AI in eCommerce: Hype or Reality?

by Zee Aganovic | Nov 13, 2017 | Commerce Strategy, System Integrations, Web Development

AI in eCommerce: Hype or Reality?

The whole business universe, eCommerce included, is buzzing about artificial intelligence, the capability of a machine to imitate intelligent human behavior.

Unfortunately there is no exact classification of what makes a solution an artificial intelligence solution. Any computer based system that is using data to make some decisions can be labeled as an artificial intelligence solution. Artificial intelligence means many different things for many different people.

This gives almost every software vendor a license to add an artificial intelligence ‘spin’ to the description of their product. This is counterproductive. Instead of attracting new buyers this just confuses the marketplace.

So, let us help you develop a pragmatic framework for understanding and use of AI in your eCommerce strategy.

Why AI in eCommerce?

Manufacturing, information systems and services are rapidly commoditized. The next industrial revolution is fueled by a need to deliver memorable customer experiences. It’s called the experience economy.

Remember Steve Jobs and his decision to recruit John Sculley as Apple’s CEO. People were confused about the logic of hiring Pepsi-Cola’s president who has nothing to do with technology to run the pioneering computer company. At the time only Steve had the vision and understanding that it is not about computers or cola but about the experience that customers have while using the computer or while drinking cola. That passion for memorable customer experiences is what eventually made Apple into one of the most successful companies in the history of mankind.

Delivering memorable customer experiences is a business of treating the right audiences with the right experiences at the right time.

Technically speaking this is a problem with many moving parts and infinite number of permutations. Each visitor to your site can be classified through use of hundreds of attributes. Buying journeys and consideration paths are different for different audiences. And, finally the business owners can act and treat visitors with a myriad of promotions, product recommendations, messages, content or layout changes.

To illustrate the size of the customer experience problem let’s compare it with a simple chess board.

Here is how. Let’s imagine that our problem is only to create the best customer experience on eight pages of your site. Let’s further assume that each page has only 8 elements, and that we have only one variation for each element. Visually this problem can be represented as 8×8 chess board where each column is a web apge with 8 elements and where each field can only have a black or white color (white is for old and black is for a new version of an element).

The total number of permutations of how the experience chess board can look like is equal to 264.

To help you visualize the size of this number let’s assume that each number is a kernel of wheat.

This number represents a huge pile of wheat. As a matter of fact this number is so big that it represents more grains of wheat than was ever farmed on our planet since the beginning of time.

This is precisely why we should use AI in eCommerce.

A problem of this magnitued can’t be solved through brute force. Conventional A/B split testing or if-this-do-that visitor personalization may sound nice but they are grossly inadequate solutions. To remain competitive you must arm yourself with much more powerful tools that are built on the foundation of big data analytics and machine learning algorithms.

Trust, but Verify

Product recommendation solution providers have done the most to apply advanced machine learning techniques and to deliver consistent results. But not all product recommendations are equal.

If you speak to the best of breed solution providers who invested tens of millions of dollars into research and development you will hear loud complaints about newcommers who are using open source packages to mimic their solutions. Unfortunately, many eCommerce brands are choosing a vendor on the basis of feature completeness of its offering rather than on the basis of the quality of the individual solution components.

If you listen to a vendor talk you will logically think that irrespective of the type of product recommendation solution you will get good value. The only difference between two products is how much revenue lift you will materialize. When a vendor tells you that ‘visitors who engage with recommended products are 10%-15% more likely to buy’ you are inclined to believe such a statement.

There is a small problem. The statement above could be true but at the same time you might be losing money with your product recommendation solution.

Here’s how. If you read the statement above more carefully you will realize that it does not compare results of those visitors who have seen product recommendations against those who were not presented with product recommendations. This would be the most accurate measurement. Instead, each visitor is presented with product recommendations which prevents you from knowing if the same visitors who enaged with recommended products would buy if there were no product recommendations at all.

That’s why we recommend you always test and verify the impact of each of the add-on solutions to your eCommerce site.

Start Small

In developing your business strategy do not view AI as a ‘silver bullet’ that will magically make your eCommerce business better. Instead, think of it as a new set of technologies that are dramatically changing the competitive landscape.

Over years the ability of your company to adapt and effectively use AI solutions will correlate with your ability to effectively compete and win in the marketplace.

The logical question is where to start and how to build a long term strategy.

We always like to be practical and recommend actions that are satisfying 20/80 rule: 20% of effort (cost) that provides 80% of benefit.

The list of such ideas includes:

• Customer experience assessment: in addition to a quite common UX audit we also recommend use of visitor data to perform an experience health check (FYI: if you do not have resources or do not know how to do it we offer a free health check service)
• Validate ROI of add-on applications: we see on a daily basis how brands are spending a lot of money on expasive user generated content applications, reviews, product recommendations – validate and make sure you are getting ROI from each one of them (FYI: we have advanced testing capabilites and we can assist you with this task).
• Uncover persuadable audiences – a fatal flaw of conventional personalziation solutions is targeting of potential customers vs. persuadable audiences, those who will become a customer only if you do something for them.

As a final note…

Consider this perspective from innovator Adam Morris, CEO of Redstage Worldwide:

“For better or worse, humanity is witnessing ever-increasing evidence that AI provides profoundly better solutions to problems we as humans lack the cognitive processes to conceive. Over the past decade, what we consider ‘modern’ or ‘state-of-the-art’ UX for online shopping hasn’t changed all too drastically. Of course, we’ve experienced improvements in search, filtering, and personalization, but an eCommerce sitemap still looks remarkably similar to the sites of old. Since AIs lack any preconceptions regarding how things ‘should’ be done, I believe AI will work to our advantage, ushering in a fresh age of progress that we ourselves could not possibly create alone in the same scope of time.”