New floors, less revenue

I can produce more analysis in a week than I used to produce in a year. You probably can too now.

That is not the same as knowing more. I came close to publishing a proforma that proved it.

I was evaluating a hotel opportunity. 12 months of financials across more than 100 expense lines. Historical performance data going back to the property's prior brand. Market rates and market share pulled across 7 different views. 13 competing properties researched on room type, ratings, and rate structure. Work that would have taken a team weeks, finished in hours.

And I was buried.

More capability did not give me fewer questions. It gave me more of them, faster than I could resolve them.

I changed the numbers that follow for confidentiality, but they are directionally correct.

Hotels measure themselves against the hotels they compete with using an index where 100 means parity. The subject hotel runs an occupancy index near 150. It runs 1.5 times the occupancy of the competing hotels. Revenue per available room sits near 90, slightly below the group. Average daily rate sits near 60, well below it.

It runs full and it runs cheap.

To an underwriter that pattern reads as opportunity. The product is stronger than the rate suggests, so there is room to raise rate toward what the competing hotels charge and capture the difference. Bigger rooms, full kitchens, a brand guests recognize. Every piece of analysis I had supported that thesis.

None of it was wrong. All of it was old.

So I did the one thing I know how to do that the analysis could not do for me. I went to the buildings.

I toured the competing hotels in person, on a weekday. At one property the housekeeping staff had propped the room doors open while they cleaned.

The floors were new. Luxury vinyl plank (LVP), wall to wall. The last time I stood in that building, those rooms had rusted carpet.

At a second hotel I found a man touching up paint on the exterior. I asked him about the refresh. He was proud of it. He told me half the hotel had been renovated, new flooring in all of those rooms, and he walked me through the rest of what they had done.

I sat in the parking lot checking the Google profile for both properties. No LVP. Then I started doing arithmetic.

Flooring is the tell. Nobody replaces flooring on its own. New plank means the room was refreshed around it, and a refreshed room competes directly for the rate premium I was about to underwrite.

Two things follow from those floors, and both of them cost money.

Either the subject hotel is worth less than I assumed, because the competing hotels have closed part of the quality gap I was pricing. Or that gap can still be held, but only with capital I had not underwritten, spent maintaining and upgrading rooms to stay ahead of a group that is now moving.

In the first case the rate assumption comes down. In the second the capital plan goes up.

Either way the asset is worth less than my proforma said.

Had I published that proforma the week before, I would have underwritten a premium that was already being competed away, with 100 pages of accurate analysis sitting behind it.

Here is the part that should bother you. Those floors were not a secret. In a few months they show up in updated photos, in reviews mentioning the new rooms, in the Google listing I was staring at. A phone call to the front desk would have surfaced them that day.

The information was not hidden. It was newer than the record.

AI works on the record it is given. The record lags the building. No amount of processing speed closes a gap that opens before the data exists.

The lag cuts both ways. In the same market a General Manager told me how busy they had been after an event that just concluded. The subject hotel captured none of it. That is revenue sitting on the table, invisible in every report I had. How much of it can actually be captured is a judgment call.

Even then, the thesis was not clear.

I had the field work and I had the analysis, and what I was holding was a pile of accurate conclusions pointing in different directions. That is what unorganized capability looks like. Volume of data, without a soul.

So I stopped working. I stepped away from the screen, then I got out a whiteboard and wrote the thesis by hand. What the opportunity actually is. What the risks are and how each one gets mitigated. What we would execute, and in what order.

Then I took a photograph of the whiteboard, handed it back to the machine, and the plan came together in an afternoon.

That sequence matters. The analysis did not produce the thesis. The thesis made the analysis useful.

I also had to keep checking its work. One model concluded that rising costs would compress flow-through to 62%. That is a defensible answer. It is what most operators are living through right now, and the model was reasoning from that general pattern.

The asset was not the general pattern. When I made it test the claim against the actual financials, cost per occupied room had fallen 8%.

That is the dangerous kind of error, because it made the deal look more conservative rather than less. A number that makes you cautious rarely gets challenged. It would have sat in the plan unquestioned and priced the opportunity too low.

None of that is unique to hotels. Here is why it matters if you are building a platform.

You should be AI-first and AI-native. That does not mean AI everything. It means the leverage is real, and refusing it is a decision to move slower than the people you compete with.

But be clear about what it does to your edge.

AI has made parts of my work harder, not easier, because it hands the same leverage to everyone I compete against. The operator across town now runs the same analysis I run, at the same speed, for the same money.

You will not see that from your desk. Everyone's reports look better and are more accurate than they used to be.

So the tool cannot be the edge. The tool is table stakes now.

AI has never built a portfolio, led a team through a bad quarter, raised capital, or grown a platform. I have. You have. That experience is what tells the machine where to point, which of its answers to trust, and when the number on the screen does not match the building.

So go look. Walk the buildings you compete with this quarter, not the report about them. Look at the floors. Talk to whoever is holding a paintbrush.

Then write your own thesis before you ask a machine what it thinks.

AI is great. It cannot build the platform.

You can.

-Damon

Damon C. Healey, Founder, Eternal Companies I help proven real estate operators build the institutional platform that makes capital come to them. If you want to pressure-test your platform against an institutional standard, that is what a Platform Edge Session is built for. Book a Platform Edge Session | Get the 2026 IC Stress Test​

P.S. If someone on your team is building the model this week, forward them this issue before they finish.

​