top of page
  • Irina Mikhaylina

How online stores can boost sales via product recommendations

You are looking for an armchair in an online store - and the algorithm recommends shower curtains? You're looking for some fancy high heels – and the recommendation is a pair of Birkenstocks?

Looking at the product suggestions in online stores often leaves us rubbing our eyes in astonishment. Classic recommendation technology fails. Visual Intelligence software such as vviinn could solve this problem - and generate more sales.

Our CEO Philipp Derksen wrote a guest article about this in the leading German trade title Horizont.

Most online stores are teeming with useless product suggestions. Yet meaningful product recommendations can be an effective sales lever. Visual intelligence could help to make better use of this potential. Philipp Derksen, who has developed Vviinn, an AI-based software solution for visual product search and product recommendations for online stores, explains how it works.

"Often people don't know what they want until you show it to them". This perhaps most famous quote of Apple founder Steve Jobs is still true. Especially in online retail, showing products is more important than ever before. Recommendations now account for more than a third of sales at e-commerce giants like Amazon. Netflix generates 75 percent of its reach based on recommendations. The sales concept "If you're interested in this product, these alternatives will probably help you, too" can be a real sales booster - if used correctly.

But here is the problem. Many online stores do not exploit product recommendations as a sales lever. Browsing the average suggestions in online stores often makes for a stunning experience. See some examples from German e-commerce reality:

If you put a cocktail chair in your shopping cart at online furniture retailer Poco, it is suggested that you also get: a bar stool, a cutlery insert and a shower curtain.

If you click on a green summer dress at the online fashion retailer Meinfischer, the category "Let yourself be inspired" suggests black trouser suits and white blazers.

If you search for red pumps on, you will see a selection of quite profane health sandals under the heading "You might also like this".

The series of such unfortunate recommendations, unintentionally funny as they are, could be continued endlessly. Retailers, though, are probably not laughing about it. In fact, it is a serious problem.

Most store operators are unable to cope with the issue of recommendations. Neither do they seem to have the capacities nor the possibilities to make meaningful product recommendations to their customers. Even if they had the massive amount of data required for that task – and they don’t - they could hardly muster the necessary capacities to analyze them meaningfully and process them in real-time. Big players like Amazon and the like can do it. The majority of European retailers can't.

As a result, shower curtains are recommended to customers interested in cocktail chairs. One may safely assume that it wasn’t a complex algorithm that calculated a large overlap between green armchair buyers and lovers of white shower curtains. If I'm wrong and there is some higher science behind the examples given, I'm happy to acknowledge defeat and welcome instructive feedback.

However, I think it is probably just the phenomenon that I call the bestseller bubble. Based on thin data, most online stores simply suggest products that many people have bought anyway.

With the imminent end of third-party cookies, the problem is getting worse. Two-thirds of all customers are already annoyed by cookie layers. Already, personalization in its current form is unable to reach a large part of the potential clientele. The concept of real-time recommendations based on historically collected user data is reaching its limit.

One way out of the recommendation trap is visual intelligence. Image-based recommendations do not require cookies at all. Product suggestions are usually much more precise and better reflect the image that customers really have in mind when they go shopping online.

Visual intelligence analyzes more than 1,000 features of a product image within milliseconds, suggesting stylistically suitable products as a result. Back to our green cocktail chair: visual intelligence would suggest a selection of similar, primarily green, seating furniture, which would very likely be much closer to the searcher's intention. In reality, the use of this technology has been proven to generate higher sales figures. Tests we conducted with customers showed the conversion rate of recommendations increasing by 258 percent.

It's worth suggesting the appropriate product to people at the very moment they intend to buy. Let's show them. And to all online retailers: let yourself be inspired. You might also like this.


Philipp Derksen, 43, is founder and product owner of Vviinn, an AI-based software solution for visual product search and product recommendations for online stores. The Berlin-based entrepreneur and graduate engineer has been working in online business for over 20 years and in e-commerce for 15 years. His company Mediaopt has been developing software solutions for the e-commerce industry since 2009 and advises companies on their digital sales strategy.

The link to the original article in German you may find here.


bottom of page