Price Optimization Models for Retail Enterprise

How Price Optimization Models Boost Retail Enterprises’ Revenue

There is a difference between the price optimization flow of SMBs and that of enterprise retail businesses. SMBs, or small and mid-size businesses, usually optimize their pricing strategies by adopting dynamic pricing algorithms. Retail enterprises have already automated everything there is to automate. What they need are laser-focused improvements: little changes in prices which significantly change their sales volumes.

The article covers the most advanced price optimization techniques for retail enterprises.

Alexander Galkin

Price Optimization Solutions for Retailers

Optimization of Portfolio Pricing

The Problem

Retailers need to ensure they factor in all the interrelations between products when changing prices. It is challenging to determine which products and how many of them require price changes. Besides, sometimes a price change can damage KPIs instead of fueling them.

By changing a price on one product, the retailer triggers a chain reaction across a group of ‘neighboring’ products in the perception of the shoppers. Thus, the fine-tuning of portfolio pricing architecture becomes a rocket-science task, given the thousands of latent relationships between product sales.

The Solution

In order to maximize the KPI of choice on a portfolio level, be it volume, revenue, or profit, the retailer needs optimization models that factor in both the elasticity of the product and its cross-elasticity with related products. An elaborate equation combines all of these coefficients and provides a solution of optimal prices across the retailer’s portfolio.

With such an optimization model, the retailer implements the so-called 'differentiated' pricing recommendations across selected products in order to predict and maximize the portfolio target of choice – volume, revenue, or profit – for each SKU.

Thus, retailers change prices for each SKU individually, while taking into account the competitors’ actions and their own business needs and goals.

Data needed for this model: sales volume and price by product, and competitive prices for similar products (by day or week) for no less than three years.

Optimization of Prices on Individual Products

The Problem

Today, setting prices based on the retailer’s business goals alone is not effective anymore. As they do not live in a vacuum, businesses need to factor in a variety of factors which influence the price, including price elasticity.

Price elasticity is a non-linear relationship between the sales volume and price: the more elastic the prices, the more they influence the sales. It is difficult to calculate the right price that would maximize revenue (i.e., will maximize the price per product and the number of products sold) without knowing exactly which are the elastic categories of products, and which of them should be repriced.

The Solution

The retailer identifies price points to predict with high precision how many items of product shoppers will buy at different price points, and maximize the parameter of choice, be it volume, revenue, or profit for a specific product.

Factoring in the price elasticity by identifying relevant price points helps:

  • To simulate the ‘real life’ market with the help of elasticity curves for the products of choice by price points within an interval.
  • To make more precise sales volumes, revenue and profit (optional) predictions for specific price change scenarios.
  • To create a list of competitors who impact the sales of the retailer, and measure the power of the impact.

Data needed: sales volume and price by product, competitive retailer prices for similar products by period (by day or week) for the past three years.

Models’ Visualisation To Quickly Identify Opportunities

Price Partition

The Problem

It is difficult to predict if customers will perceive the price as justified, and easy to set either too high or too low a price. In both cases, the potential profit is lost.

The Solution

Products with similar value and set of features ‘stick together’ in a buyer's perception in clusters or segments. Shoppers are very specific about the minimum, average and maximum price they are ready to pay for products representing each specific segment.

Img legend: Scatterplot of the top 100 SKU. The X-axis is the price per piece. The Y-axis is the sales. Optional visualization—the size of the bubble is equal to sales.

Price optimization algorithms group products within a category into 'clusters,' the centers and external borders of which are visible on the chart. This graph immediately determines which price segment a particular product belongs to, and the price variations interval which is ‘safe’ for sales.

The price optimization formula and visualization, data about product sales and median price by product, by period (by day, week, or month) for the past year is needed.

‘Magic’ price points

A price point is a retail price that allows keeping a relatively high demand for a product.

The Problem

Buyers classify products into price segments or categories with clear “thresholds” according to their subjective 'value' indicator. The sales are maximized around the center of these segments and strive towards zero on their borders.

The seller’s assortment does not always take into consideration these patterns, which makes the sales suboptimal. Thus, retailers usually lose profit by increasing the price or offering a discount for “inappropriate” products.

The Solution

Price points in categories indicate the highest or zero sales for the products within a category. Knowing these “highs” and “lows”, the retailer can significantly increase sales by offering an optimal price for the right product.

Option 1. The top X of the SKUs are taken, and they are ranked by sales. Axis X - the price of the products. The Y-axis is sales per unit. Visual price breaks along the X-axis are thresholds of price segments.
Option 2. The sales of all SKUs in the market are taken. Products are grouped into price 'baskets', and then the 'top-X SKU' approach is applied.

With a price optimization system, the manager will be able to detect different price levels:

  • Those which maximize sales
  • Price “thresholds” which should not be crossed.

Also, this helps retailers avoid the trap of following their competitors entering the “dead zone” and losing sales.

The model requires the data about sales and prices (either median or weighted by sales) by product, by period (by day, week, or month) for the past year.

'Price Ladder' or Optimal Price Intervals

The Problem

As in the game of chess, the retailer needs to set optimal prices to neutralize the competitors or to milk their weaknesses. This task becomes complicated, given the proliferation of categories and the high dynamics of prices.

The Solution

To outplay competitors, the retailer needs to understand the category leaders “alignment”:

  • At which price points do shoppers buy them most often?
  • What is the price range: from the “premium” price of expensive competitors to the “convenience” price with a deep discount or a regular price in a discounter?

It is challenging to keep a hand on the pulse since prices are dynamic and change every day.

Buyers are willing to pay the same price for similar products. The same rule applies to the price range (max — a convenience price, min — a promo price). The leader's most frequently encountered price is the benchmark from which to build the optimal price. The optimal “spread” is the leader's interval from the maximum to the minimum registered price.

Img legend: This type of chart is a 'candle/ladder chart.' The three points represent the minimum, maximum and median price for the observed period. In the chart, there is either the top-seller in the category or the same product of different retailers.

The ‘price ladder’ chart allows retailers to indicate the following:

  • the current price points;
  • the price points to play;
  • the retailer’s price positioning against competitors;
  • typical discounts in the category.

Data for this price optimization technique: the minimum, maximum and median price by product by period (typically, a week or a month) by seller.

Optimal Promo Pressure

The Problem

Promotion pressure or volume per deal has significantly increased in recent years. Vendors (brands) and retailers often start “promo wars” which usually lead to the loss in sales or the market share. It is impossible to determine the strategy to use at the beginning: whether to activate or deactivate deep discounts, or offer low prices every day.

The Solution

Retailers should analyze the share of the volume sold per deal across the total category and among the top-sellers to evaluate promo pressure that would be optimal for the category.

Scatterplot of the top 100 SKU. The X-axis is the % of sales per deal (or the % of registered promo prices to all tracked sales cases). The Y-axis is for regular sales (or tracked sales). A scatterplot visualizing sales items and volume per deal will highlight the products that have optimal promo pressure, and those that are overselling with a discount.

After analyzing such a chart which highlights the “highs” and “lows” of promo pressure, retailers should choose the moderate path.

Data necessary for such a model: the minimum, maximum and median price by product by period (typically, a week or a month) by seller.


There are many ways retailers can improve their sales while keeping their margins healthy. All of these approaches require the adoption of price optimization processes, including data management, algorithmic decision-making, and expert supervision.

Competera uses the power of AI technologies to predict the impact of every pricing or promo campaign based on the data about all the transactions in the retailer’s pricing history, seasonality, customer behavior, and competitors’ actions. In addition, it factors in retailers’ business needs and goals to help them increase revenue.