How to increase customer satisfaction with a unique recommender?
AI choice assistant for e-commerce
Summary
Especially now during the after-Christmas sales, it is important for e-commerce customers to recognize what products are worth buying and why.
We developed a native e-commerce choice assistant inspired by the offline shopping experience. The choice assistant compares the product alternatives, evaluates them (from point of view of price-performance ratio), and tells you, what their advantages and disadvantages are.
The AI model behind the choice assistant produces
- the price-performance ratio for all products
- advantages and disadvantages of individual products
You can have a look at the data coming from the model trained on the mobile phones category from one of our customers in order to get a better understanding of our solution.
The final outcome used by one of our clients can be seen in the screenshot below — the so-called “Advantageous alternatives” to the currently viewed product with a comparison of advantages.
Motivation
How to bring offline shopping experience into the digital world using AI?
Imagine how your purchase in a good specialized physical store works. Imagine how you want to buy e.g. child car seat, TV or car. What is it that you appreciate the most about offline shopping? We would guess that it is:
- You can see what the product looks like
- You can touch the product and maybe even try its functionalities
- You can ask the store experts about the comparison to other similar products, the benefits and disadvantages and their opinion about the best fit for your needs
We believe this shopping experience is crucial for making the customers satisfied and for building a long-term relationship between stores and customers. When customers purchase without understanding the product or its alternatives, this creates even more problems later, at home, when they start using the purchased product. Some customers are disappointed and some even decide to return it. This adds costs to your business and employees waste a vast amount of time and money. Moreover, unsatisfied customers may leave negative reviews and feedback and harm your credibility and search ranking.
The Solution
We are on a way to bring this offline experience into the digital world in a native way. And we started with the third point above (You can ask the store experts about the comparison to other similar products). This offline service is important for situations when you are looking for a product, which you don’t buy regularly and you are therefore not sure about the right choice (e.g. electronics, specialized sports equipment, car, etc.). You don’t want to be pushed or persuaded, you appreciate the help, the help from an expert for the given area. You just describe your general requirements and the expert shows you, what the possibilities are, explains to you the differences, recommends you the best choice, and explains why it is the best for you. This is the experience many shoppers miss in e-commerce.
We have managed to transfer this shopping assistant experience to the digital world in a digital-native way. We developed a machine learning solution that provides customers with similar but better products and compares them in a credible and transparent way. The engine automatically evaluates the advantages and disadvantages of individual products and can assess their price-performance ratio. This is done by the ML model calculating the performance of each product. The model is trained on the set of all products (even past ones that are no longer offered) — their parameters and prices. The engine is therefore suitable for products that can be described well by parameters; and it works also for products without history (new products, one-instance products). Therefore the algorithm does not suffer from a cold start problem.
If you are interested in more details, how it works, check the plot above (in the Summary or here). The performance calculated by the ML model is plotted on the y-axis while the prices are on the x-axis. We are focused on one selected product in the plot (Samsung Galaxy S10 lite). Based on the shown metrics we are able to assess, which products have a better price-performance ratio than the viewed one (Samsung Galaxy S10 lite). To make it more useful for the customers we look also to the similarity of the products, plotted by the color and the popularity of the individual alternatives, plotted by the size of point in the chart. Going over the individual points, you can see the top 3 advantages and top 3 disadvantages for each alternative (compared to the viewed product).
We call the engine Betterfy as it can reveal better choices. The practical usage of this engine is shown further.
How is the engine used in e-commerce?
Better choice recommendation. Betterfy advises more advantageous products with credible and easy to understand explanation. Highlighting the advantages of recommended products will help the customer to make the buying decision. As a result, customers get a shortlist of products that have similar (or better) parameters to the one they are looking at. And they get an explanation of why the recommendations are better. Just like with an expert in a physical store, the customer will understand the exact reasons why the recommended products are better than the previewed ones.
Above-standard product features. Betterfy automatically generates the features of products, which are specific for them among other products in the same category. The customers can understand quickly, how individual products differ, and what is special about them. For example, this car is special due to its low consumption within its price level. This can be placed either directly on the product detail page or in the products listing, which enables users to efficiently make the best choice for them.
Explanation, why customers should buy selected products. Many products are very good (have a good price-performance ratio) and there are no (or not many) better alternatives to be recommended by Betterfy (see the first feature above). In this case, Betterfy can work vice versa, i.e. saying why the viewed product is the best in a given category and price level. The explanation, why the products are good, can be done by the comparison to similar products and presenting features, which are better for the selected product. Or simply by just saying that the product is a good choice and why it is like that (like shown below).
Price-performance index. Betterfy produces a price-performance index for each product together with an explanation (why it is high or low). Customers can clearly see what products should get their attention. The ratio can be shown e.g. using a clock icon as shown below.
We developed this solution together with one of the key e-commerce players in Europe, the Mall Group, reaching great results in revenue. We believe these results are driven by customer satisfaction.
„The solution, which DataSentics developed with us, enables our customers to choose alternative products transparently based on their advantages and disadvantages. The solution is unique in the space of recommenders thanks to using machine learning to generate transparent and explainable recommendations to our customers. This led to an increase in revenue of up to 20% in certain categories while increasing customer satisfaction at the same time. DataSentics was a partner for us along the whole journey from the initial idea to a scalable solution in production and worked with us as one team.“
Roman Dušek, Head of Mall e-shop.
For more details about Betterfy and how you as an e-store can get it, see our web betterfy-ai.com.
Our Vision
We believe that currently, more than ever before, a customer-oriented approach is the key to success.
We are DataSentics inRetail, an AI studio focused on retail and e-commerce. Our vision is to bring the best of customer’s experience in the physical world to the digital one and vice versa and by using AI make this experience native for the given world. We are a part of the DataSentics family having 80+ machine learning and cloud data engineering professionals, based in Prague. We help our customers to generate profits and savings with the power of data analytics, machine learning & cloud technologies. We offer our tailored products, services, and custom agile solutions with consulting.
