Recommendation engines - how traders can use them profitably

Recommendation engines - how traders can use them profitably
Published on: Mar 18, 2019
Updated: Aug 19, 2022

Zalando takes the place as fashion advisor and About You plays the best friend. Yes, we all know it: "Your style", "This item suits you..." and "We recommend these products...". But well, admittedly, they're almost always right. Because this shirt really does go well with the jacket you bought in the shop the other day. But how do online shops know exactly what we like? The answer is "recommendation engines". They run in the background. The software is supposed to determine as accurately as possible whether a visitor likes a product or not. To do this, the system filters products according to various criteria.

###Performance of the 100 largest German online shops

What are recommendation engines?

[tooltip title="Recommendation Engines" color="blue"]Are software systems that aim to make the most accurate prediction possible about a customer's interest in a product . They aim to suggest to the consumer exactly the products they are likely to choose[/tooltip].

Recommendation engines have become as much a part of e-commerce as cheese on a pizza. There are different types. On the one hand, there are the systems that work content-based. The characteristics of the product are compared with the browsing and buying behaviour of the customer. If the system finds matches, products are recommended to the buyer. For example, a customer has already ordered a CD by Bob Marley and, ideally, has given it a good rating. Other articles in the genre of reggae are now suggested to him.

On the other hand, there are systems that work collaboratively. These are a little more complex because they also involve other users. The system evaluates behaviour patterns of different user groups that have given similar ratings as active users. Products that are still unknown to them are then suggested to them. For example: Lisa and Lena are friends. Lisa gives a positive rating to the last hairdryer she ordered. The product is now also suggested to Lena.

In content-based systems, the properties of the product are relevant, in collabroative systems, the relationships between the users. However, technologies that combine both are particularly effective. This way, only really relevant and personalised recommendations can be made to customers.

These product recommendations can be displayed directly in the online shop or sent to customers via customer communication, e.g. in the form of individual shipping messages or personalised newsletters.

###Experience shipping messages in action!

What benefits do recommendation engines bring to online retailers?

The software systems have some advantages and can have a positive impact on the web shop:

  • Increase sales
  • Increase basket value
  • Keep customers loyal to the shop
  • Higher Coversion Rate
  • Increase repurchase rate

So what data is most worthwhile for retailers to capture and contextualise? If you base your product suggestions on location, gender, age and income, this is not very helpful. After all, an old man may be just as likely to buy the latest Nike sneakers as a young man. The solution: the use of recommendation engines. They compare the data of products that are often bought together with the customers' search and buying behaviour.

Of course, it is important that the system does not suggest a mobile phone to the consumer if he just bought one last week. For new customers, ads such as "top sellers", "products from the same category", "similar products" or "products already viewed" are particularly recommended. After all, the traders still have to get to know the visitors. Top lists, editorial recommendations or product lists, however, do not count as personalised product recommendations because they are not specifically made for the individual customer. Amazon, for example, operates customer loyalty with the help of targeted product recommendations.

Image Amazon suggests other similar items to products already searched for. (Source: Amazon)

What are the possibilities?

There are various forms of recommendations that merchants can suggest to their customers:

  • Top seller
  • Other customers also bought
  • Our experts recommend
  • You might also like
  • Because you bought XY, you might also like this
  • .
  • Matching products
  • Combine or complete products

Depending on the area, other recommendations may be appropriate. For technology products, recommended accessories perform best. For books, films and music, topseller recommendations are most effective, and in the fashion sector, the matching shirt and shoes to the trousers are most likely to be offered. It is particularly important to always keep your product recommendations up to date. This means that shops should only offer their customers products that are really available. Otherwise, this could lead to customers being annoyed and possibly abandoning their shopping trip. In addition, online retailers should not overdo it with the number of recommendations. Websites can quickly appear overloaded and this can lead to confusion among consumers.

Fittkau & Maaß Consulting surveyed 120,000 internet users on their attitude towards personal product recommendations. 15.4 per cent have a positive attitude towards these services. 24.3 per cent, on the other hand, reject recommendations. 60.2 per cent of the respondents said they were neutral on the subject.

Of course, there are places in the webshop where recommendations work particularly well. These are above all the start, category and product pages and the check-out. Product suggestions can also be placed in shipping messages. They also make the messages more personal.

  • Product pages: Product alternatives based on shopping cart content and purchase and browsing history increase conversion potential.
  • Check-out: Here it is advisable to suggest products that increase both the average order value and the number of items per order. For example, by offering products of the same brand.
  • Search page: Here, online shops should offer items that show similarity to the search term or the shopping basket.

Image Based on the item that is already in the shopping cart, other matching items are recommended to the customer. (Source: Zalando)


Personalised product offers should be everywhere in e-commerce in the future, as they are definitely one of the most important e-commerce trends in 2019. Many large shops already use them, but smaller retailers should not do without them either. After all, they bring several advantages. In addition to increasing the value of the shopping basket, you can, among other things, bind customers to the shop and achieve a higher conversion rate. However, if you decide to use recommendation engines, you should make sure that they are used correctly. Offering outdated or inappropriate products does not go down well with customers. They are particularly effective if they are placed on start pages, product pages or in the check-out. They are also extremely useful in shipping messages, as they can bring the customer back into the shop. But if retailers use the intelligent systems correctly, it can lead to a personalised customer journey and thus contribute to an improved customer experience.

Written by


Create new reasons for people to love your brand. Build standout post-sales experiences tailored to your customers. Deliver personalized touch points that grab attention and spark loyalty.

Read more from parcelLab
More from the category Research