Revenue Management: why confusing supply and demand is costing you money

Series: Revenue Management

In this series, we aim to take a clear-sighted, grounded yet ambitious look at the major trends transforming our discipline.

Dynamic pricing, OTA, e-distribution, rate disparities, Big Data, RMS, AI, Profit, TRevPAR… The Revenue Management profession is changing. So are beliefs. So are expectations.

Wind of Changes · E01 · In short

Revenue Management consists of aligning your own supply (price, product, availability) with observed real demand, rather than simply following competitors’ pricing.

Compset pricing is a useful explanatory variable, but it should never be a standalone decision-making input. The true decision signal lies in demand dynamics, compared against a forecast.

Bring in the accused: they are confusing supply and demand

Major events often act as a revealing lens on Revenue Management practices.

The Olympic Games, Taylor Swift’s tour, an Oasis concert, a Champions League final or a Céline Dion show: each time, hotel, ticket and transport prices are scrutinised by the media.

The criticism is usually the same. Prices are said to be excessive. Yield Management is portrayed as the expression of unchecked greed. Yet these critiques often mix two very different realities: a philosophical debate about pricing and accessibility on one side and entirely avoidable commercial mistakes on the other.

Because behind certain excesses lies a much deeper confusion. A confusion that does not only concern major events but the entire year. A confusion that has been reinforced over the years by parts of the market, consultants and certain software vendors. The confusion between supply and demand.

Why is competitor pricing not a decision-making variable?

Modern Revenue Management has become obsessed with observing market prices, whereas its historical mission is to measure demand velocity.

No, the level of market supply (average price, median price, or the compset’s pricing positioning) is not, in itself, a decision-making variable. And even less so the sole decision input of a dynamic pricing algorithm or an AI system (a set of algorithms).

The original mistake is to believe that Yield results from the intersection of market supply and demand. In reality, it results from the intersection of one’s own supply (price, product, availability) and real demand.

It is the dynamics of transactions (the match between my offer and demand), compared to a forecast, that provide the signal. An acceleration in pick-up may indicate an upward lever (optimising for higher average rate or shifting volume to adjacent dates). A slowdown may suggest action on restrictions, inventory or pricing.

The decision signal lies first and foremost in observed demand, not in the price displayed next door.

What do last-minute price drops during the Paris 2024 Olympic Games tell us?

The Paris 2024 Olympic Games have become an emblematic case. A real-world demonstration at scale.

For several months, the pricing levels displayed fuelled the idea of exceptionally strong future demand. Yet when last-minute price drops appeared, one question emerged:

How should we interpret a sharp price decrease just days before the event?

Two possibilities exist:

• Either cancellations were underestimated.
• Or the projected occupancy was overestimated.

In both cases, the issue is not competitor pricing. The issue is a forecasting weakness: cancellation forecasting, on-the-books projection, or demand interpretation. In other words, if high prices were supposed to confirm exceptional demand, what do last-minute price drops actually tell us?

Most likely, the initial signal was less reliable than assumed. The decision-making value of compset pricing then loses much of the weight typically assigned to it.

What will a Revenue Management AI actually learn?

The rise of AI is above all the rise of learning. Claude, GPT or Mistral  have no native understanding of elasticity. They do not inherently know what strong or weak demand is. They learn from the data we provide them.

Game theory teaches us that agents try to anticipate the decisions of others. Each decision depends partly on the expected behaviour of competitors.

Applied to Revenue Management, this logic can lead to a dangerous line of reasoning:

« If competitors are significantly increasing their prices, it is probably because they anticipate strong demand. »

The problem is that this assumption quickly becomes self-fulfilling. Everyone observes everyone else. Everyone interprets the same signals. Everyone validates their neighbour’s decisions.

Until real demand comes in and reminds everyone that customers were never part of that conversation.

An AI trained primarily on market pricing behaviour risks learning the market’s biases rather than learning demand itself.

What is market pricing actually useful for?

Stating that supply is not a decision-making variable does not mean it is useless. On the contrary. Supply remains an excellent explanatory variable, a reference point.

It helps to understand market context, to validate a decision, to observe competitive reactions, or sometimes the complete absence of strategy from certain players. It therefore deserves its place in an RMS, but in its proper place as an indicator, not a compass.

Because when a Revenue Manager starts driving performance based on observed prices rather than observed demand, they are no longer doing Revenue Management, they are doing market tracking. And while they believe they are anticipating, they are simply reacting, without any additional context.