Revenue Managers Often Can't Explain Why Their Strategy Worked

A hotel can post ADR up 12% and RevPAR up 15% in a single month and still have run a bad pricing strategy. That is the core argument of a sharply worded critique circulating among revenue management professionals this week, which warns that the discipline systematically confuses good outcomes with good decisions — and that AI tools are about to industrialize the problem.

The author, who signs the piece "Fabi," builds the case around a familiar scenario: a hotel closes the month with ADR up 12%, RevPAR up 15% and improving market share, and the revenue meeting becomes a celebration of strategic brilliance. The praise, the author argues, is often unearned.

What does a good result actually prove?

"A good result doesn't prove a good strategy. It only proves you achieved a good result," the piece states.

The counterexamples are concrete. If the market grew 20%, the hotel underperformed its potential. If the biggest competitor had 100 rooms out of order, share gains came from supply disruption, not pricing skill. If a major event generated demand nobody forecast, or an OTA algorithm suddenly improved the hotel's visibility, the numbers would have moved regardless of strategy.

The sharpest version of the argument: the strategy may have left money on the table. Doing absolutely nothing might have delivered 18% RevPAR growth instead of the 15% the team celebrated.

Hindsight bias runs the post-mortem

The critique identifies hindsight bias as the structural flaw in how hotels review performance. Whatever happened, the narrative adjusts to flatter the decision-maker:

• Occupancy rose after rates were cut — the discount worked.
• Occupancy rose after rates were raised — premium positioning worked.
• Rates were held and occupancy rose — patience was rewarded.

"Whatever happened, we can usually find a story that makes us look intelligent," the author writes.

The danger compounds over time. Once a team believes its explanation, it repeats the strategy — because it worked last time.

Correlation keeps masquerading as causation

Two worked examples anchor the argument. A hotel raises rates 10% and revenue follows. The increase may deserve the credit, but demand may have been strengthening anyway, competitors may have raised rates even more, channel mix may have shifted toward higher-paying customers, or the old rates may simply have been too low.

The reverse case is a promotion that lifts bookings 30% and triggers a celebration. The missing questions: how many of those customers would have booked anyway, how many existing customers just paid less, how many bookings were displaced from another channel, and what happened to profitability after acquisition costs.

Without those answers, the author argues, teams measure activity and outcomes while assuming the relationship between them.

What changes with AI in the room?

The most pointed section addresses technology. AI can now analyze large datasets, identify patterns and generate explanations in seconds — covering pricing, demand, segmentation, booking windows, competitor behavior and market conditions, complete with a chart.

"A convincing explanation isn't necessarily a correct explanation," the piece warns. If the underlying data cannot isolate the impact of a decision, AI cannot manufacture the missing counterfactual. It can test hypotheses, surface competing explanations and estimate possible effects — but it should also be able to say when the evidence is insufficient.

The author's concern is that humans love certainty and AI is convenient at supplying answers that sound certain. The result: "revenue managers who never understood why their strategy worked now have an AI agent confidently explaining why it worked." The author's one-line verdict: "We've automated the bullshit."

A practical fix, and a tolerance for not knowing

The proposed discipline is straightforward. Before changing a price, launching a promotion or adjusting distribution, document what you expect to happen and why. Define the behavior you intend to influence, the other factors that could affect the result, and the evidence that would prove the hypothesis wrong.

Then measure the intended behavior, not just revenue: did the promotion create incremental demand, did the rate increase genuinely improve profitability, did the distribution change improve customer mix or merely shuffle bookings between channels?

And sometimes, the author concedes, the correct conclusion after all that work is "we don't know." That answer is acceptable — and, in the author's framing, considerably more intelligent than inventing certainty.

The piece closes with a structural warning for the profession: the most dangerous revenue manager is not the one who makes a bad decision, but the one who makes a bad decision, gets a good result, and spends the next five years repeating it. The interesting question, the author writes, is whether you understand the difference.