Start with the graveyard.
In-room tablets. Robot butlers. Chatbots in 2017. Blockchain loyalty. Lobbies in the metaverse. Every one of them arrived with a keynote and a vendor explaining that the properties who moved now would own the decade. The general manager who believed all of it was maximally open to change and set six years of capital expenditure on fire. He wasn’t a laggard. He moved. He moved into a ditch.
Notice who fills that ditch, by the way. Not the people counting down to five o’clock. The ones reading every trade publication and attending every conference. Enthusiasm is not analysis. It is just what analysis is most often mistaken for.
So being first is not a virtue. Anyone who has been in this business twenty years can produce a list of the things he was smart enough not to buy, and he is right to be proud of it.
You Can’t Buy Back 1998
On March 15, 1998, Americas Great Resorts published a warning about online travel agencies. The sentence was: today’s helpful partner can become tomorrow’s dominant gatekeeper. We named the merchant model. We said the intermediary would end up controlling pricing, presentation, and the customer data produced by every transaction, and that once the guest relationship moved, it would not move back easily.
The industry called it progress. The original is still posted.
Twenty-eight years later the commission is not a line item. It is the structure. For a great many independent hotels the discovery relationship belongs to Expedia and Booking. The repeat guest is frequently re-acquired through the same channel that delivered the first stay. The hotel supplies the room.
Here is the part almost everybody gets backwards.
The hotels that lost that fight were not the technophobes. They were the early adopters. They signed early, they filled rooms, and they were congratulated at conferences for being ahead of the curve. They did precisely what the innovation literature tells you to do.
Then they let the channel stand in for the relationship. Adoption was never the mistake. Substitution was. The channel was rented. The relationship would have been theirs.
So both ends of the spectrum lost in 1998. The refusers lost slowly and quietly. The enthusiasts lost while being applauded. Speed was never the variable.
One question
Take a housekeeping team that has worked together for eight years. They know each other’s shortcuts. They know which rooms turn slowly and which supervisor to ask. That team shows up in your flow-through every month, and there is no amount of capital that produces it by Thursday.
You can hire an experienced housekeeper. You can hire an experienced general manager. You cannot hire eight years of the same people learning each other. The events had to happen, in order, one at a time. Money buys forty things at once. It cannot buy a sequence that has not occurred yet.
Which gives you the whole test, blunt enough to use on a vendor in a Tuesday meeting. Could you erase your competitor’s lead by spending more, faster?
If yes, being late costs money. Annoying, survivable, finished.
If no, being late costs something that is not for sale.
Most of what gets sold to hotels as urgent is the first kind. A website rebuild is the first kind. So is most paid media. Paid search can be enormously profitable, but there is no first-mover advantage in buying tomorrow’s click today. You pay, demand arrives, you stop paying, it stops. A hotel that sat out three years of paid search can start buying tomorrow’s demand tomorrow. It does not have to go back and recover three years of missed clicks.
A hotel that sat out three years of building its own customer relationships cannot buy three years of customer history tomorrow. Or ever.
Same budget line. Two completely different kinds of late.
Which brings us to this year
Hotels are buying artificial intelligence. Agentic booking, AI revenue management, an assistant on the website, a concierge that answers at two in the morning. Some of it is genuinely good. What comes in the box is the purchasable kind. Your competitor can license the same stack next quarter, and a ninety day head start on a subscription does not compound.
Meanwhile, whether a machine recommends your hotel when a traveler asks where to stay is not something anybody sells you. It comes out of what already exists about the property: who published it, how long it has been consistent, how much of it you control, and whether the machine has any reason to treat your version as the reliable one. That accumulates. It accumulates in sequence. And it is hardening right now, while the industry shops for tools.
Same shape as 1998. The tool is the OTA contract. The thing that compounds is what nobody had time for.
The part that has nothing to do with intelligence
A general manager working on a shorter clock than the investment he is approving puts the cost on his own P&L and the benefit on his successor’s. Nobody was ever fired for running the same stack as the property next door.
1998 was not a failure of intelligence. It was the perfectly rational decision to fill rooms now instead of building something for later. The commission felt manageable at the time. It always does.
That is a scoreboard problem, and the same scoreboard kills these things twice.
Anything that accumulates in sequence looks terrible at the start. You spend first. The compounding arrives later. Which means month nine of a program that is working and month nine of a program that is failing photograph exactly the same.
So it gets killed. And killing it does not take you back to the starting line. It takes you back to the starting line with the money already gone. That is how compounding programs earn a reputation for not working. Maybe they killed something real. Maybe they killed a dog. Nobody ever found out.
Which is why anything worth starting gets its ending written down first. What would count as progress, when you check, and what you do if the answer comes back yes, no, or still muddy. Without that, “we’re piloting it” is just refusal wearing a lab coat.
Late is not finished
Most people reading this are already late to something that compounds. Fine.
The lead may never close. Your own accumulation still begins, and it still belongs to you. Permanently second in something that compounds beats permanently absent from it by a wide margin.
That notice went out in 1998. The industry filed it. The filing runs 15-25% of a booking.
The dumbest response to being three years late is spending a fourth year explaining why waiting made sense.
You can buy the software later. You can hire the vendor later. You can find the budget later.
You cannot buy 2026 later.
