How the ‘average user’ has ruined ecommerce experiences

As consumers, we understand that every website we visit, as well as every interaction experienced on that site, is delivered as the result of untold hours of behind-the-scenes work.

No website is an accident; everything has been tested thoroughly to deliver optimal outcomes, especially in online retail.

The problem is that so much of the thinking that goes into these tests is based on an outdated mindset around an “average user” and what that theoretical user is going to like, click on or buy when they arrive at the site.

While the data and analytics may show averages, there is no real, definable “average consumer.” Therefore, building around the average is a less-than-optimal way to construct a site experience.

The ‘Super Pareto’

Everyone is familiar with the Pareto principle, better known as the 80–20 rule. Today, some leading online retailers may be seeing a new “Super Pareto” emerging, where typically fewer than five percent of the users contribute more than 90 percent of the revenue. Call it the 95–5 rule.

If ecommerce marketers understood this 95–5 rule as the new normal and had the tools to identify the five percent, it would fundamentally alter how ecommerce sites are delivered and experienced by consumers. In today’s competitive retail economy, sites must be constructed to emphasize conversions with that crucial five percent, while focusing on product discovery for the remaining 95 percent.

Illustrating the Super Pareto with product sorting order

An easy way to exemplify this problem is by looking at the sorting order of products in the category or search results page, which plays a big role in conversions. Retailers who tailor their sorting order to each user could see a significant uplift in conversions and revenue.

Today, the common sorting approaches are price, from lowest to highest and highest to lowest; newest items; relevant items; best selling; and best rated. When building toward an “average” user, a retailer could decide to pick a default sorting order that may result in the highest economic outcomes, and then apply that sorting order across the entire site.

That retailer may find that sorting in terms of highest to lowest price yielded more revenue on average and then rush to apply that sorting order to all users.

But a proper segmentation schema will segment users based on information like traffic source, prior visit behavior, past purchase history and conversions, and revenue from this segmentation will invariably lead the retailer to draw different conclusions.

To put it very simply, picking a default sorting order is a bad idea. Ecommerce managers are leaving money on the table if they go with that approach.

Not only does the best-performing sorting order change for each customer segment, but it also changes based on other contextual factors like geography, weather, day of the week and more. It is simply not possible for an ecommerce manager (or team) to pick a winning sort order and deploy it to all users, not even at the customer segment level.

Automated algorithms & machine learning

Enter machine learning. The permutations and combinations of the default sort order have already become a problem that is not possible for a human being to address all alone.

The solution lies in machine-learning algorithms that constantly collect all user data and signals and use that information to deliver the best potential sorting order for that particular customer. This is vital, because even consumers who fall into the same audience segments may respond differently depending on where they arrive from.

What if the same retailer personalized the sorting order for each user comprising the “fitness enthusiast” segment as soon as they landed on the page? By leveraging user-specific behavioral data, that retailer could create targeting conditions for users that fall into a “price-sensitive” segment (i.e., consistently manipulating the sorting order to display lower-priced items first) and automatically display a low-to-high sorting order to those users.

To dig even deeper, retailers can create more advanced segmentations and personalize the category page grid not just by sorting order, but according to user affinities for specific products and brands. If a user is a frequent buyer who has a demonstrated interest in price and a strong affinity for gray women’s Nike sneakers, then the category page can be dynamically rendered to show items that fit those exact criteria, already arranged by price, low to high.

Using affinity-based data to present products that each customer in an advanced segment is most likely to buy is a bona fide way of increasing loyalty, driving purchases and creating useful one-to-one experiences for valuable consumers, and not for “average” ones.

Retailers cannot simply increase their average by targeting more average consumers. To improve the outcome, they need to identify the consumers responsible for the most revenue and drive them toward purchase, while helping the remaining percentage discover new products.

The key is for retailers to treat each outcome as unique and dynamically respond to each consumer, rather than some predetermined (and possibly misinformed) sense of what will elicit a response from an “average” user.