start portlet menu bar

HCLSoftware: Fueling the Digital+ Economy

Display portlet menu
end portlet menu bar
Select Page


Graphical user interface Description automatically generated

Recommending that right product at the right time to all your customers can make a huge difference to your bottom line. They can help to increase cart value, customer lifetime value, and conversions. Sometimes people don’t buy something from you because they don’t know you have it, and other times, they may not know they want it. This means they will never search for that item on your site, and unless you present it to them, they will never buy it. So, how do you know what to put in front of people at the right time and place? The great answer is that you don’t have to.

With the rise of machine learning and data science, you can finally rely on artificial intelligence to drive your revenue as high as you possibly can. This isn’t to say that it’s all just magic happening behind the scenes, or that you and your merchandising teams don’t have to put any effort into it. Unfortunately, there are no “one size fits all” product recommendation solutions, or at least not the ones you need to truly maximize conversions. There are three key areas that you will need to focus your efforts on to be successful here:

  1. User Experience Design
  2. Data Science
  3. Analytics/Monitoring

Of course, mastering each of these areas will yield financial fruit for many, many use cases, I will illustrate in this article how you can specifically take advantage of them to nail product recommendations.

User Experience Design

It all starts with understanding how your customers think and shop. This is different across retail and B2B verticals, but there are also some universal learnings that apply across the board.

Here are a couple concrete examples of hyper specific vertical-specific thinking:

  • A cattle farmer shopping for a new set of tractor tires is also interested in 300 pounds of hay seed that is resistant to northern leaf blight and would love to know about a new John Deere hat.
  • An investment banker in Miami is looking for a custom band for his $12,000 Panerai Luminor divers watch before his next flight to Micronesia.

In these scenarios, you would likely implement two very different product recommendation user experience strategies. For example, you may want to label your recommendation areas vastly different. In the farmer example, a recommendation area titled “Other Blight Resistant Seed” would be appropriate on a seed page. You may also have several different product recommendation areas on a single product page, whereas on the watch site you may want to heavily limit the number of options available in order to minimize decision paralysis.

However, in both scenarios, you would benefit from a “Customers Also Bought” recommendations area on the order confirmation page.

How many recommendation areas, what you label them and where you place them across your site is driven by generic best practices and by domain-specific user experience approaches. The thing to keep in mind is that the “formula” for each of your recommendation areas is potentially very different which leads us to Data Science!

Data Science

To me, this is the most exciting area of product recommendations. It is also the hardest to master and the easiest to underestimate. You should think about data science as a way of better understanding the context, affinities and needs of your customers so that you can better personalize their experience. In the cattle farmer example, you may not know what kind of farmer they are when they first show up on your site. After they search for 10 sets of $5,000 tractor tires, you may get a sense of their scale. When they search for seed used to grow cattle food, you can start to infer a lot more. The more and more they interact with your brand across channels, they give off tremendous amounts of data over time. Assuming you are capturing all that data (please say you are!), you can feed that to your machine learning models and train them to deliver better and better recommendations.

Again, this is not magic, and there are no turn-key solutions that will have you producing 1:1 personalized recommendations on day one. You must implement a full data science lifecycle in your organization that leverages your business domain knowledge and drives a data acquisition strategy across all your customer touchpoints. Then you will need to employ skilled data scientists to leverage that data to build “models”. Some companies have hundreds and thousands of different machine learning models that they use across their enterprise to infuse intelligence into a vast array of business functions. Whether it’s financial forecasting models, recommendation models, image classification or fraud detection, there’s a model to be had for it. Managing all these models and deploying them effectively across your enterprise is essential to getting the most out of data science but start small. Just implement a few recommendation models on your site and get all the plumbing right for those, and then expand out.

Analytics & Monitoring

What good is all this work, if you don’t know if it’s making a tangible impact on your bottom line? Implementation of a data science lifecycle isn’t cheap, and you will want to know, not only if it’s returning on investment, but also is it creating any kind of bad experiences for customers. Is it returning completely ridiculous results? Is it returning blank results? Keeping an eye on things after you deploy your models is key to success.

Here are some of the things you should be measuring for product recommendations specifically:

  • Bounce rate changes on pages that have recommendation areas
  • Total impressions of recommended product areas
  • Clickthrough rate of recommended products by area
  • Purchases specifically tied to click throughs on recommended products
  • “Assisted” purchases – purchases that occur on non-recommended products after clicking on a recommended product
  • Top-performing product recommendation areas
  • Audience/Segment performance

You will also want to keep an eye on qualitative results. For example, make sure your Commerce system allows you to shop on behalf of a customer segment so that you can see the kinds of products being presented. Perform session replays if you have that capability and monitor changes in customer feedback through pop-up forms and post order surveys.

Click here to learn more about how HCL Commerce helps companies ‘sell more’.


Comment wrap
Marketing & Commerce | December 22, 2023
Is Customer Behavior Analytics on Your Radar? With Increasing Digital Sales, It Should Be
Last year, behavioral analytics has been one of the hottest topics with Commerce customers. It will enable reporting and analytics capabilities.