I have previously lightly touched on how pricing and promotion were key domains of an e-commerce data strategy, but never got around to explaining in details how to effectively act on it.

Concepts of elasticity of demand

Experimentation — in-Session conversions

Facebook allows us to set up split tests to test the performance of different ad sets. These split tests allow us to get a sense of the elasticity of demand across various metrics, using in-session performance. Similar approaches can be performed on an existing customer base, using a CRM or Marketing automation platform.

Looking only at users who have initiated a session directly after having been exposed to an advertisement, and only using that session information can give us an excellent proxy performance as to the type of uplift we can expect from the different pricing.

There are a few caveats to this, however:

1. it requires that the test be done in a small enough subset of the population, so that there is no massive spillover effect across the offers

2. it requires that your target customers aren’t predominantly bargain hunters, scouring the internet for specific discount codes or offers

3. it does require that the product you offer to be purchased through some “impulsive” type of purchase, we are not talking about an experiment here that would work on a car or house purchase for instance

The raw number that we obtain from running this type of experiment allows us to get an excellent first pass estimate on the impact of a price change on the overall conversion rate. It is however, important to understand that this type of experiment allows us to observe the blended behavior of different customer groups based upon their price sensitivity and that each increase in discount rate allows us to reach another customer segment strata.

Offering higher discount rates will, for most goods, increase the overall conversion rate, but it also means that we are reaching different customer groups with potentially different behaviors. Behaviors, which might not be fully compatible with the experience on the website. Getting an understanding of the full purchase funnel can help us unlock the true elasticity of demand.

A few key: can help us get a better grasp on the different key metrics and help us try to understand what could happen within the conversion funnel, and help us identify where we could bootstrap further the price-elasticity of demand.

• First-order conversion — # sessions, if all ad-sets are constrained to the same budget, the relative number of sessions attributed to a specific campaign (a, b, or c) should give us an indication as to the impact of lower pricing in attracting a new customer to the site.
• Second-order conversion — Add to cart/session, add to cart rate provides us with an indication in on the purchase intent of the traffic brought to the site.
• Third-order conversion — Purchase / add to cart, drop-in that metrics from reaching a broader base of customers, would typically point to checkout issues, such as a payment or shipping options gap. Perhaps that new customer segment that we are pushing onto the funnel is expecting a cash on delivery option?

There exist other variations of this type of experiment, such as a “saved offer” variation trigger, where a popup offer with a specific discount percentage might be offered to a subset of the population on leaving the website. These tend to shed light on slightly different behavior and certainly doesn’t quite capture the impact on customer acquisition.

Experimentation — product/price variation

It is possible to run experiments around demand elasticity in a different manner when a decent catalog size exists. Assuming that the products on offer have what is called a low substitutability, i.e., that they are not easily interchangeable, and that within a category behave similarly, a variation in product price of certain items up and of other items down can help estimate the impact of pricing on demand.

Statistical techniques, such as difference in difference, can help estimate the causal relationship between price and demand with this type of experimentation.

Price discrimination

As we saw before, setting up a higher discount rate allows us to reach a more substantial proportion of the population, but at the cost of offering a higher discount than intended for certain strata of customers. Being able to differentiate pricing across different type of customers, help us maximize our revenue to the customer’s willingness to purchase.

Customer groups

Optimal pricing can be further achieved through the use of a customer group segmentation. Typical customer groups used for this purpose are students and retired. Most e-Commerce platforms, such as Shopify, provide native support for offering discounts to specific customer groups.

Behavioral — through segmented product offering

One of the ways to be able to able to price discriminate effectively is to offer slightly different products for different audience/target segments.

Adidas, for example, has applied behavioral, discriminatory pricing by introducing limited edition at a significantly higher price than its primary collection. Limited Editions sneakers easily retail at over 300e vs. less than 100e for their collection.

Apple is another company that has segmented its’ product offering, an already premium offering to be able to further price discriminate. Each of its’ main phone products is further segmented into different variants by memory size, for instance.

Although a bit less reliable, and sometimes impractical, it is possible to set up some experiments around a segmented product offering. Depending on the situation, it might be possible to create dummy products to measure the overall interest with proxy metrics such as product detail page views.

On the other hand, if we wanted to backtrack the impact of offering certain items, hiding the specific product, or using an out of stock message can provide an estimate of the effect of this offering. If we were to take the example of apple as previously mentioned this is an approach that might be feasible for certain’s model memory variants, but might not be practical for testing the impact of a full model category.

Behavioral — through different terms

Besides offering different products, service terms also enable price discrimination, offering faster/slower delivery, lower return window, early access to new products. These terms and conditions might not necessarily cost you significantly more, or anything at all, however customers, who value these product benefits would be more willing to pay a premium.

Prime customers on Amazon, might only consider items eligible for Prime delivery for their orders, doing so will for a certain portion of their orders increases the average ticket price.

Pricing across the product lifecycle, typically falls into one of two categories high introductory price for products that do not tend to be bought at a high frequency, think of consumer electronics goods like a computer or a PlayStation, or low introductory offers in order to raise awareness for repeat product purchases.

For the first category, it is possible to further segment and extract an additional premium for certain customers by providing early access terms to these items. Think of sales as a particular case of this pricing, customers who buy before the sales are essentially paying for some sort of early access to these items.

For the second category, strategies such as free trials and discounted offers for new customers are typically used. Identifying what should be the right offer to customers for product introduction, is a constant juggle between three different sets of metrics, CPA, CLV, and conversion rates.

Behavioral — Affiliates

Affiliates can help drive traffic to the website of particular customer subsegments. Price comparison websites and coupon code websites are able to bring price-conscious customers.

Optimizing the pricing through dynamic pricing to reach this kind of customers helps further price discrimination.

Cost / Profit pricing

Even the most aggressive market penetration strategy has some sort of cost/profit pricing strategy, if only to manage an overall budget.

Traditional economics pricing strategies include contribution margin, margin pricing, cost plus, pricing, average cost pricing, penetration pricing. And while for some niche e-commerce shops, these may be valid, for a large number of e-commerce shops, these pricing strategies do not prove effective.

For them, the game is to control cost while gaining scale, ultimate it is scale that will bring profit. Scale allows for better amortization of fixed costs, more efficient processes, more buying power towards suppliers. All of these are driving a lower cost structure. The Amazon flywheel is the iconic example of this approach:

This approach, however, requires a compass to get an understanding of when we could finally be profitable, taking a lower price strategy to drive scale. Entitlement values, can help get a better understanding of what could be the cost structure to target at scale and use that information in order to set the pricing of products to reach this goal.

Bundling

The original goal of bundling was to extract what is called “Consumer surplus,” i.e., the difference between what the customer would have been willing to pay and the price that the customer ended up paying.

The theory here is that most consumers will tend to value a product in a bundle more than others. If the price has been individually set to minimize each individual items’ consumer surplus, there would still be room to up-sell customers without impacting the sales of each individual item by offering a lower pricing to those having already purchased another item.

For instance, I might be willing to pay $3 for coffee, but only $2 for a croissant. Wonderful coffee, normally sells each at $3 each. Under normal circumstances, I would not buy the croissant. But if an offer for a $5 breakfast combo was made, I would be tempted. This offer could reduce the overall sales turnover of customers who normally purchase both coffee and croissant together by decreasing the total price, but this would, in most cases by exceeded by the increase in overall sales made.

For e-commerce activities bundling has certain additional benefits:

• Lower operating costs: Bundle helps driving better cost structure by helping to amortize fixed costs, drive lower logistic cost/unit (more items in a box for instance), simplifying operations and put your site in a better negotiating position versus suppliers (more items sold overall)
• Increased exposure: bundles can increase exposure on different marketplaces through the creation of new SKUs. The bundle may also be returned on a higher number of search queries as containing more items as search terms.
• Pricing Flexibility: Allows for more flexible pricing, when competitors use a price matching strategy or items have a minimum advertised price policy

Bundles can have a significant impact on e-commerce sales, having a grasp on their profitability, their contribution to improving the cost structure as well as how to price them appropriately, can be critical in setting up a proper pricing strategy.

Competitive Pricing

Information on the internet travels quite fast, and a lot of vendors attempt to price match their known competitors on a series of products. Amazon offers a way through their API to get the price of a given SKU, and multiple solutions exist to scrape prices and match offers.

Competitor’s prices will impact the amount you will be able to sell, and having both the visibility and the ability to act on this new information can be the difference between a sale and a lost sale.

There are different ways to handle the price matching. The base of price matching involves about three variables: a standard price, a price threshold that represents the lowest price you would sell the product at as well as an acceptable price difference between you and your competitor/s . More advanced competitive pricing strategies would include only matching prices for users coming from certain channels, such as price comparison websites, where there is evidence that a user is particularly price sensitive.

Mixed Approach

In most cases, a mixed approach will be the way to go. The complexity will lie in getting the right understanding as to how each approach impacts the others. As it is the case when dealing with any complex systems taking a stand towards continuous improvements and experimentations will be key to identify what pricing will be right for your business.

‍