Suggestions to customers for products they may be interested in based on products they’ve already bought or viewed online, allowing your online business to increase conversions, conversion rates, average order values and decrease cart abandonment, while improving user experience. For example, if a customer bought the same tank top in 3 colors, there’s a good chance they’ll like the same tank in a new color. Personalized product recommendations, which mathematical calculations called algorithms determine on the back end, are a key feature of websites for cross-selling and upselling based on customers preferences and behavioral data.
When you recommend products you think your customers will love, you can improve their shopping experience and get more sales. But there’s a lot more that goes into those little boxes than you might think.
Personalized product recommendations are generated with user data taken from your website. Ideally, you’ll only show your customers products that they actually would like to buy. The more personalized the experience, the better.
When done right, personalized product recommendations can save you money and increase your customer retention. A 2018 report shows that returning customers who engaged with a suggested product (by clicking on it and reading the description, for example) were 55% more likely to make a purchase during that shopping session. For new customers, that figure was a cool 70%.
Recommendation engines, which are programs that analyze data on both products and users, drive features like this. Not only do retailers employ these engines, video streaming websites use them to generate lists of recommended clips for their users to watch.
There are 3 main types of recommendation engines:
This method analyzes data about multiple customers to predict what products a given customer will like the most. If a user viewed a particular DSLR camera, for example, it might show them lenses that other users bought along with that model.
Collaborative filtering systems also take other data into account. For instance, they might consider what other equipment the person had searched for, whether they were a first-time buyer, and where they lived. It’s distinct because it looks at the behavior of more than 1 customer, cross-referencing their purchase histories with each other.
This filter looks at a customer’s past choices to recommend products. Rather than pairing them with products that similar users liked, it uses their own actions to form a preference profile. This is the kind of engine that’s usually behind those “If you liked this, you might also like. . .” recommendations.
Content-based filtering systems are different from collaborative filtering systems because they only look at data from 1 customer at a time. Content-based filtering does not identify or analyze trends present in a group of similar customers.
As the name suggests, this method combines the previous 2 filtering systems, using data from similar users as well as that specific user’s past preferences to generate a list of recommended products.
These systems usually run collaborative and content-based predictions separately, then combine them. For example, a video streaming service might use this method to cross-reference data from users who watch similar programming to you with a list of the videos that you’ve watched in the past. The service then can come up with suggestions for you to watch next.
Of the 3 systems, this one looks at the largest pool of data. Consequently, it’s the most versatile and usually generates the most accurate results.
With so much user data at your fingertips, you can use a recommendation engine to offer very specific suggestions to your customers.
Recommendation engines usually look at data such as:
Your engine might use their geographic location to recommend a good coat when fall rolls around in their part of the world. In the case of our photographer, it might recommend a high-capacity memory card to go with their expensive camera after they’ve had it for a while.
To get the most out of your personalized product recommendations, do your best to:
For your personalized product recommendation engine to generate relevant results, it needs to focus on user behaviors that are, well, relevant. Not all of your customers buy based on the same criteria. Your engine should always be looking at what those criteria are and figuring out the “why” behind what people are buying.
If you sell T-shirts, some of your users will prefer particular colors or styles. Others will be loyal to their favorite brands. Others might even go out of their way to buy T-shirts that were made in the United States.
Does one of your customers prefer shirts sewn from eco-friendly fiber? That user probably cares about the environment and would be more interested in other “green” products than another shirt that’s a similar style. They might also be interested in things like non-toxic household cleaners and glass food-storage containers.
As with any marketing campaign, testing and benchmarking are paramount. If you don’t test out your strategies before using them in the real world, you’re missing a valuable opportunity.
Recommendation engines are powerful, but they still need direction from human beings like you. Always watch and make a note of what seems to be working. If you’re not sure, consider A/B testing.
Pay attention to whether certain kinds of recommendations perform better on some pages than others. A “trending products” box might test well on your homepage but not on the shopping cart page. If somebody is on the shopping cart page and is about to buy a new pair of shoes, maybe a “complete the look” box suggesting complementary products like handbags and belts would be effective. The only way to know is to test and make changes according to what works and what doesn’t.
You don’t have to only place recommendations on pages that already feature products. Why not place them on your 404 page and give shoppers an easy way to click back to your main site? A recommendation box on your checkout page could inform the customer that they only need to spend a little more to have their entire order shipped for free.
You can also put recommendations in your emails. An abandoned cart email might suggest products based on what the customer has waiting in their shopping cart. Many companies send follow-up emails after someone has bought a product, suggesting things that might go with it. If someone has looked at the same product several times, you can send them emails asking if they’re still interested and show similar products around the same price point.
How heavily you incorporate recommendations into your website is up to you. Some companies have personalized product recommendation boxes on every single page of their site.
Marketers and designers who follow this approach believe that maximizing the customer’s exposure to the recommendations makes them more likely to purchase the featured products. Detractors of the approach argue that featuring too many recommendations distracts from the purpose of the page and risks annoying users.
If you build personalized product recommendations in everywhere, be careful that you’re doing it right. Make extra sure your recommendation engine is performing correctly and displaying relevant results, or your customers might start to feel bombarded and tune them out.
Even if the recommendations are spot-on, some customers may be inclined to ignore them if you place them front and center on every page. It bears repeating: The purpose of a personalized product recommendation is to suggest something to your customers that they actually will find useful.
Some marketers advocate a tightly focused effort. You could choose a “quality over quantity” approach and only display a few curated recommendations on pages that your research suggests will be effective. Alternately, you can incorporate recommendations across your entire site, but do so subtly, placing them below the fold or in a sidebar.
When you check a product’s reviews before buying something or ask someone you trust for their recommendation, you’re looking for social proof. You want to be sure that the thing you’re about to buy is worth your money.
As a business owner, you can use your personalized product recommendations to provide social proof. Add little badges beside each product that show how many people viewed or purchased it that day. If a customer can tell that hundreds of other people viewed a product, it might nudge them a little closer towards buying it. Labels like “best seller,” “top pick,” or “staff choice” also lend credibility to your products. This process is collectively known as “badging.”
Some websites notify their users in real time when someone buys a product they’re considering. For example, “Jenny in California just bought a carved hematite ring.” This creates a sense of urgency and can motivate people to either make a purchase or investigate other products.
Personalized product recommendations make it easy for customers to find what they want to buy. The smoother you can make their experience, the more positively they’ll think of you.
When you use your sales data to tailor your recommendations and showcase relevant products, the more likely shoppers are to make a purchase—and come back again.