How to build a hotel projection that survives due diligence

A projection is easy to build and easy to puncture. When a lender, an investor or a buyer pulls one apart, the weak ones fall over in minutes. The job is to build one that does not.

Due diligence is not hostile, it is thorough. The people reviewing your numbers have seen hundreds of models and they know exactly where optimism hides. A projection that holds up is not the one with the best headline return. It is the one where every assumption is defensible, every number traces back to a source, and the downside has already been tested. Here is what that takes.

Build it on the standard, not a blank spreadsheet

The first thing a reviewer checks is whether your statement is even in the right shape. A projection built on USALI 12 lands every revenue and cost on the line the industry expects, which means the reviewer can read it without translation and compare it to benchmarks directly. A bespoke spreadsheet forces them to work out what you meant, and anything they cannot follow, they discount. Starting on the standard is the cheapest credibility you will ever buy. If the format is new, start with the operating statement guide.

Make the assumptions realistic, and show where they came from

Most projections die on two inputs: occupancy and rate. An aggressive ramp to a high stabilised occupancy in year one, paired with an ADR well above the local market, is the first thing a reviewer flags. The fix is not to be pessimistic, it is to be grounded. Ramp occupancy over a sensible period, set ADR against real comparable hotels, and be ready to say where each number came from. A projection where you can name the source of every assumption survives. One built on round numbers and hope does not.

Model the actual deal structure

A projection that ignores how the hotel is actually run is incomplete. If there is a management agreement, the base fee, the incentive fee and the hurdle have to be in the model, on the correct lines. If it is a lease, the rent mechanics matter. If it is a franchise, the full fee stack belongs in the numbers. Reviewers know that the band between gross operating profit and the owner's return is where deals are won and lost, so a model that glosses over the fees is a model they do not trust. See how the structures differ for why this changes the bottom line so much.

Carry the reserves and the capital

A common gap is leaving out the replacement reserve, the money set aside to keep furniture, fixtures and equipment in good order. A projection with no reserve overstates the owner's return and signals inexperience. Include it, usually as a percentage of revenue, and include any planned capital works. Reviewers expect to see the hotel maintained, not run into the ground to flatter a number.

Show your working

The single biggest difference between a projection that passes and one that fails is traceability. Every output should trace back to an input the reviewer can find and change. If they ask "why is year three GOP this figure", you should be able to walk them from revenue to that line without hand-waving. A black box gets distrusted no matter how good the answer is. A model that shows its working invites scrutiny instead of fearing it, and that confidence is itself persuasive.

Test the downside before they do

Reviewers will stress your model. It is far better to have done it first. Show what happens if occupancy is five points lower, if rate softens, if costs rise. Show the break-even, and if there is debt, show the cover, the ratio of cash flow to debt service. A projection that already answers "what if it goes wrong" is one a lender can say yes to. A single optimistic case with no downside reads as naive, and naivety is the easiest thing in the room to reject.

The red flags reviewers look for

In short, the things that get a projection torn apart are familiar: occupancy and rate above the market with no justification, no ramp in the early years, missing or token reserves, fees and structure left out or wrong, a single case with no sensitivity, and numbers that cannot be traced to a source. Avoid those six and you are most of the way to a model that holds.