Dynamic Pricing in 2025: the AI shock serving your turnover

Dynamic Pricing: what is it?

« Dynamic Pricing » refers to a set of practices that can be used to optimize the revenue of a business (Revenue Management.) The general idea is to be able to offer the optimal price, which maximizes revenue, i.e. turnover, at a given moment t, based on demand.

Simple. Basic.

A subset of Revenue Management

« Dynamic Pricing » is therefore part of a broader function known as « Revenue Management », whose purpose is simply to optimize the revenue of a business. It's a discipline that remains confined to certain business sectors for now.

At Junr, we believe every company should tackle this function head-on!

Which includes Yield Management

« Classic » Dynamic Pricing can be applied to any product/service to boost your revenue thanks to Science. Magic.

So « Dynamic Pricing » could be implemented in any business. But that's not yet the case.

That's where Junr comes in! We promise no more ads until the end

So, « Dynamic Pricing » means varying prices based on demand. The topic has therefore naturally been dominated by companies whose… stocks are limited, perishable, and ideally intangible.

Why? Because it's in these businesses that a pricing error is the most violent : no way to « make up for it » over time, impacts that are proportionally higher in value.

You've guessed who we're talking about, haven't you? Yes, your plane ticket that goes up day after day, just like your hotel room, and, less visibly, in your favorite hypermarket.

« Dynamic Pricing », when applied to a perishable inventory, is called « Yield Management ». It's probably the most sophisticated pricing method that exists. And yet it's too rarely implemented, and above all, often without taking advantage of modern technologies: AI, real-time,… We'll come back to that later, we promise.

Before that, a bit of history…

A touch of history, it's a fun one!

In reality, the notion of « Dynamic Pricing » has no clear history depending on what you consider a dynamic price to be: does a business based on negotiation not ultimately have the effect of creating variable prices based on demand, i.e. the depth of the sales pipeline?

The notion of « Yield Management », however, is clear. In the 1980s, the American government decided to deregulate airline fares. Following this, « low-cost » carriers emerged. To fight them, Delta Airlines invested massively in Revenue Management and was the first company to implement « Yield Management », closely followed by American Airlines.

The CEO of People Express Airlines, a « low-cost » carrier of the time, would declare:

« We were a growth and profitable company from 1981 to 1985, then we switched to losing $50M a month. We were profitable from the day we started until American [Airlines, ed.] attacked us with Ultimate Super Savers », their Yield system.

Just goes to show, it's powerful. Don't be the next People Express, be the future American Airlines!

If the history interests you, more details on wikipedia: Yield Management

Are you concerned?

Whatever the case, you'll need robust sales histories to set up « Dynamic Pricing », in order to predict the level of demand

By Revenue Management

Everyone is concerned by Revenue Management, since it's about optimizing prices to maximize turnover. Despite this, many companies still don't have rationalized and automated analysis as support.

By Dynamic Pricing

« Dynamic Pricing » is applicable to any product or service whose demand varies over time (day, week, month, year..). In the end, few are those who are not concerned.

It nonetheless remains necessary to have solid sales histories in order to consider predicting future demand.

By Yield Management

In the specific case of « Yield management », it is essential to sell a product whose quantity is limited and perishable, and for which you have or will implement a centralized inventory.

This product or service can be tangible or intangible, as long as it is limited and a quantity not sold on a specific date cannot be put back on the market: from the can of peas to the hotel room, by way of dated ticketing, a large number of businesses are concerned.

Dynamic Pricing: the mandatory components

Demand prediction

The entry point of « Dynamic Pricing » is the analysis of future demand, in volume.

That's where the normalization of Big Data with AI comes in. In recent months/years, solutions to store and process large data models have become widely democratized, as has accessibility to artificial intelligence solutions.

It is now relatively inexpensive to create real-time predictive models, and any company can aspire to it, contrary to popular belief.

Market demand

Market demand is the total number of people who need/want your product/service independently of the company that produces it. This demand can be and is influenced by many factors (historical demand, real and perceived economic conditions, market situation, weather,…).

AI is a formidable tool for processing all this data, eliminating human biases, and delivering a reliable prediction.

Product demand

Starting from market demand, we'll need to know which portion is reasonably likely to buy from you based on your ability to convert, your market share, and plenty of other factors (updated positioning of the competition, ongoing promotion, specific event, …).

Here again, AI is a formidable tool for compiling these factors in real time and providing a demand prediction for your company.

Internal and external substitution

When we talk about demand, competition, market share, it is also necessary to talk about substitute products/services. In a « Dynamic Pricing » system, it is essential to know, for a price change, the number of people who will fall back on another product, whether internal or external to your company. Here we are not talking about elasticity but rather substitution, or shifting of demand.

Unlike elasticity, which we'll talk about next, substitution concerns consumers who will still consume, but a different equivalent/competing product.

In the case of internal substitution, this can be beneficial or even desired to balance demand across all products and/or on products with higher incremental value.

Determination of elasticity

Elasticity is the measure of the variation in demand relative to the change in the price of a good. By good we mean the set of products or services in direct competition on a market. As seen previously, the shifting of demand concerns demand « sliding » onto a product B when product A increases in price, whereas elasticity is the outright loss of demand for a good on its market when the price generally increases.

Elasticity is essential to measure in a « Dynamic Pricing » system. Along with demand, it is the second measure that allows predicting the expected sales volume at a given moment t, for a product p.

Elasticity can be predicted thanks to Machine Learning models, as well as measured thanks to A/B testing. The two are generally combined to strike the right balance between the robustness of the analysis and the disruption of optimal prices.

Price elasticity

Price elasticity is the purest measurable elasticity, namely how many consumers the market loses when prices increase by €1.

Cross-price elasticity

There is also another form of elasticity, cross elasticity, which tends to measure the demand that will « fall back » onto a good B (indirect competition) when market prices increase.

To illustrate the differences between internal substitution, external substitution, elasticity and cross elasticity, nothing beats a little example:

Let's imagine a campsite, with tent pitches and mobile homes available.

• If by increasing the prices of tent pitches, my company's demand « slides » onto the mobile home offer, this is internal substitution: demand « slides » from a product A to a product B within the company itself;

• If by increasing the prices of tent pitches, my company's demand « slides » toward the tent pitches of the campsite across the road, this is external substitution: demand « slides » onto the same product of a direct competitor;

• If, when hospitality market prices are inflationary, the market loses value, my market has a positive elasticity: demand decreases as prices increase;

• If, when hospitality market prices are inflationary, the restaurant market gains value, the restaurant market has a cross elasticity with that of hospitality: demand increases as hospitality demand decreases.

Elasticity effects are often the consequence of trade-offs. For example, the food market is very inelastic because consumers rarely make the trade-off of reducing their budget on this item in the event of inflation, but rather the trade-off of reducing their spending on other items to compensate.

It is very interesting and often surprising to analyze these effects and trade-offs in your demand.

Optimization

In « Dynamic Pricing », optimization is the phase during which, knowing the level of demand and its elasticity, a system takes care of finding the perfect price that optimizes Revenue.

Revenue function and optimality

The revenue function is the mathematical function that determines the optimal price. In theory this function is relatively simple.

An example being worth a thousand words, here it is, under the assumption of a positive and linear elasticity over the price interval.

If you want to have a little fun, below is a very simple simulation file.

But here's the thing: not only is elasticity not always positive (even if that's the most common case), but above all it is never linear over the price interval. Furthermore, we imagine here that only a single product is available, which is also rarely the case. It will therefore be necessary to break down this approach across several products, whether or not they are indexed to the reference price, and according to the different price-dependent elasticity curves.

Constraints

When setting up a « Dynamic Pricing » strategy, we don't want to let the models navigate without any rules. It is therefore appropriate to define some, the first being the price interval within which we want to leave the model free. Here, each constraint, as its name suggests, will introduce sub-optimality, but may be necessary in terms of image and strategy.

Resolution

Each system is therefore ultimately built to respond no longer to the optimal price, but to the best price given the constraints we impose. It must nonetheless be kept in mind that, by extension, each constraint is a step toward a loss of revenue, because it forces the model not to purely apply the optimal price.

Dynamic Pricing: what gains can you expect?

Price and volume, beware of biases

It may seem like stating the obvious, but in « Dynamic Pricing » revenue is not a function of volume, nor of price, but of the product of the two. The goal is not to sell as much as possible, but rather to sell at the best price, the one that maximizes total revenue.

Using the same example as before, we can clearly see that beyond a certain volume level, revenue decreases.

Yet we observe a large number of Yield teams always chasing occupancy, at the expense of Revenue.

The gain is therefore best appreciated in revenue, rather than in volume or price. Since revenue is a calculated indicator, it can be difficult to get past biases: it is therefore appropriate to build the right visualizations to track the right KPIs, at the right times.

This being said, we observe from experience an average gain of 9% in turnover, and an operational cost reduced by an average of 3%, which brings an average gain in commercial margin of around 10%.

Optimality gains

To sum up, artificial intelligence tools and mathematical optimization allow us to ensure we reduce biases, automate « Revenue Management », while ensuring the best optimality gains, which ultimately do not depend on price alone, but on where we sit in the revenue curve:

New practices

Although few yet resort to AI in « Dynamic Pricing » models, we are already seeing new practices emerge among the pioneers. As part of building a system, it would be sensible to consider them from the outset.

Margin optimization

Optimizing revenue alone is very limiting insofar as margin is not taken into account. This may seem obvious, but the « Revenue Management » discipline, it's even in the name, generally focuses solely on turnover. It's an approach that is starting to become dated, and creates perverse effects.

Accounting for ancillary revenue

Another limiting subject: not considering the revenue or ancillary margin generated by a customer. For example, optimizing the revenue generated by a seat on a plane does not consider other expenses (food, drinks, etc.). It's quite simple, and we sometimes realize that a child's ticket reduced by almost half is actually more profitable than an adult ticket, because the child is an admirable prescriber of additional consumption!

Conclusion

10% gains on a company's overall margin, that's huge! Especially fully automated. But the change it implies should not be taken lightly, nor should the techniques and technologies to be implemented to get there.

To implement the « Revenue Management » strategies described, you must meet the following criteria:

• For « Dynamic Pricing »: have demand that varies over time
• For « Yield Management »:
  • Have demand that varies over time, and
  • Have a perishable inventory.

At Junr, we support SMEs and Large Groups in implementing automation and artificial intelligence strategies.