This is a whitepaper I wrote for Peak on price elasticity. The objective was to provide sales reps with a short but detailed piece they could send to prospects to build credibility for our offer.
Artificial intelligence (AI) is transforming the world of retail and provides an incredible opportunity to build competitive advantage in these turbulent times. Every decision a retailer makes can — and we believe will — be informed by AI. This will result in unprecedented leaps in operational efficiency. Those that don’t invest today risk the same fate as those that ignored the e-commerce trend: they’ll disappear into the retail history books.
“Gartner® predicts by 2025, the top 10 retailers globally will leverage AI to facilitate prescriptive product recommendations, transactions and forward deployment of inventory for immediate delivery to consumers.”
To leverage the full transformational potential of this emerging technology, retailers need to invest in building their own AI. Not something generic or off the shelf, but AI that is specific to their unique business, made up of composable blocks that can be re-configured as needed. This will allow AI to ingest data from all systems and push decisions back to all systems. Putting AI at the heart of the business in this way provides flexibility and creates opportunities to infuse AI across the entire retail value chain.
For any retailer looking to build their own AI “nervous system”, pricing is a great place to start. Because pricing is by nature a numerical pursuit, it lends itself well to the data-driven and probabilistic approach of AI. Building AI for pricing can also help lay the foundation for additional AI use cases in a retail business.
Pricing is a critical factor that can significantly influence a business’s success. Finding the optimal price point for a product or service can lead to increased revenue, market share and profitability. Understanding price elasticity plays a crucial role in determining the appropriate pricing strategy.
In this article, we’ll explore what price elasticity is, delve into traditional approaches to calculating it and then shift our focus to more modern techniques that leverage the power of AI. We’ll also discuss best practices for using price elasticity models effectively, including leveraging data from different sources, isolating the effects of price on demand, selecting the right algorithm and ensuring effective model training and tuning.
Pricing strategies over time
Business leaders have had to consider pricing strategy since the days the first market traders hawked their goods.
1970s – 1980s: Cost-plus pricing
Add a fixed markup to the cost of a product to determine the selling price. Simple, but limited in scope.
1980s – 1990s: Competitive pricing
Retailers started to take competitor prices into account when setting their own prices. Leads to price wars.
1990s – 2000s: Price segmentation
Retailers segment customers based on demographics, purchase history and other factors, allowing them to offer different prices to different customer groups. Data is a limitation.
2000s – 2010s: Dynamic pricing
The rise of e-commerce and the availability of ‘real-time’ data allowed retailers to adjust prices based on market conditions, customer behavior and inventory levels. Availability of data and technology adoption is a limitation.
2010s – now: Personalized pricing
With the advancement of analytics and AI, retailers can now offer personalized prices and/or promotions to individual customers based on their preferences, shopping habits and willingness to pay.
Now and in the future: AI-powered pricing
Leveraging advanced AI algorithms, tuned to unique business logic and guardrails, to make pricing decisions across the full product lifecycle based on a huge number of factors and features, such as customer behavior, competitor actions, market trends and product attributes.
Understanding price elasticity
Price elasticity measures how sensitive demand for a product or service is to changes in its price. It quantifies the percentage change in quantity demanded resulting from a 1% change in price. If demand is said to be “highly elastic,” a small change in price will lead to a significant change in demand, indicating that consumers are sensitive to price fluctuations. On the other hand, if demand is “inelastic,” changes in price will have a relatively minor impact on demand.
The formula for price elasticity is:
Price Elasticity = (% Change in Quantity Demanded) / (% Change in Price)
A price elasticity greater than one indicates elastic demand (sensitive to price changes), while a value less than one suggests inelastic demand (insensitive to price changes). A price elasticity of exactly one represents unitary elasticity, where changes in price and quantity demanded are proportionally equal.
Traditional approaches to calculating price elasticity
Historically, businesses have relied on various traditional methods to estimate price elasticity. Some common approaches include:
Percentage change method: This straightforward method involves calculating the percentage change in quantity demanded resulting from a percentage change in price. While simple, this method may not capture complex demand patterns and may not consider other factors influencing demand.
Regression analysis: Regression analysis is a statistical method that helps identify relationships between variables. By analyzing historical data on price and quantity demanded, businesses can estimate the price elasticity coefficient using regression models. However, traditional regression may not be able to handle non-linear relationships and interactions between multiple variables effectively.
Consumer surveys and experiments: Businesses often conduct consumer surveys or experiments to directly assess how changes in price impact demand. These methods can provide valuable insights, but they may be time consuming, costly and may not fully capture real-world market dynamics.
Modern approaches: Machine learning for price elasticity
The age-old adage that high prices scare away buyers is of course still true, but determining the effect of price on sales independent of other variables remains elusive without the help of some sophisticated data science. To accurately estimate price elasticity, it’s crucial to isolate the effects of price changes from other demand drivers. This involves controlling for factors like marketing activities, seasonality, stock and changes in the product itself to ensure that the model attributes demand changes specifically to price variations.
For example, stock availability and competitor activity can mask the actual price elasticity of your product. If you plot your prices and promotions over time against sales history for the same period, the trends you observe may not tell the whole story. It’s important to set promotional prices with the full knowledge of how price impacts demand in your market and category.
With advancements in machine learning, businesses can now leverage more sophisticated approaches to price elasticity that can process vast amounts of data, identify complex relationships and provide more accurate predictions.
Some popular machine learning techniques for price elasticity include:
Gradient boosting and random forest: Ensemble learning methods like gradient boosting and random forest can handle non-linear relationships and interactions between variables. These methods are particularly useful when dealing with large and diverse datasets, making them valuable for capturing real-world demand patterns.
Econometric models with machine learning: Combining traditional econometric models with machine learning techniques can lead to powerful hybrid models. For example, using regression models with feature engineering derived from machine learning algorithms can provide more accurate elasticity estimates.
Best practices for effective price elasticity models
While machine learning offers significant advancements in estimating price elasticity, using these models effectively requires careful consideration of best practices.
Leveraging data from different sources: Price elasticity models should incorporate data from various sources, including historical sales data, customer behavior data, marketing spend, competitor pricing and economic indicators. Combining diverse datasets can enrich the model’s understanding of demand drivers.
Choosing the right algorithm: There is no one-size-fits-all algorithm for estimating price elasticity. The choice of algorithm depends on the complexity of the data, the presence of nonlinear relationships and the overall business objectives. Careful experimentation and model evaluation can help identify the best algorithm for a specific pricing problem.
Model training and tuning: Machine learning models require proper training and tuning to perform optimally. Splitting the dataset into training and validation sets helps evaluate the model’s performance on unseen data. Hyperparameter tuning ensures that the model is optimized for the best possible price elasticity estimates.
Continuous monitoring and updating: Market dynamics change over time, and consumer preferences evolve. To maintain accurate price elasticity models, they should be continuously monitored and updated with fresh data. Regular retraining of the models can help keep them relevant and effective.
Setting the right price, the first time: Whether we’re talking about homewares, footwear or denim, setting the right initial price must be done using a range of data inputs. Whilst a traditional cost-plus technique applies a crude margin uplift to COGS, modern approaches take into consideration a wide range of datasets. For example, competitor pricing, customer expectations, demand signals, availability across the market and brand metrics can all be used to help determine the optimal initial selling price. Another consideration is the role of distribution partners and the possibility that your initial prices may actually only be a recommendation (i.e. MSRP).
Pricing in cohorts and categories: In traditional retail workflows, pricing decisions are often made only considering the impact on that individual product. There can be great benefits from optimizing price in cohorts. Data clustering can help determine the ideal cohorts to use, be that category, sub-category, price band, demand or some other combination of factors. You can then optimize the pricing and promotion of products so that the impact of changing the price of one product is considered in the context of the full cohort. Optimizing in a cohort means you can compare the simulated demand and metrics to your targets to understand if the selected strategy keeps you on-plan. This ensures individual price changes do not happen in a silo. Changing the price of one product can affect whether another product’s price is optimal — so to get the best results you need to consider the whole picture.
Pricing for customers: Traditional measures of price elasticity are aggregated by nature, with each product’s elasticity based on data from many customer interactions. These measurements therefore inherently lack the granularity that they could have if we were to measure price elasticity for every product and customer combination. A new and superior approach is to model “customer elasticity,” in other words to individual customers’ sensitivity to changes in price. With this information you can create price sensitivity scores for each customer and offer them promotions programmatically and contextually to achieve the desired result. If you understand the elasticity of a customer, you can understand how large a percentage off is needed to get them to buy.
Managing the proliferation of selling channels and competition: The multitude of ways for consumers to shop means pricing is more competitive than ever. In the future retail pricing will become increasingly contextualized, real-time and programmatic. Retailers must be equipped for more frequent price changes in order to keep up with competition in digital channels, requiring both a change in mindset and adoption of new technology. Retailers will need to monitor prices and demand across the market, adjust prices and operationalize those price changes both in store and online. In the fullness of time, a more data-driven and algorithmic approach to buying, merchandise planning and customer segmentation can also help build competitive advantage.
Conclusion
Pricing is a crucial lever for businesses to optimize revenue and profitability. Understanding price elasticity, which measures the sensitivity of demand to price changes, plays a pivotal role in devising effective pricing strategies. While traditional methods for calculating price elasticity have been valuable, the rise of AI has opened up new possibilities for more accurate and data-driven approaches. By harnessing the power of AI and adopting these best practices, businesses can make better pricing decisions and gain a competitive edge in their respective markets.