Price Optimization

Price optimization is the use of mathematical analysis by a company to determine how customers will respond to different prices for its products and services through different channels. It is also used to determine the prices that the company determines will best meet its objectives such as maximizing operating profit. The data used in price optimization can include survey data, operating costs, inventories, and historic prices & sales. Price optimization practice has been implemented in industries including retail, banking, airlines, casinos, hotels, car rental, cruise lines and insurance industries

Retail Pricing Strategies

One major common denominator that runs through all of the pricing decisions made by retailers is the concept of “markup”. Markup is simply the difference between the cost of the product to the retailer and the price at which the product is sold by the retailer divided again by the retail price. It is usually expressed as a percentage figure, so the calculation is made like this:

*Retail price minus cost price divided by retail price*

So if your item cost is Rs. 4.00 and you sell it for Rs. 10.00, you would calculate markup as:

(Rs. 10.00 – Rs. 4.00 = Rs. 6.00) /Rs. 10.00 = .6 or 60%

Markup is a concept that every retailer understands and factors in consideration somewhere in every pricing strategy.

One of the most traditional retail pricing methods is called keystone pricing. Keystone pricing is simply the retailer doubling the cost amount to arrive at a 50% markup. For example, if an item costs a retailer Rs. 3.00 to buy, the retailer will set the price at Rs. 6.00.

Premium pricing is another retail pricing strategy. In this method, the retailer takes a larger markup on a product in order to establish higher perceived value for that product. For example, a new designer brand being introduced by a department store might see 70%- 80% markup levels initially (especially if the store has an exclusive arrangement with the vendor so no competitors have the same products).

Discount pricing is a prevalent retail pricing strategy. Retailers such as Kmart, Target, Wal-Mart and others pioneered this method, setting their sights on moderate-priced competitors and setting prices below them. Retailers can expect markups to drop below 20% and even lower depending on the product category. The latest wave of discount retailers have simplified the discount strategy even further by featuring entire stores with goods all priced at Rs. 1.00 or even 99 cents.

Psychological pricing refers to taking advantage of human perception to convince customers of a more attractive price. For example, instead of placing a price tag of Rs. 200 on an electronic product, a retailer may mark the item at Rs. 199. Or a dress shirt may be marked at Rs. 29.99 instead of Rs. 30.  Although it is a small difference in price, it is believed that people pay more attention to the first number in the price.

Another common retail pricing strategy is bundle pricing. This term refers to grouping multiple items and pricing them together. There are many variations of this strategy as well. “Twofor” pricing (2 for Rs. 10), “BOGO” (Buy One Get One Free), “Get 50% OFF the Second Item”, etc.

The last retail pricing strategy we will discuss in this section is tiered pricing. Tiered pricing is the practice of establishing set price-points within a product category and marking all the products in that category at those price-points. For example, men’s ties from different manufactures could be priced at Rs. 11, Rs. 12, Rs. 16, Rs. 18, Rs. 22 or Rs. 25 depending on their different costs. In a tiered pricing scenario, a retailer may offer these ties at Rs. 10, Rs. 15 and Rs. 20 to simplify their price structure.

Various other strategies are as

• Penetration Pricing: The starting price is set extremely low- lower than is normal- in order to gain market share and persuade customers or clients to choose their product over the competitors’ products that they may normally purchase. After the business has achieved this, increasing its market share, it is able to choose a more profitable price.
• Premium Pricing: Where there is a unique brand, the business charges a high price in the knowledge that they have a substantial competitive advantage. By marketing their product as higher in value, customers may perceive the product to be higher in value, whilst also having a higher profit margin.
• Economy Pricing: Low prices. Costs are kept low in production in order to keep down the selling price. Other overheads, such as marketing their product and branding, are also low cost. This thus allows the product to be priced lower in comparison to competitors, and still retain a healthy profit margin.
• Value Pricing: This strategy is used where external factors force business’ to provide value products and services e.g. a recession or increased competition. This strategy makes it seem like you are getting a lot for your money.
• Price Skimming: This is where a business chooses a high price for their product, which is often newly launched, and has little or even no competition. Consumers will be willing to pay the higher price in order to have the advanced product. This strategy does not, however, last long due to a gradual increase in competition over time.
• Geographical Pricing: Simply, geographical pricing is a strategy whereby the seller will alter the price depending on where the buyer is located. This may be due to shipping costs, different operating costs, or competitors pricing in that particular location perhaps due to what the customer is willing to pay.
• Product Line Pricing: When there is a range of products or services, the customer expects to pay for what is fair incrementally over the range. In this way, the pricing of this strategy rarely reflects the cost of making the product. For example, in an ice cream cabinet, some flavors could be a lot more expensive to produce than others, however consumers expect to pay the same price for all of the ice cream, and more only if the quantity increases.
• Promotional Pricing: This strategy is well-known and is where prices are reduced, or offered for free as part of a deal. This is adopted in order to attract customers and increase sales volume. This includes sales, buy one get one free, special offers, vouchers, discounts etc. Although bringing in more customers, promotional pricing also reduces profit margins.
• Product Bundle Pricing: Another form of promotional pricing. Often used to shift old stock or slow-selling items by selling them together (at a lower price) with better selling items.
• Captive Product Pricing: When you buy a product for a cheap price but the essential components that are required to make the item useful are sold at an expensive price.
• Optional Product Pricing: Including pricey optional extras to offer with a low cost product or service.

Price Optimization Techniques

Price optimization utilizes data analysis to predict the behavior of potential buyers to different prices of a product or service. Depending on the type of methodology being implemented, the analysis may leverage survey data (e.g. such as in a conjoint pricing analysis) or raw data (e.g. such as in a behavioral analysis leveraging ‘big data’). Companies use price optimization models to determine pricing structures for initial pricing, promotional pricing and discount pricing.

Market simulators are often used to simulate the choices people make to predict how demand varies at different price points. This data can be combined with cost and inventory levels to develop a profitable price point for that product or service. This model is also used to evaluate pricing for different customer segments by simulating how targeted customers will respond to price changes with data-driven scenarios.

Price optimization starts with a segmentation of customers. A seller then estimates how customers in different segments will respond to different prices offered through different channels. Given this information, determining the prices that best meet corporate goals can be formulated and solved as a constrained optimization process. The form of the optimization is determined by the underlying structure of the pricing problem.

If capacity is constrained and perishable and customer willingness-to-pay increases over time, then the underlying problem is classified as a yield management or revenue management problem. If capacity is constrained and perishable and customer willingness-to-pay decreases over time, then the underlying problem is one of markdown management. If capacity is not constrained and prices cannot be tailored to the characteristics of a particular customer, then the problem is one of list-pricing. If prices can be tailored to the characteristics of an arriving customer then the underlying problem is sometimes called customized pricing.

Price optimization uses data analysis techniques to pursue two main objectives:

• Understanding how customers will react to different pricing strategies for products and services.
• Finding the best prices for a given company, considering its goals.

Current state-of-the-art techniques in price optimization allow retailers to consider factors such as:

• Competition
• Weather
• Season
• Operating costs
• Local demand
• Company objectives

to determine:

• The initial price
• The best price
• The discount price
• The promotional price

Price optimization vs dynamic pricing

It is important to differentiate price optimization from dynamic pricing, given that these terms are sometimes used as synonyms. The main difference is that dynamic pricing is a particular pricing strategy, while price optimization can use any kind of pricing strategy to reach its goals. Despite having many advantages and being quite used, dynamic pricing has some disadvantages when used in an extreme way.

Simply put, using a dynamic pricing strategy, retailers can dynamically alter the prices of their products based on current market demand. In contrast, price optimization techniques consider many more factors to suggest a price or a price range for different scenarios (e.g. initial price, best price, discount price, etc.).

We all know and somehow accept because it seems reasonable, that the price of a hotel room or a plane ticket varies according to the season, the day of the week or the anticipation with which we booked. However, when prices change too fast – sometimes in the course of a few hours – some customers might have the feeling that prices are unfair or that the company is practicing price gouging. Dynamic pricing is, therefore, a strategy to be used with caution.

Machine Learning and retail price optimization

The pricing strategies used in the retail world have some peculiarities. For example, retailers can determine the prices of their items by accepting the price suggested by the manufacturer (commonly known as MSRP). This is particularly true in the case of mainstream products. Another simple strategy is keystone, which consists in defining the sale price as the double of the wholesale price or cost of the product.

While these and other strategies are widely used, ML enables retailers to develop more complex strategies that work far better to achieve their KPIs. ML techniques can be used it in many ways to optimize prices. Let’s have a look at a typical scenario.

Machine Learning Example

Imagine an online or brick-and-mortar retailer who wants to estimate the best prices for new products for the next season. The competition is hard, so their prices and promotions need to be taken into consideration. Therefore, the retailer adopts a widely used strategy: competitive pricing. Simply put, this strategy defines the price of a product or service based on the prices of the competition.

Let’s see the steps needed to develop a ML solution for this use case.

Gather input data – First of all, we need data. To train Machine Learning models, it is necessary to have different kinds of information:

• Transactional: a sales history that includes the list of the products purchased and, eventually, the customers who purchased them.
• Description of the products: a catalog with relevant information about each product such as category, size, brand, style, color, photos and manufacturing or purchase cost.
• Data on past promotions and past marketing campaigns.
• Customer Reviews: reviews and feedback given by customers about the products.
• Data on the competition: prices applied to identical or similar products.
• Inventory and supply data.
• In the case of physical stores: information about their geographical location and that of the competitors.

Depending on the set KPIs and the way of modeling the solution, some of this data may not be necessary. For example, if there is little or no information about customers, which is sometimes the case for brick-and-mortar retailers, the model can nonetheless be trained.

In contrast, information about the competition is crucial for a competitive pricing strategy. In many cases, it is even possible to connect via APIs to this information or monitor it online.

Define goals and constraints – The next step is to define the strategic goals and constraints.

Retailers may pursue a unique, clear objective of profit maximization. However, they may also be interested in customer loyalty (e.g. increasing the net promoter score or the conversion rate) or in attracting a new segment (e.g. young people).

Restrictions may be of legal nature (e.g. if some type of control of sale prices is carried out), they may have to do with the reputation of the company (e.g. fearing a bad image for applying favorable prices only to a certain segment of customers) or be related to physical aspects such as the capacity of a store or the average time of supply.

Each particular scenario will impact the way the problem is modeled. It is possible, and usually very interesting, to test different scenarios for the same retailer, which implies using different models.

Modeling and training – In this step, the data previously gathered is used to train the ML models. There is a wide variety of models that can be used in price optimization. Historically, Generalized Linear Models (GLMs) have been used (in particular, logistic regression. However, for a few years, more complex and powerful methods have been developed. For instance, depending on the volume of data available, it could be possible to use Deep Learning methods.

In this case, in which we are dealing with new products for the next season, there is an additional difficulty since there is no previous product data. The interesting thing is that the ML models will know how to find similar products and be effective despite not having specific prior data. The same happens in the case of retailers that sell rare or exotic products.

Execute and adjust prices – Once the model is trained, prices can be estimated for the new products. Depending on the modeling, the estimate may be an exact price or a range. The prices obtained by the model can be subsequently adjusted manually by the retailer.