Dynamic pricing is nothing less than the holy grail of e-commerce price management: a dynamic pricing strategy that sets prices based on a predictive algorithm fed with available data. Most e-commerce players are using some sort of algorithmic predictive pricing with great success. However, there are a few pricing truths often forgotten once the data science buffs start taking over the show.
When to do Dynamic Pricing
Let’s start by giving dynamic pricing its well-deserved credit. Summing up, there are three main applications where dynamic pricing is likely the way to go:
#1 Complex portfolio
You have an online shop listing hundreds, if not thousands of products all at different prices. Even the best pricing manager would be unable to optimize pricing for all of them simultaneously. Naturally, some automated solution is needed.
#2 Tough competition
The products you sell are also sold by other easy-to-find online or offline vendors, and so you need to stay on top of their pricing constantly to be able to adjust yours and compete.
#3 Dynamic customer price acceptance
Dynamic pricing allows maximum price differentiation, so it can most optimally exploit customers’ price acceptance if it varies by context or point in time. This is the case in many areas such as travel booking, mobility, fashion, or eating out.
Most e-commerce players face some combination of the above. So a dynamic pricing strategy makes quite a lot of sense. However, they should never just blindly implement some run-of-the-mill data science model and trust that it will increase growth, market share, or EBITA.
How to do Dynamic Pricing
When you implement dynamic pricing, make sure you adhere to a couple of rules:
#1 Understand the decision motivation
The starting point for any activity that tries to influence customer decisions must be those very customer decisions you are trying to influence. Start by asking: why are customers visiting our online shop? Are they likely to browse around for a better price on the same product?
If your products are not offered anywhere else, competition’s prices should not be of as much concern to you. Is buying on your side mostly a spur-of-the-moment decision, or is a customer likely to look at your product several times before finally putting it in that virtual cart? In the latter case, changing prices will likely make your customer more focused on price in the long run, or even postpone the decision, waiting it out for a better deal later. In the worst case, this may paralyze the decision completely, as the customer gets the feeling that there is never the best time to buy.
Vocatus’ GRIPS typology will give you the most comprehensive understanding of your customer’s actual purchase motivation. Knowing GRIPS will help you decide whether dynamic pricing is even the way to go.
#2 Clean up the choice architecture
Picture the following customer journey: A shop visitor with a given set of characteristics (Android phone, visiting at 9.30 pm, coming from Google, etc.) browses through your products, looks at a few of them, then leaves the webpage not having bought anything. That’s the most generic possible description of a non-converting visit. So what will the pricing algorithm by its very design assume is the reason for that non-purchase? It will assume that surely the price must have been above that customer’s willingness to pay
Behavioral economics, however, tells us that a price is never perceived independently from the context in which it is put. In fact, we’ve argued before that willingness to pay is even the wrong concept to price against, to begin with. Maybe the product was not too expensive for that customer at all, it just happened to be the only one presented in a row with other products, which was not discounted, thus merely appearing to be bad value in comparison. Maybe the description of the product just didn’t do a particularly good job at conveying the product’s use to the customer.
In short, dynamic pricing will not make up for bad sales shop architecture, and may even end up maximizing lost revenues this way.
#3 Search for causes, not correlations
Dynamic pricing is often understood as a data-mining exercise. Somewhat simplified, the premise of data-mining is often this: “Look, we collect enormous amounts of data all the time – can’t we do something with it?” You surely can, but for reliable pricing decisions, this is rarely enough.
The more data you analyze, the more likely you are to find correlations in the data. For example, you might find that iOS users have a higher conversion than Android users, and that Android users are younger on average. This might draw you to the conclusion that younger customers are more price sensitive. However, it might be the case that the webshop just looks more compelling on iOS than on Android. So the most readily available transactional (what did which customer buy before?) and contextual (when did the purchase happen, on which brand and operating system of the device was it made?) data are often not telling you anything about the true reasons (not) to buy.
For pricing decisions, you need to find causes instead of correlations. Cause-and-effect insights can for one be induced from longitudinal data, which is from data about purchase and price comparison behavior (Is the browsing very directional, or there a lot of back-and-forths? How much attention is paid to product details and small print?). More ideally, the data should be complemented with insights from controlled experiments to create the full picture. Both longitudinal data and experimental data require a behavioral, hypotheses-driven approach to data analytics.
Implementing dynamic pricing should therefore not start with just the statistical relationships in the available data, but with developing hypotheses about customer decision behavior that you can test against the data. Even more importantly, it should reflect on what additional data may be needed to prove causation. If you are to make your pricing (out of all things) dependent on a data-driven algorithm, we suggest not sparing efforts in sourcing data that validly predict customer decision behavior.
#4 Align with your pricing strategy
Your pricing is not just another aspect of your business. It may be what makes or breaks your bottom line. Always remember: If you are currently operating at a gross margin of 10% and you can increase your prices by 2% without losing market share, you have increased profits by a whopping 20%! Vice versa, if your price level suffers by a mere 2%, you have slashed profits by 20% also. This effect maximizes the thinner the margins are that you operate on.
Things like price image matter, too: If your company position is as the save, reliable, no-frills provider, having prices change multiple times a day or between people visiting the shop is just not a good look. If dynamically changing the prices seems odd to you, likely it will not fare well with your market either. For more information on that matter read: What does Brand Positioning have to do with Behavioral Economics?
Make sure your dynamic pricing is built, managed, and monitored by at least one pricing expert who is on top of the company’s long-term goals and market positioning. A short-term sales boost through dynamic discounting for example can be ruinous in the long run as it may end up triggering a price war. For example, in the German e-commerce market for kids’ fashion, the average discount level increased from around 10% in 2015 to more than 20% in 2018. Just because automated dynamic pricing was programmed to undercut competitor prices. This turned a whole fashion category from being a Price Accepter’s to being a Bargain Hunter’s market, to speak in terms of GRIPS.
To sum up
When applying dynamic pricing, make sure that your pricing is based on your customer’s actual decision-making and is also in line with your overall pricing strategy. Both are only possible with a true understanding of your customers’ motivation and behavior. Then (and only then) will your dynamic pricing ensure growth at sustainable margins in the long run.
