Dynamic pricing is currently conquering e-commerce, helping retailers achieve higher margins despite growing competition. With dynamic pricing, prices are not fixed, but are regularly adjusted to current market conditions. This is a strategy that pays off. On average, companies achieve a 25 percent increase in sales when utilizing dynamic pricing. In this article, we cover what you should look out for when implementing a dynamic pricing strategy, and share successful use cases.

How do you develop a dynamic pricing strategy?

There are as many dynamic pricing strategies as there are companies in the market. For a strategy to be successful, it must be tailored to the specific company. An algorithm that adjusts prices at fixed intervals has little added economic value. Not even industry-specific recommendations can be defined: dynamic pricing models that boost sales in the travel industry, for example, can be counterproductive in the fashion industry.

Despite the customization required for effective utilization, some basic requirements exist for effective dynamic pricing strategies in e-commerce and beyond.

Alignment with business objectives

Retailers should design their algorithms so that price adjustments are always in line with their corporate goals and brand identity.  Firms known for low prices should therefore define maximum prices in their algorithm that are below the market average.

Conversely, companies selling luxury items should not go below certain minimum prices, as this would run counter to their image.

An example: An online store for discounted branded clothing could possibly raise the prices for a certain winter coat model significantly due to seasonal peaks in demand. However, this would irritate customers and not fit the brand image. If they have programmed a maximum price into their pricing application, the algorithm will only raise prices to a level that is compatible with the brand identity, and at the same time best supports the underlying corporate objectives, even if there is potential for an increase.

Consideration of relevant factors of price elasticity

In order to take advantage of the potential of dynamic pricing in e-commerce, companies should include all factors in the algorithm of their pricing application that have an impact on their business. To what extent do certain factors reduce or increase demand? The more accurate the analysis, the more precise the algorithm can be.

Some important factors to consider include:

  • Competitor prices
  • Seasonality
  • Marketing budget
  • Weather
  • Consumer trends
  • Inventory

Competitor analyses can reveal which products they sell very well despite comparatively high prices. These or similar products should also be offered at premium prices themselves. Demand for products in different designs or colors often differs significantly. Here, a market and competition analysis can reveal opportunities to identify which product variants are suitable for higher prices.

Analytics and usability requirements

Companies usually associate the desire for a dynamic pricing strategy with more than just the goal of optimizing their prices for higher profits. Those responsible should clarify what other expectations they have for the implementation of the strategy.

What functions should a dynamic pricing application map?

  • Profitability analysis
  • Automation of price management
  • Forecasts of price developments
  • Analyses for personalized pricing
  • Market and competition analyses
  • etc.

Only after you have created a detailed catalog of requirements can you select the right tools for your goals.

How do you implement rule-based dynamic pricing?

When talking about dynamic pricing in the realm of e-commerce, online retailers usually mean rule-based dynamic pricing. This is the current standard in the e-commerce industry.

With rule-based pricing, companies use software to define if-then rules for price adjustments based on their own business goals. How they design the rules is determined by their own in-house expertise. The application configured in this way then “observes” the market. If it detects a triggering variable (If), it uses an inference engine to perform a previously stored price adjustment (Then).

Example: If a competitor reduces the price of product A by 10%, a retailer reduces the price of their own product C by 5%.

The system relies solely on its internal knowledge base to interpret market conditions. Once defined, rules are executed until they are manually changed.

If you want to introduce rule-based dynamic pricing in your company, you can base your process on the following four steps:

  1. Formulation of corporate goals
  2. Deriving a pricing strategy
  3. Definition of pricing rules
  4. Implementation and testing
Formulate the business objectives

Before you price your products, consider your corporate goals and values: What image do you want to give your company? How does it differentiate itself from competitors? Are there profit and revenue targets to be achieved? Which other KPIs (e.g. sales rates) play a role?

It is important to clarify these fundamental questions at the beginning in order to program the pricing algorithm later, so that it acts in your favor and does not irritate customers or stakeholders with unpopular measures.

Derivation of a pricing strategy

The individual dynamic pricing strategies in e-commerce can essentially be assigned to four categories.

  1. Segmented pricing: prices are adjusted according to product type, market segment or sales region.
  2. Peak pricing: prices are adjusted according to demand for a product or product category.
  3. Time-based pricing: pricing depends on the length of time certain products have been in the range. Online stores that sell fashion often use time-based pricing because of changing trends.
  4.  Penetration pricing: prices are set slightly lower than for comparable competitive products. The strategy is often used for new products that want to achieve high market penetration quickly.
Definition of pricing rules

Once the basic strategy has been defined, pricing managers define the specific rules for adjusting prices. These can apply to individual products, product categories or the entire product range. In most cases, the pricing rules refer to costs, competitors or demand.

– Costs: Prices are adjusted dynamically to ensure a consistent margin despite fluctuating company costs.

– Competition: prices are kept in a certain ratio to competition.

– Demand: prices are increased when demand increases and reduced when demand decreases.

The rules are formulated in if-then structure and stored in the pricing application.

“If competitor X runs a marketing campaign for product A, increase the price of own product B by 15%.”

“If demand for product C increases by more than 10% in a month, increase the price by 5%.”

“If a product is in the FOOD category, let its price end at 99.”

Most companies combine different rules, resulting in a complex rule system based on the entire product range.

Implementation and testing

Once you have established the rules, store them in your pricing application. Check that the application reacts correctly to market changes and according to your specifications before you go live with the pricing system.

You should continuously monitor the market so that you can quickly adjust your pricing rules as changes occur and not give away revenue. Some rules may not be useful year-round or only during a specific growth phase of the business.

Disadvantages of rule-based pricing

Rule-based pricing can optimize revenue, but there are limits to its effectiveness.

Consider an example:

An online retailer offers 1000 products in its store, many of which are private labels. The company’s goal is to attract customers by offering low prices and thus gain a competitive advantage. The private-label products are already low-priced. Now the company wants to optimize prices for other products as well. The company decides to lower prices on a rule-based basis by orienting itself to the competition. It identifies five relevant competitors and establishes rules for different product categories: “If a product is grouped in the yoga category, apply Rule X.” Rule X states that the price should always be 10% lower than the competitor. The company monitors the price movements of other market participants with the help of tracking software.

The disadvantages of rule-based pricing

The larger the assortment, the more complex this approach is. More and more pricing rules have to be formulated, and the formulations become increasingly complex. To keep the rules manageable, they refer to sub-ranges or categories. As a result, the special features of individual products are not taken into account and price optimization falls short of its potential.

The system does not react to unpredictable changes in the market, such as changes in trends or the weather. Pricing managers have to monitor the market. Often, manual adjustments do not succeed fast enough, so short-term sales opportunities remain untapped.

If new products are added, the applicability of the rules must be checked manually. If necessary, pricing managers have to add new rules or revise old rules. Over time, this process becomes more and more time-consuming and it becomes more difficult to keep track of whether all the rules really still meet the company’s objectives.

Rule-based pricing is ineffective overall: it requires a high level of manual maintenance and delivers inadequate results.

Specific disadvantages of competitive pricing

In the example, the company links its prices to a competitor. This makes for simple rules, but it is not a stand-alone dynamic pricing strategy. You fuel a downward pricing spiral where all suppliers only lose. In some cases, it may make sense to align prices with those of the competition. This applies above all to products that are subject to high price elasticity.

However, companies that take the effort to analyze the market more closely realize that there are many products for which customers are willing to pay premium prices – regardless of the pricing policy of the competition. They pay higher prices because they value the quality of the product, the company’s service or the brand. They may also fear that the product will soon no longer be available or that they will not find a comparable offer on the market.

Special caution applies to private labeling products: manually identifying comparable products from the competition here is time-consuming and difficult. The configured algorithm is prone to errors. Companies may be able to achieve higher sales in this way, but profitability may decline due to the increased effort.

Online retailer Vitafy has implemented a dynamic pricing strategy with the help of 7Learnings. Vitafy offers a variety of fitness and weight loss products, including under the Bodylab and GymQueen brands. The result of the price optimization: a profit increase of more than 10%.
Learn more in the Vitafy case study here:

How can machine learning-based pricing be implemented?

Machine learning applications are fundamentally different from traditional software: they develop solutions on their own instead of just executing previously programmed rules.

Machine learning algorithms evaluate databases for this purpose: they recognize patterns, derive conclusions, and ultimately find solutions to previously defined problems. They do not need explicit instructions to solve the problem. The more comprehensive the database, the more learning material is available to the algorithm and the better its results. Only since the advent of machine learning (ML) has it been possible for analysts to evaluate such huge amounts of data in a short time as part of a dynamic pricing strategy. Previously, these extensive evaluations took too long to be used economically.

Pricing applications based on machine learning are able to consider diverse factors when calculating the ideal price.

Internal factors: Manufacturing costs, product characteristics, pricing history, and others.

External factors: season, competition, region, among others.

Companies do not need to create complex rule systems, but can work with simple instructions to optimize prices on specific KPIs. If they aim for a certain margin or inventory level, the pricing application automatically adjusts individual product prices to achieve these goals. The ML-based pricing application observes the market and continuously optimizes various prices without requiring manual intervention from pricing managers.

Price elasticity is a key factor in price optimization. This is because it indicates how demand will evolve in response to price changes. Machine learning algorithms can capture the relevant factors of price elasticity much more accurately than previous rule-based applications, starting from existing data. These only take into account some obvious factors such as competitors’ prices and leave a lot of revenue potential untapped.

If you want to introduce an ML-based pricing application in your company, you should proceed in four steps. The steps provide a good orientation as to what economic effects can be expected and what investment will be necessary, and what time should be planned until the launch.

Depending on how the evaluation turns out, companies should opt for a third-party solution or an in-house development for pricing optimization.

Collect and prepare data

The first step to implementing a dynamic pricing strategy is to create a solid data base. The more meaningful data companies collect, the better the machine learning algorithm can work later.

A practical example: An online retailer for sportswear and fitness nutrition wants to expand its online store to other countries. Various factors have to be taken into account for ML-based price optimization: For example, prices should adjust in the future depending on trends and stock levels. Depending on demand, there will be discounts for some products, but these may apply to different product categories. These requirements must be incorporated into the database, as must thousands of data records on sales and inventories of individual products and the millions of data on the behavior of competitors.

Pricing managers need to know their company and their market well in order to identify and compile relevant data. This is no easy task, as the data is stored in many different formats and sources, must be cleansed of errors and made analyzable. Transactional data, product attributes, information on inventories, competitors, historical prices, and more are needed.

Challenges in data preparation
  • Data volume: The amount of data is key to the subsequent quality of results. Companies should make sure that the pricing application can handle sufficiently large volumes of data.
  • Data quality: the best machine learning algorithm is only as good as its database. Before data is used, databases should be checked for relevance and errors.
  • Data analysis: Analyzing and evaluating the various data manually takes a lot of time. Although software is available for this purpose, the applications are often error-prone, so that manual checking is preferable in individual cases.
  • Data management: Preparing the data for the machine learning algorithm of the pricing application can take several weeks to months if not done by experienced data scientists.

After data collection and preparation, the ML algorithm needs to be trained. You can read how such training can proceed in our detailed blog post “How Dynamic Pricing Works.”

Selection of the machine learning model

A dynamic pricing system autonomously detects patterns in data so that it can incorporate price elasticity factors into the pricing model that a pricing manager would not have recognized as relevant at all.

The more accurately the price elasticity factors are captured by the pricing model, the more accurate the subsequent forecasts of how price adjustments will affect demand and sales.

High-quality pricing models are characterized by two features:

  • Performance: The application is able to capture, analyze and communicate data quickly.
  • Precision: The application provides accurate forecasts of the impact of price changes for each product.

Once the model is trained, the predicted prices can be tested for accuracy and the algorithm can be manually readjusted. Such optimizations should be performed regularly by online retailers or software vendors. Thus, ML-based pricing is not only more accurate from the start, its performance also improves with regular training over time.

To start the learning curve from a solid level, the initial programming of the algorithm is critical. In-house developments often require more intensive testing here as compared to specialized pricing software that has been developed by experts and whose efficiency can be quickly proven.

7Learnings offers A/B testing to its customers to prove the higher effectiveness of their software compared to in-house or competitor applications.

Training the team

Selecting a powerful ML-based pricing application is essential to the success of dynamic pricing in e-commerce. However, the expertise of the staff programming and training the application is just as critical. Companies need experienced Data Scientists, Engineers and Pricing experts if the investment in Dynamic Pricing strategy is to pay off.

The educational background and job title of the employees are less important, as long as they bring in-depth know-how in data engineering and machine learning.

Data Engineering

Ideally, companies should have a team of data scientists. Their work provides the foundation for the effectiveness of ML pricing. The tasks of a data science team include:

  • Data selection: Data scientists identify all data relevant to the implementation of the dynamic pricing strategy. They collect data from various sources, are able to process different data formats and data of any size.
  • Data Preparation: Data scientists must prepare the collected data so that it is available in a central data pool or other repository and can be used by the ML algorithm. The data must be integrated so that it updates automatically at regular intervals.
  • Build a data platform: Most companies store historical operational data in a data warehouse. If one is not yet in place, data scientists are responsible for creating these structures. They also ensure that this in-house data flows smoothly into the central data pool accessed by the pricing system.
  • Maintaining the data ecosystem: A company’s data structures must be monitored to ensure that interfaces function, data updates take place, and applications work with the correct data.

Machine Learning

If companies want to develop their own machine learning-based pricing application, they need to employ experts who know how to use the technology.

  • Developing and validating a machine learning model: The task is to build an algorithm that can make predictions for various events and trends, for example, churn rates, customer behavior on marketing campaigns, purchase probabilities, and certain conditions. The algorithm can work with the help of data prepared by data engineers. ML experts must validate it in test scenarios and minimize errors with the help of training or data corrections.
  • Deployment: If the algorithm has been sufficiently trained, it can be deployed in live operations. Regular training optimizes the results and ensures that the ML algorithm always reflects the company’s goals.

Above all, the development of the algorithm and its training can take many months in the case of in-house development. Companies that want to save time and money can turn to third-party applications – and save themselves the trouble of learning the background of machine learning know-how.

Regular optimization and governance

When the pricing application is finally in use, it is time to train the pricing and category managers. After all, the added business value of the application depends to a large extent on how well these managers manage to work with the systems.

Managers need to understand how the algorithm works so that they trust it, even if they can’t understand every decision it makes. Because in the vast majority of cases, the recommendations and predictions of dynamic pricing systems are more accurate than human assessments. At the same time, managers should maintain a healthy skepticism. Instead of blindly following the algorithm, it is possible to set limits for it and test suggestions before executing them in real operations. Experience shows that companies that combine the experience of their managers with the reliability of the pricing algorithm achieve the best results.

In which industries is dynamic pricing used?

Dynamic pricing can be used to add value in e-commerce in many industries. In some, ML-based price optimization is already widely used.


Airlines were among the pioneers of dynamic pricing. Airline tickets cost different amounts in their online booking portals for the same route at different times for different customers. Since users were already used to ticket prices varying according to the time of booking from travel agencies, it was easy for airlines to integrate a dynamic pricing strategy into their online presences right from the start. Users accept the fluctuating prices and even see potential savings in them.


Dynamic pricing strategies are increasingly used in e-commerce. They have long been commonplace at major online retailers such as Amazon or Ebay. Their pricing algorithm recognizes a customer’s purchase frequency. Regular customers are shown higher prices than new customers, who they want to retain with low prices.

For fashion mail order companies that want to maximize their profit, a dynamic pricing strategy is nowadays without alternative. Algorithms calculate customers’ willingness to buy and strategically apply discounts and price increases. Customers today want to find bargains on a regular basis. Retailers give away sales if they only offer generous discounts on current fashions at the end of the season in order to empty their warehouses.

This is where ML algorithms come in: they independently calculate favorable times during the year to sell swimwear, winter coats and the like with low discounts in the best possible way. This improves the company’s sales and at the same time satisfies customers’ need for regular discount campaigns.

Hospitality Industry

In the hotel industry, it has always been common for prices to fluctuate according to season. Assessing customers’ willingness to pay correctly is critical to the success of the business. With the help of a dynamic pricing strategy and machine learning, hotels can evaluate their own historical data, keep an eye on the competition, and also take external factors such as weather and social events or events in the region into account in their pricing. Machine Learning does a much better job of integrating this multitude of data than rule-based pricing optimization systems.

AirBNB, for example, follows such a dynamic pricing strategy: hosts can use an ML-based tool that gives them suggestions for accommodation prices based on variables such as season, day of the week, condition of the rental property, and hotel prices in the surrounding area.

Passenger Transportation

Passenger transportation prices have always varied depending on what time of day and month a customer books a ride. It is no different for cab alternatives Uber and Lyft. Here, too, rides just after midnight on December 31 are significantly more expensive than rides at lunchtime on an uneventful weekday of the year. Using a dynamic pricing strategy, providers can further optimize their prices without compromising demand.

For example, Uber uses machine learning to display forecasts of what demand will be like at certain locations at certain times. The clearer demand exceeds the number of available drivers in an area, the more expensive fares become. The pricing algorithm determines whether a rider will be paid a premium, and if so, what that premium will be, fully automatically. It monitors the relevant factors and updates prices in real time.


Price optimization has long been used in e-commerce. But previous rule-based pricing applications are expensive because they require a high level of manual effort – for selecting data and maintaining pricing rules. Despite all the effort, the rules are inaccurate. As a result, revenue and profit increases fall far short of what is possible using current technology.

The new gold standard in price optimization is Dynamic Pricing based on Machine Learning. With its help, online retailers can significantly reduce application configuration and monitoring efforts, incorporate significantly more factors into the calculation of optimal prices, and determine valid forecasts for a wide range of events. In some industries, companies that do not use this technology are already at a competitive disadvantage. This trend will spread to other industries.

Many companies have reservations: they think that only corporations with their own experts in artificial intelligence and machine learning can afford the smart applications. But these reservations are unfounded. In the meantime, there are ML applications for dynamic pricing strategies that companies can configure for their individual use case, completely without AI expertise.

7Learnings offers such a solution for dynamic pricing in e-commerce and promises customers a number of benefits.

How can 7Learnings benefit your business?

  • Easy integration: our application is compatible with any backend system, so no expensive technical adjustments are needed upfront.
  • Latest technology: AI-based application gives you a competitive edge and is an investment-proof solution.
  • Approx. 10% sales and profit: You will see the effect within a few weeks. Sales and profit increases of up to 10% are realistic.
  • Implementation support by experienced Data Scientists: We ensure a stress-free introduction and support you until go-live and beyond.
  • Accurate price optimization and forecasting: Properly trained, the AI algorithm beats any human estimate and saves you from costly wrong decisions.
  • Automatic optimization to business goals: You define your goals, the AI algorithm does the rest. It optimizes your prices fully automatically and for the fastest possible target achievement.