Why $270 Billion Still Runs on Spreadsheets

The global in-destination experiences market, covering tours, activities, attractions and events, is approximately $270 billion in size as of 2024, and is forecast to reach approximately $342 billion by 2029 at a compound annual growth rate of roughly 8%.^¹ By travel-industry standards this is a large, well-documented and visibly growing category. By the standards of any comparably sized category in which digital systems have had twenty years to take hold, it is extraordinarily under-digitised.

The headline figures are well-rehearsed within the industry but worth stating plainly. Approximately 33% of the industry’s bookings flow through online channels, against roughly 64% for broader travel.^² Among operators themselves, around 39% still run without any vendor booking system, with another small share on in-house tools; their availability calendars, reservation lists and post-booking communications live across a mix of email, spreadsheets, paper diaries and telephone calls.^³ A separate Arival finding, attached to a 3,500-plus operator-response base, describes this as the industry’s “dirty secret.”^⁴ McKinsey, surveying the same operator layer for a different metric, puts the share of operational data that remains unstructured at 40–60%, a related readout of the same digitisation gap.^⁵

The reflexive explanation for this state of affairs is cultural. The experiences industry is said to be slow to adopt software because its operators are small, locally rooted and proudly analogue; because the people running these businesses are guides and curators rather than technologists; because the category has always lagged other travel verticals and will continue to lag them for the foreseeable future.

This framing is widespread, partially true, and ultimately the wrong level of analysis. The assertion that operators “just need to catch up” presupposes that there is a structurally similar rung they could step onto, that the software stack appropriate to a boutique hotel or a regional airline is, with minor adjustments, also appropriate to a ten-person kayak operator or a family-run vineyard tour. That presupposition does not survive a look at transaction volume: a mid-sized hotel will typically handle more bookings in a month than a typical tour operator handles in a year, and a regional airline will handle more in a day than a typical tour operator handles in a month. The same software stack does not face the same per-booking economics across those scales. The reason is simple: most software costs arrive as fixed or semi-fixed costs, while the operator’s ability to absorb them depends on booking volume.

The more useful reframe is to ask a quieter but harder question: what is the unit economics of enterprise software adoption for a tour operator, and how does it compare with the unit economics that drove digitisation in hotels and airlines? Hannes Werthner and Stefan Klein, in their 1999 study of information technology and tourism, and Dimitrios Buhalis, in eTourism, both treat tourism as an information-intensive sector: one in which the product itself is shaped by, delivered through, and marketed by information flows.^⁶ On this description, tours and activities are at least as information-intensive as hotels or flights; arguably more so, because the product is more heterogeneous and less standardisable. If information intensity alone drove digitisation, the experiences category would be as digitised as the adjacent categories, or more so.

It is not. The reason, on closer inspection, is that the economic surface area over which software costs can be amortised is fundamentally different. Across the operator population, approximately 86% of businesses run below $250,000 in annual revenue, and approximately 50% have been operating for less than five years.^⁷ At those scales, the per-booking cost of a conventional booking-system, channel-manager and revenue-management stack is not a marginal operating expense; it is a proportionally significant fraction of total revenue. The software decision is therefore not a question of willingness. It is a question of whether the software exists in a form whose unit economics work at the operator’s scale.

A worked sketch makes this concrete. Consider an operator typical of the long tail: roughly two hundred thousand dollars in annual revenue, around four thousand bookings a year at an average ticket value of fifty dollars, two employees and a founder-operator. A booking-system and channel-connectivity stack that costs five to eight thousand dollars a year before implementation, configuration and ongoing administration is not a rounding error. It is two to four percent of revenue and a meaningful share of operating margin. The same annual software cost, deployed by a larger attraction or a hundred-room hotel handling tens of thousands of bookings a year, amortises across an order of magnitude more transactions. The software is not the variable. The denominator is.

When this frame is applied, the 33% digitisation figure stops being a puzzle of cultural lag and starts being a predictable equilibrium. Operators that can make the economics work, such as higher-volume attractions, larger multi-site tour companies, and the upper tail of the operator distribution, appear from the same operator data to digitise at rates closer to the rest of travel, even if the cohort cut is not always made cleanly in surveys. Operators below a threshold that the industry has not cleanly quantified, but which hovers somewhere around the long-tail line, adopt patchier, cheaper, often consumer-grade tools that they do not themselves describe as “booking systems” in operator surveys. There is also a measurement artefact at work: a spreadsheet used systematically, a shared messaging channel used as a dispatch tool, or an off-the-shelf calendar tool wired into a payment provider by the operator themselves can all be functionally digital operations, even if they do not register as such in an Arival or Phocuswright response. The cultural-lag reading double-counts the gap.

The important consequence of this reframe is that the usual remedies do not follow from it. If digitisation is a cultural matter, the remedy is patience: eventually operators will modernise. If digitisation is a unit-economic matter, the remedy is structural. Either the software has to become an order of magnitude cheaper per booking (which is bounded by the underlying cost of building and maintaining it), or it has to be bundled with a distribution mechanism that carries the cost (which introduces its own distortions in distribution economics), or the economic frame itself has to change. The direction in which current indicators point is the third.

Myskiv and Nycz-Wojtan, in their 2022 treatment of reservation systems in small tourism enterprises, observe that in the SME operator segment, booking systems are often adopted primarily as marketing surfaces rather than as operations tools.^⁸ The booking form on a website is the public-facing reason for the software; the internal use is patchier and weaker. This is not a failing of the operator. It is a rational response to a system whose per-booking cost is higher than its per-booking operational value. The marketing surface provides a return; the operational depth, at the operator’s scale, often does not.

Nine years ago, Phocuswright’s Global Travel Activities report made the same observation in a slightly different register. Most operators were either not digitised or were “minimally” digitised, and the structural reason was the mismatch between the cost of enterprise-grade tooling and the scale at which most operators run.^⁹ The 2017 reading was, by the standards of the time, cautious. The 2026 reading is less cautious on the scale, since the market is half again as large and growing faster than broader travel, but arrives at structurally the same description of the operator layer.^¹⁰ That nine-year continuity is the most important single fact about the digitisation question in this industry. It is not that the catch-up has been slower than expected. It is that the catch-up frame was the wrong frame to hold the question in to begin with.

What, then, is the right frame? A category whose unit economics do not support conventional enterprise software adoption will either remain under-digitised indefinitely, or will digitise through some mechanism whose cost is not borne one-operator-at-a-time. Two candidate mechanisms, each of which moves the cost off the operator’s per-booking budget but by a different route, have been visible in the industry for close to a decade. The first is distribution-bundled software, in which a distribution partner absorbs the fixed cost of the stack in exchange for the downstream booking flow: the model that the industry’s major online travel agents pursued through a cluster of reservation-technology acquisitions in 2018. The second is infrastructure-layer software, in which a neutral layer provides the functions below the operator-level stack and removes them from the operator’s budget line altogether.

Of the two, only the first has been materially tested at scale, and the test is instructive. The distribution-bundled model was initially built on a zero-operator-fee pricing structure, with economic return captured through consumer-side booking fees: a design that made commercial sense if adoption scaled fast enough for the distribution flow to cover the cost of maintaining the platform. That original design is beginning to shift. Connectivity fees have been introduced in selected regions by OTA-owned reservation-technology providers, and the public pricing profile of the category has started to resemble that of a conventional paid-for software product. That shift is, in itself, testament to one of two underlying realities: either adoption did not scale to the volume the original economics required, or the unit economics of distribution capture alone did not carry the platform at the scale the acquirers had modelled.

There is a second structural issue alongside the unit economics. A reservation system owned by a distribution partner is, at the margin, optimised to route bookings through that partner’s channels, which is not the same thing as a reservation system whose commercial interests are aligned with the operator’s. Operators who adopt OTA-owned reservation technology have, whether the choice is framed this way or not, made a distribution-channel decision at the same time. The distribution-bundled model therefore resolves the unit-economic problem by transferring rather than removing it, and introduces an incentive structure that does not align neatly with the operator’s own economics.

None of this is an indictment of the model. It remains live. But the proposition that the distribution-bundled approach has cleanly solved the unit-economic problem, and that the industry should expect to converge on it, is harder to defend in 2026 than it would have been in 2018.

The cultural-lag reading would have the industry wait. The unit-economic reading does not. If the gap is economic rather than attitudinal, it will close only when the economic frame changes: when the cost of the layer operators need is no longer carried one operator at a time. Distribution-bundled software changed that frame in one direction. Whether an infrastructure layer can change it in another, and in what form, is the question the next essays in this sequence take up.