As AI Shopping Agents Shorten The Booking Window, Tourism Marketers Are Rethinking What They Measure

A media plan is a bundle of assumptions about audience, channel, and timing, and in tourism those assumptions are going stale faster than most reporting cycles can catch. A Hotel Online review of industry figures found that 34% of US travelers finalized bookings within two weeks of travel in the third quarter of 2025, up from 29% a year earlier. Sentiment in an inbound market can turn within days, and platform dashboards keep reporting activity through all of it. The marketers who adapt fastest decide before launch what normal looks like and choose a short list of signals that would show a departure from it. They also assign a person to interpret those signals before anyone moves the budget. Without that discipline, a team ends up reacting quickly to whichever number is easiest to see.

Spencer Gibboney, Associate Digital Media Director at The Zimmerman Agency, leads multi-channel media strategy for a portfolio that includes convention and visitors bureaus and individual hotels. He previously directed insights and analytics at Clarivoy, where he led development of a multi-touch attribution platform for automotive and consumer clients. He began his career in integrated media planning on Ohio Tourism and Safelite AutoGlass campaigns. That mix of measurement experience and hospitality clients gives him a close view of how booking behavior is changing, and of what marketers lose when the numbers run the plan.

“Blindly following the numbers on things like ROAS, or even worse, click-throughs or view-throughs, leads to a lot of wasted effort and missed opportunity,” Gibboney says. Metrics like return on ad spend record what a campaign did inside the platform, and a hotel’s business depends on what happens after the click. For his tourism clients, the earliest warnings tend to arrive in conversations with customers and in CRM data, well before a dashboard shows a problem, and his approach to media planning starts from that gap.

Define normal first

Gibboney looks for a model of contingency planning in an unlikely place. “Waffle House is the patron saint of contingency planning,” he says, pointing to the Waffle House Index. Craig Fugate devised it in 2004 while leading Florida’s emergency management and kept using it after he became FEMA administrator. A full menu suggests minimal damage nearby, a limited menu suggests the area may lack power or water, and a closed restaurant points to unsafe conditions or severe destruction. The index works because responders know what a normal day looks like and have chosen in advance which signal to watch.

For media teams, Gibboney says, the lesson is to “have a plan, know what normal looks like, and be able to define those signals that matter.” He recently applied that logic to a tourism client that draws some of its visitors from Canada. Social feedback from that market began showing resistance, which left the team weighing whether to keep marketing where sentiment had cooled or shift budget toward stronger markets. “You listen to what the market is telling you, and you listen to what the consumer is telling you,” he says. The habit comes from his parents, who told him to “just shut up and listen to people.” In media planning, that means treating customer sentiment as a signal in its own right, with the same standing as any metric on the dashboard.

Guided by data, led by people

“I cringe when people say we’re data-driven,” Gibboney says. The alternative he prefers is a line he credits to Jason Packer, a Columbus data analyst: “It should be guided by data, led by human intelligence.” The difference shows up in how teams treat platform metrics. An algorithm left in charge of spend will chase ROAS, click-throughs, or view-throughs because those are the numbers it can see. Gibboney argues that the wasted effort benefits only the large ad platforms, a pattern he calls “mal-optimization.”

A person still reads the data, listens to what customers are saying, and translates both for the client in a form the client can act on. Some clients know exactly where they’re headed. Others believe they want one direction while customers point toward another, and the planner’s job is to carry that message back.

Earning the client’s trust is what makes that translation land, and Gibboney says the opening comes from how little time clients have for marketing. “We spend 100% of our days in marketing, but our clients are just a fraction of their time,” he says. His approach is to offer relief: “let me help you take this off your desk for a little while.”

Heads in beds

For hotels and destinations, the useful measure sits far from the click. “Don’t confuse site performance or vanity metrics with heads in beds,” Gibboney says. He asks hoteliers to go deeper into their CRM and see whether guests drive in from nearby or travel from out of state, and whether a stay is a one-time trip or a repeat weekend. He also wants to know whether the guest booked a street-side room or a deluxe oceanfront one. Each answer changes how the property should speak to that guest next.

His years at Clarivoy gave him a vivid example of reading customers. One visitor to an automotive client’s site came back nearly 150 times, Gibboney recalls, always looking at performance cars, and then bought a minivan. “He just told you a story. He’s a family man,” he says he told the dealer, adding that the details belonged in the customer’s file because the same buyer might return in 15 or 20 years for a very different car.

Once a guest is in the loop, Gibboney sees the front desk as part of the media plan. In another CMO Wire interview, SG-retail’s Tim Fairs says a shopper’s memory of in-store service can outweigh everything else about a visit. Gibboney tells hoteliers that a booking is only the midpoint, since retention marketing continues through the stay and after the guest leaves. A staff member who asks at checkout whether the guest had a great trip and would like to book the same time next year can turn a visit into a repeat booking. He says clients who tried it have seen results. The post-stay survey is one more chance to reach back out.

Agents shorten the booking window

Gibboney sees AI as a reason the narrowing window will last. Economic pressure started the change, he says, but travelers have since learned they can ask ChatGPT, Claude, or another assistant to build an AI shopping agent that watches hotel rates. Hoteliers check their numbers daily and may lower the average daily rate to fill rooms. A traveler’s agent can wait for that drop and book five, seven, or ten days out instead of three or six months ahead. The same agent can check 20 online travel agencies in an instant, work that used to mean cruising through Expedia and its rivals by hand. “Even if everything turned around tomorrow and it was rainbows and unicorns, this is now a learned behavior. People are going to wait,” he tells his hotel clients.

He expects the pattern to spread. “It is fundamentally changing tourism, and I can see that it will probably spread into other industries,” Gibboney says. Automotive retail is his first example, since pre-owned dealers generally don’t want inventory approaching 90 days old, and shoppers can build agents to find the best price on a particular vehicle.

He compares the moment to the early days of social media. Car dealers he worked with resisted it, and he told them they had to participate anyway. “You have to be there because the conversation’s going on, whether you’re there or not, about your dealership.” Bad reviews carried a lesson too. Sometimes, he told them, “thank God for a bad review, because that shows a chance for conflict resolution.” A pile of them signals something systemic, he adds.

Gibboney expects AI to follow the course social media took, with conversations about a brand happening in places the brand doesn’t control. When asked what else marketers should take from the conversation, he went back to Packer’s advice and added his own. “Be quiet, and listen to what the customer is telling you. They will tell you their story,” he says.