AI for Resort Activity Booking: Filling Every Slot at the Right Price

The Most Perishable Inventory on Property Is the Least Managed

Every resort GM knows the discipline of room revenue management. Rates move daily, pace reports arrive every morning, and an RMS quietly reprices the house against demand. Now walk down to the marina, the golf shop, the spa desk, or the kids club and ask a simple question: what did an unsold 10 a.m. snorkel-boat seat cost us last Tuesday, and what did we do about it?

At most properties, the honest answer is nothing and nobody knows. Activity inventory — tee times, spa slots, cabana rentals, ski-school sessions, guided excursions — is more perishable than a guest room. A room that doesn't sell tonight can at least anchor a discounted two-night package tomorrow. A 2 p.m. paddleboard lesson that departs half-empty is revenue destroyed at 2:01 p.m., along with the fixed labor, fuel, and equipment cost the resort paid to stage it anyway.

The market around this inventory is exploding. Arival and Phocuswright now size the global experiences market at $271 billion in 2025, headed to $342 billion by 2029 — an 8% compound growth rate that has made tours, activities, and attractions the fastest-growing segment of the entire travel economy, ahead of hotels, air, short-term rentals, and car rental for the first time. Online booking penetration for experiences is forecast to climb from 33% in 2025 to 42% by 2029, growing faster than any other travel product.

And yet pricing practice has not caught up with market reality. Roughly 70% of experience operators still price statically — one rate card, set before the season, untouched by weather, occupancy, pace, or events. Arival's research shows only 3% of large attractions and 12% of enterprise attractions price dynamically today, though both figures are expected to triple or quadruple in the coming years. For resorts, that lag is an opening: the property that yield-manages its activity inventory the way it yield-manages rooms captures margin its comp set doesn't know exists.

A vacant guest room is a lost night. A half-empty catamaran is a lost night plus the crew, the fuel, and the dock fees you paid to sail it anyway. Perishability with a fixed cost of goods is exactly the problem yield management was invented to solve — and resorts already own the playbook.

Why Activity Slots Are Harder Than Hotel Rooms — and Why That's the Opportunity

It's tempting to assume room-style revenue management maps directly onto activities. It doesn't, and understanding the differences is what separates a pricing engine that works from one that quietly breaks guest trust.

Capacity is granular and heterogeneous. A resort sells one product type upstairs — room nights, segmented by category. Downstairs it sells dozens: an 18-hole tee sheet in 8-minute increments, six massage rooms in 50-minute blocks, twelve cabanas, four departure times on a dive boat with eight seats each. Each has its own capacity curve, marginal cost, and demand elasticity. Spreadsheets cannot price this; algorithms can.

Demand is weather-dependent to a degree rooms never are. Guests sleep inside regardless of the forecast. They do not snorkel in a squall or ski in rain. Weather forecasts 5–7 days out swing last-minute leisure bookings by as much as 40%, and the swing cuts both ways: a sunny Saturday forecast fills the tee sheet by Wednesday, while an approaching front empties the beach and floods the spa. Static pricing treats those two Saturdays identically. AI does not.

The booking window is compressed and mobile. Room nights book weeks or months out. Experiences book late: roughly 38% of activity bookings happen the same day or within two days of the activity, 48% happen after the traveler has arrived in destination, and more than 80% of same-day bookings are made on a phone. The guest deciding what to do tomorrow is standing in your lobby tonight, holding the booking device in their hand. Whether they book your excursion or a third-party operator's depends on whether you reach them first with the right offer at the right price.

The guest is already acquired. This is the structural advantage. An OTA-dependent tour operator pays 20–30% commission to reach a stranger. Your activity demand walks through the front door pre-paid, pre-profiled, and on property for a known number of nights. Every dollar of activity revenue you capture from an in-house guest is nearly acquisition-cost-free — which is why properties that systematize the capture see 2.1x more ancillary revenue per guest than those that leave it to the concierge desk and chance.

From Rate Card to Algorithm: The Three Pricing Maturity Stages

Most resorts don't need to leap from a laminated price list to full algorithmic pricing in one season — and shouldn't. Arival's guidance to operators is "walk before you run": introduce variable pricing rules first, then graduate to true dynamic pricing as data and guest communication mature. The three stages look like this:

The revenue physics are the same ones that transformed rooms twenty years ago. During high-demand windows, price captures willingness to pay. During soft windows, price stimulates volume that would otherwise evaporate — and in activities, stimulated volume carries an extra payoff rooms don't have: the half-empty boat was sailing anyway, so every incremental seat sold at any price above marginal cost is nearly pure contribution. The broader market has noticed: the attraction dynamic pricing software category alone is growing from $2.8 billion in 2025 to a projected $4.9 billion by 2030.

The Demand Signals an Activity Pricing Engine Actually Uses

What makes activity pricing an AI problem rather than a spreadsheet problem is the number of live signals that move demand — and how differently they move it for each activity type. A competent engine ingests, at minimum:

Forecasting on combined signals like these is mature technology now — hotel-side AI models routinely claim 95% forecast accuracy, and visitor-forecasting systems that blend weather, holidays, and events report up to 97% accuracy on footfall predictions. The practical planning horizon is two to six weeks, which conveniently matches the activity booking window almost exactly.

Weather deserves special emphasis because it is the signal resorts systematically underuse. The forecast doesn't just predict demand — it creates it. When the sunny-weekend forecast lands on Wednesday, the resort that reprices its Saturday tee sheet and beach inventory that afternoon captures the surge; the resort running last season's rate card gives it away. And when the front rolls in, the same engine becomes a recovery tool: automated rebooking of the cancelled catamaran into spa slots and indoor experiences turns a refund queue into retained revenue. Weather-aware scheduling also flows into labor — the same demand forecast that prices the slot staffs it, so instructors and therapists are rostered against predicted utilization rather than habit.

Where the Money Is: Pricing Levers by Activity Type

Not all activity inventory responds to the same lever. The matrix below is how we frame the portfolio at a full-service resort — every revenue center on one page, with its perishability profile and its primary AI play:

The guest deciding what to do tomorrow is standing in your lobby tonight with the booking device in their hand. Forty-eight percent of experience bookings happen in destination — the only question is whether they happen with you or with the operator down the beach.

The In-Destination Window: Winning the Booking That Happens on Property

The single most actionable fact in activity revenue is the shape of the booking curve. It is nothing like the rooms curve:

Guests wait deliberately: they want to see the weather, gauge their energy, and decide in the moment. That behavior punishes resorts that treat activity sales as a pre-arrival email and a binder at the concierge desk — and rewards resorts that treat the stay itself as a live sales channel. The mechanics matter: mobile-first booking flows with real-time availability, integrated payment against the room folio, and offers that arrive through the channel the guest is already using. Every step between "I want to do that" and "booked" leaks conversion, and the leak flows to whichever third-party operator made booking easier than you did.

Cross-Sell Intelligence: The Recommendation Layer

Dynamic pricing fills slots at the right price. The recommendation layer decides which guest sees which slot — and it is where the compounding happens, because activity data is the richest behavioral data a resort owns. The guest who booked the sunrise snorkel tells you things the reservation never did: early riser, ocean-oriented, experience-spender, probably receptive to the sunset sail and unlikely to want the 7 a.m. spin class.

In practice, the resort recommendation engine works three moments of the journey. At booking, the confirmation page and pre-arrival flow surface two or three activities matched to party composition, stay dates, and season — a family of four in July sees kids club and the dolphin excursion, not the couples massage. Pre-arrival offers like these convert at 8–15%, versus near-zero when upselling is left to chance during the stay. In stay, the engine reacts to behavior: a booked activity triggers a complementary suggestion, a rained-out excursion triggers an instant spa alternative, an empty tomorrow-morning tee sheet finds the golfers in the house tonight. Post-activity, high-satisfaction moments become the trigger for the next sale — the guest walking off the boat grinning is the easiest rebooking on property.

Two design rules keep this welcome rather than intrusive. Relevance over volume: three well-matched suggestions per stay outperform a daily promotional blast and preserve the brand. And recommendations personalize the offer, never the price — a principle we'll return to below, because it is the difference between yield management and a fairness scandal. Done well, the numbers are consistent across the industry: digital recommendation and upsell platforms drive 2.1x the ancillary revenue per guest of front-desk-only selling, and hotels that prioritize ancillary streams overall report materially higher profitability than those that don't. For a deeper treatment of the pre-arrival window specifically, see our research on the AI-orchestrated pre-arrival experience.

A 90-Day Implementation Path

The good news for resort operators: this does not require a data science team. It requires clean booking data, a modern activity booking platform, and a phased rollout that builds guest-communication muscle before algorithmic complexity. The path we recommend:

The measurement discipline matters as much as the technology. The metric that changes behavior is RevPAS — revenue per available slot — the activity-inventory analog of RevPAR, tracked alongside utilization and attach rate (the share of room bookings that add at least one activity). Once department heads see the marina and the spa reported the way rooms are reported, the conversation shifts from "how busy were we?" to "what did we earn per unit of capacity we staged?" — and pricing decisions start compounding. Resorts that want a structured, owner-grade approach to this — demand modeling, pricing architecture, and the measurement framework — often start with our AI Revenue Optimization & Forecasting service, which builds exactly this operating system across rooms and activity revenue centers.

The Fairness Line: Pricing Slots Without Breaking Trust

One caution, because it is the difference between sophistication and a TripAdvisor incident. Dynamic pricing prices the slot — the 9 a.m. Saturday tee time versus the 2 p.m. Tuesday one. It never prices the person. Two guests booking the same slot at the same moment must see the same rate, always. Personalization belongs in the offer layer (who is shown what) and the packaging layer (what is bundled for whom), never in the base price. The airlines spent decades teaching travelers that seat prices move with demand; guests accept slot-based pricing when it is transparent and consistent, framed as early-booking value rather than surge penalty — "book early for the best rate" reads as fair, "prices doubled because it's sunny" reads as gouging even when they are mechanically identical. Guardrails — rate floors and ceilings per activity, maximum daily movement, blackout on mid-transaction changes — are not constraints on the algorithm. They are what make the algorithm trustworthy enough to run unattended, which is the entire point.

Frequently Asked Questions