The 9 Reasons Why It's So Hard to Measure Price Elasticity

A deep dive into the traditional model of supply and demand...and its limitations.

If you are reading this, I am sure you have already seen some version of the following chart at least once in your lifetime:

This is the traditional economic model of supply and demand, whereby the volume supplied increases with price and the volume demanded decreases with price, both following nice-looking curves that one can, in theory, easily quantify, measure, and predict.

Then why is it so hard for pricing professionals to answer seemingly simple questions such as “is a cup of coffee price-sensitive?” or “if we take price up 10%, what kind of impact on our volume should we expect?”.

Well, the short answer is because the traditional model of supply and demand is an over-simplification of an otherwise very complex economic environment, to the point that this model is close to useless for pricing practitioners. For the long answer, keep on reading to learn about the nine main limitations of that traditional model.

Price elasticity has a very clear and simple definition

Let us start with the basics. Price elasticity measures the change in volume (in percent) after a 1% change in price. For example, if you lose 2% in volume after a 1% increase in price, your price elasticity is -2.0. In details, it means that:

Price elasticity is a number, which means that it can be multiplied. A 2% loss in volume after a 1% increase in price, or a 20% loss in volume after a 10% increase in price mean the exact same price elasticity of -2.0. There are, however, limits to the amount of price increase (and corresponding loss in volume) that you can apply to this logic.
That number is usually negative. In other words, if price goes up, volume goes down and vice versa. There are very few exceptions to this rule (mostly luxury goods, masters’ paintings, and other extremely rare goods). More interesting (and more common) are situations in which the price elasticity is simply zero: volume does not respond to changes in price.
But it’s its absolute value that matters. Because price elasticity is (almost) always negative, practitioners usually drop the minus sign to focus on the value of elasticity itself. That value is also what truly matters to make a pricing decision: the higher it is, the more price-sensitive the product is. A high price elasticity is desirable when one aims to lower price or run a promotion (because volume will respond positively to the price decrease). Likewise, a low price-elasticity is preferable if one wants to take a price increase.

As you can see, the concept of price elasticity itself is very simple. So why is it so hard to measure in practice?

The golden rule of pricing decisions: increase price on low-elasticity products, decrease price on high-elasticity products.

Price elasticity is incredibly hard to measure…and here is why

To understand why price elasticity is so hard to measure, one needs to deconstruct the traditional economic model of supply and demand (Figure 1) into its three main components: the supply side, the demand side, and the market environment. By decomposing each component even further, one reveals the ten reasons why pricing professionals struggle so much to measure price elasticity.

The supply side

1. Prices are set by producers, not by “the market”.

When you dropped by your favorite coffee shop this morning, did you bargain the price of your medium oat milk latte with the cashier? Chances are you didn’t: the coffee shop had set the price before you even walked into the store, and you paid without asking questions. This applies to most products and markets, both in B2B and in B2C.

In terms of Figure 1, it means that the supply “curve” is in fact a straight vertical line (volume may change based on demand but the price is fixed, at least in the short term).

More importantly, it also means that customers don’t really have a say in the price setting process. Sure, they can “vote with their wallet” and go to competition, but it will be weeks (if not months) before the seller notices it. Having fixed prices set by the “supply side” means that customers’ responses to changes in said prices becomes, by definition, very hard to observe.

2. Prices don’t change that often.

To add insult to injury, in most industries, especially in B2B, companies change their prices only once per year (typically on January 1st), with maybe a small adjustment later in the year. B2B companies selling mostly through multiyear contracts, such as medical device manufacturers, do not even have the luxury of an annual price increase. Even in B2C, it is not uncommon to see businesses such as restaurants taking only annual price increases, at least at the item level. The likes of Amazon and Uber, with their near-constant price adjustments driven by dynamic pricing, are still very much the exception more than the rule.

This state of affair is most vexing for pricing analytics professionals trying to use data to build good models of demand and price elasticity: we were promised a beautiful demand curve with unlimited price points to feed our model; we end up being given 3-5 price points over a 3-year period to work with.

The demand side

3. We cannot see the choices that customers didn’t make.

Because prices change only once or twice a year, the demand “curve” is, for all intents and purposes, mostly unobservable in real life. If you consider demand at a given point in time, all you can see is a single point, at the current price. Even if you expand your observation period to, say, three years, all you can observe are a few points on that curve, corresponding to the various price points set by the supply side over those three years.

Even worse, all we can see are the actual purchases made by the customers. Every time a customer decides not to buy, which in theory would be critical information to fully understand their demand, is also unobserved in practice. In other words, the demand models that one can build with transactional data are biased from the get-go because they are based on purchasing customers only.

4. The demand “curve” is not even a curve.

To address these shortcomings, the industry has turned to surveys and customer research, which we will cover in more details in our next article. One very common technique, called the Price Sensitivity Meter (also known as Van Westendorp, from the name of its inventor), simply asks survey respondents a series of questions about their price perception, such as the price they would consider fair for a given product.

Thousands of such surveys have been done over the years, across multiple industries and countries. They all show the same thing: there is no such thing as a demand curve! Instead, a typical “price fairness” measurement would look like this:

This chart shows two clear patterns, confirmed by a multitude a consumer psychology studies:

Customers have price thresholds in mind when buying (cf. the “cliffs” on Figure 2). In other words, customers care more about a price’s overall order of magnitude (e.g., a $20k car, a $500 TV, a $3 coffee) than they do about the actual price point. This is mostly applicable to consumers, but B2B buyers are (still) humans too and have these thresholds in mind as well.
Customers have “zones of indifference” between these thresholds. This is a well-known consequence of how people read numbers: from left to right, giving more importance to the left digits in a price than to the decimals. That is why many retailers still rely on the good old $9.99 price point: if they were to set that price at, say, $9.73, they would leave $0.26 on the table on every unit they sell…without selling a single additional unit. And if they were to set it at $10.00, their units sold would drop by 20 or 30 percent overnight.

The existence of these two patterns has a clear consequence on price elasticity: it is not constant. Instead, it varies depending on where you are on the price perception “curve”: if you are in a zone of indifference, you will likely observe a rather low price elasticity; but if you happen to cross a price threshold with your price move, price elasticity will appear to be much higher.

5. Customers don’t respond the same way to price increases and price decreases.

To make matters worse, price elasticity also depends on the direction of the price change we want to observe. In other words, a 10% price increase is not the exact opposite of a 10% price decrease, as far as price elasticity is concerned. Generally speaking, price elasticity is higher when price goes up and lower when price goes down, all else being equal (we are talking about “permanent” price changes here; promotions behave very differently).

Where is this asymmetry coming from? First, we all suffer from the “anchoring bias”, covered in depth in Daniel Kahneman’s popular book, “Thinking Fast and Slow”. That bias gives a disproportionate importance to the “starting point”, which in our case would be current prices. In addition to it, we also suffer from the “loss aversion bias” (also covered in “Thinking Fast and Slow”), which states that we are more sensitive to losses (i.e., a price increase of $1) than to gain (i.e., a price decrease of the same magnitude).

Considering that most businesses almost never lower their prices, especially in B2B, it becomes even harder to measure price elasticity for decreasing prices based on transactional data because we can’t just simply “apply” the price elasticity measured on price increases.

6. In B2B, the relationship between quantity and price is reversed.

So far, and despite all the limitations we’ve already covered, we’ve, at least, assumed that the underlying causal relationship between price and volume was still holding. In other words, the traditional model of demand is still fundamentally valid, though wildly imperfect. In most B2B industries, that is not even the case. In fact, the relationship is reversed.

Think of a typical B2B transaction. The seller (either the manufacturer itself or a distributor) offers a product at a given list price. Either during the transaction, or in their contract, each buyer then gets a discount off that list price. And what is the #1 criterion used by B2B companies to determine customer discounts? You’re right: volume! In other words, the net price paid by customers depends directly on the volume they purchase, which violates the traditional microeconomic model of demand all at once.

It makes price elasticity all but virtually impossible to measure in B2B transactions. Using traditional approaches, like the ones used to measure elasticity in restaurants or retail, would in fact only measure the seller’s discounting policy…and how good they are at implementing it.

The multidimensionality of pricing

7. Buying is not done in a vacuum.

Up to now, we’ve treated our supply & demand model as if it happened in some sort of “perfect” environment where nothing else happens. We’ve assumed that demand is driven by price and price only, which would make it fairly easy to measure price elasticity, were it not for all the challenges we’ve listed so far!

This is, of course, not the case. Demand is influenced by many factors beside price, which further complicates price elasticity measurements. Some of these factors, like sales seasonality, are rather predictable and reasonably easy to account for. Others are much harder to measure, lowering demand models’ accuracy. In particular, demand for a given product can be strongly influenced by the price of its alternatives, either from the same company (substitute products) or from its competitors. This is what practitioners call cross-elasticity: the idea that a product’s price impacts its own demand but also that of other products. Think Coke and Pepsi, but also Coke and Coke Zero.

There are two main issues here:

The first one is data availability, especially when it comes to competitors’ prices. Even though things have improved on that front for a lot of B2C segments (especially with the rise of ordering apps and web scraping), getting accurate competitors’ prices is still a major challenge in B2B, even before considering customer discounting (which is usual confidential).
The second issue is complexity. The number of cross-elasticities increases dramatically with the number of alternatives considered, to the point that data scarcity can become an issue, even in data-rich industries such as retail (how many products can be considered alternatives to Coke?) More critically, most businesses usually change prices of similar products at the same time, making it almost impossible to disentangle “direct” price elasticity from cross-elasticities.

8. We are all a segment of one.

The attentive reader may have noticed that I have, so far, used the term “price elasticity”. Singular. We’ve implicitly assumed that we were only looking at a single customer, or, more precisely, that all customers were acting as one and had the exact same behavior, preferences…and price elasticity.

Again, we all know this to be untrue. In fact, the exact opposite would be truer: each individual customer (even in a B2B setting) has their own preferences, biases and demand “curves”. It would of course be totally impractical to try and measure price elasticity at the individual level, which is why companies rely on pricing segmentation to simplify their approach. Even then, identifying the right customer segments is easier said than done.

But more importantly, it now becomes evident that there isn’t a single price elasticity (nor a single demand curve) for a given product. Instead, there are as many price elasticities as the business has pricing segments, which always involves a certain level of arbitrary decisions.

9. Price elasticity changes over time.

Lastly, the traditional model of supply & demand also assumes that time is not a factor. All the decisions (price setting and quantity produced by the supplier, volume purchased by the customer) happen instantaneously, as if we were in a “perfect” market setting.

In practice, it has been observed that price elasticity changes over time and tends to decrease the farther we move on from a price change. Put differently: customers are more price sensitive the first time they see a new price (especially if it follows a price increase) but tend to readjust their behavior and become less price sensitive over time. This is again a consequence of the “anchoring bias”. It also explains why promotions are typically most effective during their first week but tend to lose their effectiveness the longer they last.

One can then measure different price elasticities for the same product and the same price change depending on when the measurement is made. This can significantly impact how a company quantifies the impact of a price increase (or a promotion) and deems it successful.

Wait, all hope is not lost!

After reading this, you might be wondering if it is even worth trying to measure price elasticity. The good news is, there are ways to overcome each of the challenges listed above. In other words, measuring price elasticity is difficult, but it is not impossible. And given its importance in making pricing decisions, it is certainly worth the effort. We will address these solutions in a future article.

Thank you for reading!