 blog look measured.

3.6 Moderate

• M-1 Airbnb. Host-side 20% is now well supported. The unasked question: homes
  carry both a host fee and a 14.1–16.5% guest service fee, with no
  guest-side experience fee documented either way. If one exists, all-in take
  exceeds 20% and the "$38 ≈ $40 CAC" equivalence breaks. Retire the old gap and
  replace it with this question.
• M-2 RACV Bike Assist. Tyre repair is limited to metropolitan Melbourne;
  outside metro the remedy is the capped $50 taxi. The "RACV has already
  commoditised rescue" counter-argument is therefore much weaker in a regional
  wine district — the only place this product operates. The conclusion is right;
  the reasoning should be corrected, not relied on as written.
• M-3 The take-rate stack is understated. A 20% OTA commission sits on top of
  the booking-software fees documented alongside it (Rezdy 3%, Checkfront 3%, Bókun
  1–1.5%) plus payment processing. Real all-in on an OTA booking is 2–3 points
  higher than modelled, before the downside case.
• M-4 "Correos Express… that is the whole insight" is not in the cited press
  release, which never mentions Correos Express. Asserted inside a [VERIFIED] block.
• M-5 "The full stack" (operator 20% → BikeTours ~15% → agent 5% → traveller
  €25/$30) only occurs with an agent; BikeTours keeps the full ~20% on direct
  bookings. The site quotes US$30; the €25 was not on the page.

3.7 Minor

• N-1 Correos season stated as "Easter–31 Oct"; the press release says the
  service activated 16 March 2026, three weeks before Easter, and states no end
  date. "Sarria→Santiago and Tui→Santiago year-round" unverified.
• N-2 BikeTours "2+ years trading and insolvency insurance" is not on the
  cited page.
• N-3 Caminofácil row internally inconsistent (€6–9 vs €12).
• N-4 Xacotrans caveats dropped: 30 km stage cap, off-route surcharge,
  O Cebreiro→Sarria by consultation (relevant to §3.3 above).
• N-5 Method note says FareHarbor's "and TourRadar's" tables came via Zendesk
  JSON; TourRadar's help centre is Intercom and fetches over plain HTTP.
• N-6 All arithmetic spot-checks passed: Escapegoat 44% · MTB Rotorua 62%/83% ·
  Adrenalin £43.13/94% · IRRC $3.89/13–22× · WHW 18.75% vs Brigantes 48.6% ·
  Maydena 15.3 · TourRadar 18.7% · break-even 6.1→7.1.