Using Machine Learning to make your e-commerce grow
You already know that more and more companies are using machine learning with different purposes. Amazon suggests products that are just right for you; Netflix will recommend you series and movies that you will probably like; Uber can always allocate a driver to you at the same time as it tries to balance the number of runs between different drivers. These seem to be trivial applications, but behind it there are techniques in the state of the art in machine learning.
With this in mind, maybe it is time for your e-commerce to start using machine learning too. In the next sections, I will present some ideas on how your business can take advantage of it.
Organizing your catalog
Suppose you have a marketplace, with several distributors publishing the same product, with different titles. How do you group all these products together?
Take, for example, the image below. When searching for “ps4 controller”, the search will return all items that represent the same product, but with different titles, such as “ps4 controller”, “dualshock” and “ dualshock playstation 4”.
A manual solution to this is inspecting each SKU in your store and edit them one by one, making all SKUs match the same product. Obviously, this is impractical for marketplaces since there are hundreds of thousands of products in the catalog, with thousands of new items being added each day.
We can also try to automate this process by grouping products by image. However, notice how in the example above no image is identical to another, even representing the same product. We would have to use a more intelligent approch. That’s where Machine Learning comes in.
You could collect all the images and product descriptions from your e-commerce and use them in a Machine Learning system. This system would extract the most relevant information from each item and group them by proximity, as the image below shows us.
A system like this not only would organize all your catalog, but also would pre-classify any new item being added to the e-commerce.
Discovering the best price for your products
If you have a marketplace, you may have noticed that many sellers end up publishing their products at prices that vary a lot in relation to the average value practiced in the market. Although you can not force a seller to determine his price, you could suggest a price that would guarantee a fast selling, without giving up a good profit margin.
Based on your own previous sales data, your system can learn by itself how to suggest a competitive price by comparing prices within the marketplace itself and on other online marketplaces. All of this works with the support of systems built with Machine Learning.
Suggesting the ideal products to your customers
You’ve certainly seen some of those online shopping suggestions based on the current product you're browsing. To implement this, you can use simple statistical inferences, such as suggesting the most bought product along with the current item. But with the use of Machine Learning, these predictions can be much more sophisticated and suggest things that basic or manual statistical inferences would not be able to, based on purchase histories of other buyers like you.
Demo
Here you can play around with a demo of a furniture image classifier, using only the processing power of your own browser!
You can use something like the demo above to catalog your entire inventory with only the images of the products, for example. You won't have to spend human time for simple and extensive tasks anymore.
Conclusion
Machine Learning and other AI techniques will continue to be extensivelly used in e-commerces. If you still doesn't have any of this in your store, it's time to starting integrating smart systems to it.
