How artificial intelligence fits into e-commerce

Three elements of artificial intelligence—data mining, natural language processing and machine learning—can help online retailers improve results.

 

According to the recent Forrester report, Predictions 2017: Artificial Intelligence Will Drive The Insights Revolution, AI will grow 300% in 2017, and “will steal $1.2 trillion per annum from their less informed peers by 2020.” Numbers like these are behind the surge in retailers betting big on AI. Particularly for retailers looking to gain a leg up in e-commerce, AI is hard to ignore.

Still, while retailers have generally recognized the importance of AI, there is also plenty of confusion when it comes to relevant terminology and real-world applications because of the marketing noise around the technology.

 

So, what exactly is AI?

Market confusion on what AI is and what it’s capable of has in large part been driven by the simple misuse of the term AI.

AI is not a singular technology. It’s comprised of multiple components, such as machine learning. These component technologies that make up AI each have their own inherent value—but as with many things when it comes to marketing cutting-edge technology, the nuance is lost for the buzz-worthy.

This is one reason AI has become a catch-all word for multiple technologies. To move AI forward, we need to embrace the nuance of what it actually is.  Here are three important aspects of AI that e-commerce businesses need to understand:

  1. Data Mining – Often also referred to as Knowledge Discovery, this includes the technologies and methodologies for identifying useful and meaningful patterns in data. This is often based on extremely large data sets.
  2. Natural Language Processing (NLP) – This is the process used to assign meaning to human sentences. In the parlance of the industry, it provides machines with the ability to understand the languages that humans speak. Within the context of the e-commerce, it allows computers and software programs to understand the sentiment of a consumer’s written word—whether it’s email, online, social media, etc.
  3. Machine Learning – This is a specific process within AI, and refers to the science of self-learning algorithms. At its core, Machine Learning is about the use of statistics to solve problems using the data from the knowledge discovery process. The driving concept of machine learning is using technology to help humans think better.

Beyond these three components, AI also encompasses Neural Networks (computer systems modeled on the human brain and nervous system) and Robotics. However, neither of these fields are relevant for e-commerce at this time.

 

Depending on your goals, find the gaps in your data and fill them

 

AI Applications in E-commerce

By better understanding the technology underpinnings of AI, you can better view the world of possible applications and the specific impact it can have on retailers. A few examples include:

  • Dramatically Improved Search Capabilities – Although retailers have made significant gains over the years, today’s search algorithms still lack the ability to understand a given search query with the nuances of language, as a human would. It just takes a look at Siri to see we have a long way to go. Machine Learning combined with NLP capabilities, which fall under the AI umbrella, can improve search engines’ ability to learn from each new interaction, better understand what a customer is querying for, and deliver more relevant results—even if the wording isn’t exactly as programmed.
  • AI-Fueled Personalization and Predictive Recommendations Drawing on all three aspects—Data Mining, NLP and Machine Learning—retailers are now able to combine data gathered from transactions across all channels with actions taken by the consumer throughout the day, even those that aren’t necessarily related to shopping, to gain a deeper understanding of consumers’ wants and needs. For example, say a consumer tweets about it being cold outside. A brand may know from purchase history that she recently bought a winter hat, and that many hat buyers also like to purchase gloves. They could then respond to the consumer with some glove recommendations, or a discount offer. Beyond the “personalization” of the past, e-commerce companies can now understand what drove individual product purchases, and use that information to predict which products customers might be interested in, and even how those products could be customized to fit their personal preferences.
  • Improved Customer Interactions63% of consumers are highly annoyed by the way brands continue to rely on the old-fashioned strategy of blasting generic ad messages repeatedly (Marketo). Retailers can avoid this customer fatigue and drive e-commerce sales today by using Data Mining and Machine Learning to understand how each customer wants to be reached, and how often.

As an example, marketers already measure open rates on email, so they know if they’re getting customers’ attention. They also measure website clicks as a gauge of customer interest. If that click is on the “thank you” page on an ecommerce site, they know the customer has already purchased something, and likely even what that item is. Using this data, enriched with other data about the individual, purchase history and other factors through Data Mining and Machine Learning, allows marketers to communicate with consumers in real time with the right offer at the right place and time. They can understand when each customer will be most receptive to and offer, and what that offer should entail in order to capture their attention.

  • Concierge as a Service – Chatbots give marketers the ability to interact with the customer in real time and learn on-the-fly what the customer needs and deliver specific prescriptive guidance and results. Though the idea of bots has been around since the ‘50s and ‘60s when Alan Turing and Joseph Weizenbaum invented the first “chatterbot” program, Eliza, AI has given the technology new teeth—enabling conversational commerce.

As an example, a customer may be shopping for a particular brand of mascara on a website. Using Data Mining on past information about the customer and others like her, and Machine Learning to react to this new data about the customer, a bot agent could pop up and give her exclusive personalized offers that meet her needs and are relative to her color preferences, demographic, time of year, price triggers and sense of style. Given the amount of data to create a personalized interaction, a mere decision-tree logic would not work. Deep-learning customer models must be created in order to “react” to the consumer in real time, and NLP is needed to interact with her in a conversational way.

 

First Steps to Implementation

The list of applications for AI could go on, but the first steps for implementing the technology are similar regardless of how you choose to employ it.

First, get a comprehensive view of your existing data points. This could include CRM data, transactional data from online or mobile, demographic information, third-party sources—any and all context you can gather on your customers and their preferences can be relevant.

Second, determine your goals. Do you want to increase sales among your existing customers? Bring in new customers? Determine when you should have sales? Figure out which products you should be stocking? Having a specific goal will help you determine the best route to get there, and gauge the effectiveness of your efforts.

Next, depending on your goals, find the gaps in your data and fill them. The power of AI is that it can uncover correlations that aren’t readily apparent, and no company has access to 100% of the data they need on their own. For example, Starbucks may not know how heavy traffic is today or what the weather is in your current location, but there could be important correlations between those pieces of data and whether you are buying a coffee because you’re tired or because it’s cold outside.

Finally, it is important to understand that the Data Mining and Machine Learning process takes time. It is still a form of A/B testing, just done much more rapidly and on a massive scale. You can’t expect immediate results as the system needs to learn from successes and failures. For those that get a head start now, however, there is an incredible opportunity to grab a larger slice of that $1.2 trillion pie in the years to come.

Grey Jean Technologies provides personalization technology to retailers.