July 24, 2018
In classical terms, elasticity is simply a link between price and volume. The more an item’s sales volume changes in response to changes in price, the more elastic that item is said to be. Items like luxury goods are fairly inelastic – they’ll still sell even if the price changes – whereas items like frozen foods are highly elastic, selling much more when the price drops.
The problem with this simple elasticity model is that it’s extremely limited in scope. An item’s sales volume depends not only on its price, but also on the location where it’s sold, the value of other items displayed near it, the price charged by the competition, the use of in-store promotions, and many other interrelated factors.
Because elasticity depends on so many influences, it’s highly unpredictable, and can only be measured in relative terms – as a measure of “X is more elastic than Y” – and not in any absolute sense. In light of elasticity’s relative nature, it makes far more sense to consider it not as an inherent property of the item in question, but as a measure of relative value for money.
The concept of elasticity is a fictitious simplification
The classic elasticity engine – a one-to-one correlation between price and sales volume – is fairly easy to measure historically. One can measure the effect of price changes on sales in each store or product category, and establish a set of relative elasticities; for example, “On the whole, sweaters are more price-elastic than t-shirts.”
This measure of relative elasticity can then inform actionable predictions, for example, “We can raise the price of t-shirts by five percent across all stores, and maintain the same revenue with a lower capital investment.”
To generate accurate pricing forecasts, we need to take a closer look at what elasticity actually measures.
In any corporate office for a large retail organization, you’re likely to see compelling presentations on product elasticities. The reports in these presentations are generated with complex Excel formulas, and their level of detail provides a reassuring sense that a products sales can be predicted reliably and consistently as a simple function of price.
But these elaborate models fail at the only test that really matters: actual predictive power. As soon as an unexpected factor enters the picture, reality deviates wildly from the elasticity model’s predictions – and retailers soon face lost revenues and margins, which can be difficult to see at a first glance.
Elasticity is about much more than just price
The problem with conventional elasticity calculations is their dependence on a simple one-to-one correlation – price and volume – that isn’t nearly so simple in the real world.
Price is just one of many factors that influence an item’s sales. If a competitor responds by lowering prices on t-shirts, for example, or stock runs out in some stores, or a certain brand suddenly becomes popular, an item’s price elasticity can shift abruptly, rendering those carefully calculated forecasts invalid.
What’s more, an item’s elasticity depends highly on its current price point. If the price is low enough to provide reasonable value for money, the item may exhibit low elasticity, selling about the same even if the price shifts up and down. But if the starting price is significantly higher than many customers are willing to pay, then the item will be extremely elastic, selling much more when the price drops into an affordable range – or not at all if the price is raised further.
Taking all these variables into account, it’s clear that an elasticity calculation isn’t enough to make accurate predictions about an item’s sales at various prices. If the model is to generate accurate forecasts, it needs to take many additional factors into account – which means we don’t just need to measure the relationship between price and sales; we need to model many relationships at once.
For example, as this heat map shows, average elasticity varies to different degrees in different locations:
So what should we actually be measuring?
The answer to that question is different for every business. For many retailers, elasticity depends on factors like store location, foot traffic, weather, holidays, promotions, social media activity, and a wide range of other factors. The good news is that many of these factors can be quantified and fed into machine learning tools, which generate new predictions based on the latest data, and learn from their mistakes to develop more accurate predictions over time.
Our team at Evo is composed not of business theorists, but of data and software experts focused on delivering tools that actually work. We’ve learned firsthand that a business’s data already holds all the keys to accurate forecasting – and when our software can learn from that data, and from feedback provided by human experts, we’re able to generate forecasts that align stunningly closely with what actually happens in the real world.
We don’t just measure overall elasticity; we monitor its evolution across time and space, and across different channels, stores, and customer segments. We can even detect strong changes in elasticity can be detected to alert the client that something is changing in customers’ perception.
The traditional concept of elasticity is a dangerous simplification. It’s intended to illustrate the link between volume and price, which is actually a highly dynamic, multi-factor link. The only effective way to forecast elasticity over time is to use a multi-signal model of customer/market demand – and that’s exactly what we do.
About the author
Ben Thomas is the Chief Writer and Brand Strategist at Evo, with a core focus on emerging technologies, Big Data, and the Internet of Things (IoT).
He loves to engage audiences about the frontiers of science, culture and technology — and the ways these all come together.