Pricing software: dynamic pricing vs. repricing and other tools

According to a study conducted by the German E-Commerce and Distance Selling Trade Association (bevh), 40 percent of German online retailers already use dynamic pricing to strengthen their competitiveness. The market offers three categories of pricing software. In this article I will attempt to outline the differences and indicate which strategies can be used with which approach.

Repricing – buy box at all costs

Online retailers use so-called “repricing robots” with the goal of achieving the highest possible listing on marketplaces such as Amazon or ebay. The best-case scenario is for the retailer’s own product to be in the buy box. This pole position is the object of all desire. However, even marketplace operators have their own objectives – namely sales, frequency, market share and commission. The logic behind the repricing tools works accordingly. The price is constantly checked and dynamically adapted. In addition to a few other criteria, price is the measure of all things. Retailers who play around on the marketplace underbid one another non-stop, promoting downwards price spirals. This takes place at the expense of retailers and manufacturing companies who forfeit margins while the marketplace consistently expands its market power.

System-supported rule-based tools – static complexity, intensive consulting and manual effort

Another type of dynamic pricing includes system-supported and rule-based tools that implement a number of the company’s own pricing rules. This is sometimes extremely complex and as the number of items increases, so too does the amount of manual maintenance. These tools do not really fall under the category of dynamic pricing as they only implement the rules in a static way. There is virtually no automated or dynamic adaptation taking place when an extremely volatile market environment is in play. Shopper insights are used to predict customer behavior, corresponding to a predictive analytics approach. More modern pricing algorithms use a prescriptive approach which not only recommends decisions but can also implement them automatically and immediately based on predictions. System-supported and rule-based tools survive thanks to consulting services and great manual effort. They are the precursor to true dynamic pricing.

Dynamic pricing – algorithmic machine learning and prescriptive automatic pricing

It can be assumed that many retailers that already use a dynamic pricing approach fall into the first two phases of its evolution (repricing and system-supported pricing). Dynamic pricing in the form of algorithmic machine learning, however, is not as widespread. This is the highest level of evolution in dynamic pricing.

So what exactly constitutes a state-of-the-art and powerful dynamic pricing algorithm? The first requirement is the ability to flexibly operate different strategies and objectives. Dynamic pricing is KPI-driven. The retailer’s objective is for the customer to make a purchase – and to make that purchase at the best possible price. That is, the best possible price based on his KPIs. This means that the algorithm can be configured for profit maximization, turnover maximization, frequency optimization or inventory optimization, enabling the use of multiple strategies.

A dynamic pricing algorithm starts learning on the basis of the retailer’s objectives and business rules. Within just a few days this kind of algorithm can make accurate predictions about the price acceptance for each article. This takes place by calculating price elasticity. Price elasticity indicates how demand changes with price changes. With fast-moving items this is relatively simple. The challenge then comes in the long tail area, in other words with items that sell less frequently. A dynamic pricing algorithm can determine the price elasticity for all items in the range using regression, cluster and decision tree algorithms. This is a powerful tool for any retailer as it provides accurate predictions about price acceptance based on consumer appreciation.

Competitor prices flow naturally into the pricing. Unlike repricing, the competitor’s price is just one weighted pricing factor amongst many others. The competitor’s price is assessed in the context of its own brand strength, which comes down to the exact and dynamic evaluation of the regional market situation at the point of sale. The right price for any item, at any given time and in any regional branch or online shop. In addition to the competitor’s price, many other pricing factors play a role – the weather, seasons, day of the week or time of day, inventory information, in-house and third-party promotions and, a crucial factor, the retailer’s own brand strength. This allows a retailer boasting good service, high customer confidence and covered parking on rainy days to benefit more than the discount store on the corner. Price acceptance for this retailer is higher. This is justified by what the retailer offers.

A dynamic pricing algorithm learns offline on the basis of an enormous amount of historical data. Comparable to learning how to play chess, the technology predominantly learns online based on real-time traffic data. This way, virtually all price influencing factors can be taken into account in real time – an incredible advantage in a completely volatile market environment. The artificial intelligence constantly analyzes the interplay between price campaigns and customer responses, drawing immediate conclusions as to the effectiveness of its own pricing. This technology is also known as reinforcement learning. The algorithm learns adaptively by measuring changes in sales and adapting prices as well as exploratively by effectively testing price acceptance thresholds.

A dynamic pricing system uses intelligent algorithms that not only support decisions but that actually automate the decision themselves (prescriptive analytics) in the last expansion phase. This is digitalization at its highest level. It goes without saying that category management always has the option to intervene. Management by exception detects outliers before the suggested price is implemented in the physical store or online. Beyond that there will be partial ranges for which the algorithm only suggests one price. This supports category management by providing predictions regarding the effects on gross profit, sales and turnover. The price decision, however, remains completely in the hands of the category manager (predictive analytics). A division of labor begins to show between the algorithm and category management; the algorithm takes on pricing tasks for items that do not require the full attention of the category manager.

A state-of-the-art algorithm sets itself apart by way of networked learning across all sales channels and multiple application scenarios. In the case of omnichannel learning, for example, the algorithm also uses the findings from online shop A for the price decisions in online shop B or in the mobile shop, call center and branches. In the case of multicase learning, the algorithm can even learn via different use cases, e.g. it incorporates findings from recommendations for pricing.

Multiple strategies with modern dynamic pricing

A modern dynamic pricing algorithm is extremely complex. What does that mean for the online or bricks and mortar retailer? What options does this complexity offer? It makes use of the entire keyboard of pricing par excellence – with multiple strategies.

Strategy 1: Value-added pricing

In the case of base and long tail items, which make up the majority of the product range for many retailers, prices are controlled fully automatically. Consumers are not focused on the prices of these items and they are generally only moderately relevant to competition. They do not require the full attention of category management. For these items, the algorithm is configured for the optimization of gross profit. The price gets more expensive where it can, always measured according to the price acceptance and thus the appreciation of the consumer. In these product ranges the algorithm has been proven to raise gross profit. The business case offers a number of advantages: On the one hand, automation unburdens category management and on the other it achieves a relevant gross profit effect.

Strategy 2: Strategic pricing

For base items on which consumers are intently focused and that are extremely relevant to competition, the algorithm is configured for frequency optimization. The price of the competitor is incorporated but is weighted in the context of the competitor’s own brand strength. Price image and market share are important. The price gets more reasonable where it has to but only to the right extent. Ultimately what counts is the customer’s decision to enter the online shop or branch and complete a purchase. Frequency and conversion are the top priority. With strategic pricing a dynamic pricing system supports the category manager using the predictive analytics described above. Pricing authority remains with the category manager.

Strategy 3: Life cycle pricing

Just as with yield management in the tourism industry, here we are talking about items with limited life cycles such as fresh items, seasonal items and consumer electronics. These items lose their value over a very short period of time. For this reason the algorithm is configured to a mix of inventory and profit optimization. The goal is to reach zero inventory by a certain point in time but until then to obtain the best possible price at any given time. This increases gross profit while lowering the costs for depreciation.

Strategy 4: Intelligent couponing

It is also possible to generate personalized coupons. Provided the customer gives his consent, the customer’s click and transaction data can be used. In this case, price differentiation should only take place in the interest of the customer, in other words with the help of personalized discounts. Powerful tools combine personalized and real-time capable product recommendations with personalized coupons. Customer loyalty and conversion rates rise when the most relevant product at a discount also relevant to that customer is provided to every customer at any given time and at every touchpoint.

Conclusion

Not all dynamic pricing strategies are created equal. Only complex algorithms enable the use of the full breadth of pricing strategies and maximize the full potential. This means that increasing earnings while simultaneously improving price image and enhancing customer value or customer retention are no longer mutually exclusive. And in the meantime, processes are digitalized and automated. That is the great art of digitalization.