deep learning explained

What is deep learning and what are the direct benefits of using deep learning in e-commerce?

Before we explore how deep learning can improve your e-commerce efforts, let’s take a moment to answer the question:

What is deep learning?

Deep learning is a branch of machine learning which has been developed to help us discover and trace user behaviour online at a more complex level than ordinary machine learning. Through it the machine is able to predict and anticipate user actions and trends and thereby make relevant suggestions before they are even asked for. There are many other uses for deep learning, but for the sake of this article we will be focusing on how it applies specifically to e-commerce and the daily lives of you and your customers.

Other names for deep learning include deep structured learning, hierarchical learning or deep machine learning. Deep learning is based on a set of algorithms that try to mimic high level abstractions in data. Basically, it gathers all of the available information and applies a set of variables. Each set of variables has a different outcome, a bit like a filtration system that separates items according to a specific set of criteria. In deep learning this goes through many levels and many different variables, allowing the algorithm to apply multiple processing layers which consist of both linear and non-linear transformations.

This is how we are able to have such advanced technologies as facial recognition and voice recognition. The fact that you can talk to Google (or Siri) on your Smart device is thanks to deep learning. It’s understandable that this kind of artificial intelligence (because that is essentially what is at a basic level) has huge implications in the world of marketing and sales.

How does deep learning work?

Although deep learning is a highly complicated and technical process, it can be summed up with a comparison to the human brain (it is learning, after all, which usually happened in the brain). According to Wikipedia:

Deep learning algorithms transform their inputs through more layers than shallow learning algorithms. At each layer, the signal is transformed by a processing unit, like an artificial neuron, whose parameters are ‘learned’ through training. A chain of transformations from input to output is a credit assignment path (CAP). CAPs describe potentially causal connections between input and output and may vary in length.”

What this basically means is that deep learning allows the machines to pick out what parts of visitor behavior will be useful for the specific goals set out for them by the programmer, even if these visitor actions seem unrelated with the end result in isolation.

How does deep learning affect e-commerce?

Huge amounts of data are available via mobile e-commerce. This data has allowed deep learning algorithms to trace the buyer journey and by doing that we (or our machines) now have a fairly clear picture of what kind of product information buyers search for when they are making purchase decisions for different things. Because the machines can hold and reference huge amounts of information at once while applying learned knowledge, they are able to predict what kind of purchases consumers are likely to make before they have made a decision.

What that means is that by following a user’s activity, your e-commerce site can offer the buyer purchase options that will appeal to them based on the specifications of other items they have viewed, how much time they spent viewing each item and what they did before and after viewing each item. It will even take into account time of day and your viewer’s location, whether they are male or female and any other relevant contexts available to the machine.

For example:

Let’s say Mary is visiting an online store and looking at Blue ladies running shoes. The shoes she is looking at are pretty but not exactly what she wants. She wants a shoe that has more arch support. Fortunately for Mary the site uses deep learning algorithms. Mary can click on the image of the shoe and she will be offered visually similar products to browse through.

The products recommended for her were chosen by analysing the image Mary clicked on, along with her other activity since she entered the site. The algorithm has suggested other pairs of shoes with similar colour and shape characteristics.

The system has worked with an input – output system whereby Mary’s actions and preferences have acted as the input, and the shoes that are recommended to her act as the output. The collection of pixels in the image that Mary clicked on serve as inputs, forcing the system to try and find similar sets of data. Simple, really.

It is clear how this is a powerful way of not only marketing your products, but of offering client service without even being present. The customer finds what they want easily, they have multiple options presented to them and all the information they may need is readily available. Basically, your e-commerce site becomes like a virtual store assistant.

It does mean that the visual integrity of your site is of paramount importance. Your images need to be very clear and easy to view with all of the relevant information loaded in the image map editor.

For you as a marketer deep learning means more opportunities to increase your conversion rates and improve your brand image through a positive customer experience. Not only are your customers being served targeted personalised content for the purchase they are making, they are also being offered related products that they are highly likely to be interested in, based on their own individual behaviour.