Consider this. You are shopping for new shirts to match your suite online. You take a quick picture of you wearing that jacket and trousers, upload it to the e-tailer’s website and immediately get matching suggestions of shirts and ties in your size.
Right now, that may sound like a sci-fi experience. But not much longer in fact. The rapid advances in AI, and more specifically the usage of neural networks can make this kind of CX a reality.
What are neural networks?
In the human brain, neural networks are electrical connections made when a person learns something new. It allows a child, for example, to ultimately know what a dog is – there are enough learning experiences that strengthen those neural connections so that, even though different in size, shape, color, etc. the concept of a dog is cemented.
Artificial neural networks are similar. Instead of neurons, there are units – input, hidden, and output. The input units are fed information; the hidden units process that information and decide what should be learned from the input; and then the output units report that learning. If the learning is right or wrong, the networks are notified through what is called backpropagation – a big word for feedback. Keeping with the dog analogy, it’s sort of like training a dog, providing the information, having it process that and then behaving in a certain way. Feedback is given based upon the dog’s performance.
Beyond artificial neural networks (ANN) are recurrent neural networks (RNN) which add memory cycles to ANN. RNN is often called deep-learning, and most applications are still in the laboratory state. For example, some data scientists have been able to produce writing that emulates Shakespearian plays or learn a language. While RNN networks are not practically useful yet, the current experimentation does demonstrate the amazing power that AI has and will have in the future to model human activity.
The implications for e-commerce
As neural networks continue to be developed, there are some interesting and potentially transforming uses for e-commerce. While online shopping continues to be an exploding human activity, there are still limitations with the experience – notably with online searches for products, absent a salesperson to help out. But neural network technology has the potential to provide a much more personalized experience that mimics more the help that a shopper would get from another live human.
Here are just three uses of AI that will optimize the online shopping experience.
1. Advanced searches
It’s pretty frustrating to conduct a generic search with what a shopper thinks is a good keyword or phrase. Often, though, it proves not to be and a lot of irrelevant results show up. The searcher must then go through iterations of keywords until the right one(s) are found. This happens because the algorithms in search engines match keywords with words in product titles and descriptions, not natural human language expressions. Neural networks that can learn natural language can put a “human” element into processing searches and allow shoppers to get results they want the first time.
Recently, three Belgian grad students developed a neural network that could learn the various features of dresses, both by text and by image. The network learned the differences of necklines, type of skirts, sleeve lengths, etc. So that when a search was conducted with specificity of attributes, the results were products that matched those specificities.
Siri is a personal assistant that retrieves information based upon natural language processing (NLP) capabilities. Users are able to ask questions in full sentences. But Siri is essentially a retrieval system that relies on basic web searches. Now, the same creators have introduced the “Son of Siri,” also known as Viv. Viv can handle more complex requests and uses neural networks that have learned enough to provide rather than just retrieve information.
One of the most important developments in AI has been the ability to gather, analyze and then use huge amounts of data to monitor consumer behaviors by a variety of demographic characteristics and then to put those consumers in groups based upon their shopping histories, preferences, etc. By doing this, marketers can target their advertising of specific products based upon those groups and even by individuals. The clearest example of this is Amazon. When shoppers search for products on Amazon, they are immediately shown other related products. And the selection of those other products is not random. They have neural networks in action that push products that the same demographics have also purchased.
3. Advanced sales forecasting and predictions
Traditionally, businesses have forecasted sales based upon their own histories. But consumers can be fickle; their preferences change. And marketers have attempted to make forecasts and predictions based upon what they think will be new trends and “moods” of their typical customers. This has always been a lengthy, time-consuming procedure, and not always accurate. But using neural networks that can gather and analyze huge data sets and learn to extract specific features and make predictions based upon those features. Two Chinese data scientists recently tested the accuracy of what they call a Convolutional Neural Network and the results validated the effectiveness of AI in sales forecasting.
What is it that e-commerce business owners want? They want increased transactions, greater customer satisfaction, increased customer retention, and, ultimately, more conversions. Neural networks that can boost e-commerce enterprises through better search results, personalized experiences and targeting, and a greater “human touch,” will provide superior efficiency and automation. Those businesses that embrace neural networking will gain a huge competitive advantage. E-commerce is a multi-trillion-dollar industry undergoing a big transformation – one that will permanently change the way we sell, and the way consumers will buy.