You’re entering one of the online bookstores. An offer of recommended products flashes on your screen right away: recent Swedish crime stories, sports gadgets and a cycling routes guide. How come one of the largest Polish shops knows that Scandinavian whodunnits is my favourite pastime and my holiday plans involve travelling by bike?

"This is personalized recommendation" Mateusz Gordon, Gemius expert in e-commerce explains. "The system is a well-established element of business strategies employed by the largest e-shops in Poland. If it weren’t for it, e-buyers would be presented with the shop’s entire product offer, which does not necessarily meet their needs or interests" he goes on.

For example, such solution is utilized by Spotify, a digital music site, which selects music pieces, performers, playlists or albums based on our preferences and history of portal usage. With this approach employed, discovering new music is easy and fun.

"Customers appreciate offers suited to their taste" comments Prof. Grzegorz Mazurek, from Kozminski University, head of e-commerce post-graduate faculty. "Research suggests that shopping recommendations are widely popular among e-stores customers. Over half of internet users considers them useful, practical and helpful" he adds.

According to the Professor, customizing website content to buyer’s needs is no longer just a marketing tool, but first and foremost, a great advantage for customers and a time-saver.

From the e-shop owner’s point of view, introduction of personalized recommendation systems is a sink or swim situation. In the times when, as IIBR research suggests, over 70 per cent of internet users declare that they did e-shopping at least once, and 19 per cent intend to buy products or services this way more often, it will be the fittest to survive, i.e. those that cater for the customer needs best.

"How long would an e-shop homepage have to be to feature all products on offer?" asks Gordon. In his opinion, this may not be a problem for shops with limited range of products, but in case of stores that broaden their spectrum and increase their customer base the story is different.

"No personalized offer on a website means a less satisfied and a less loyal customer. And often an irritated one, as he/she cannot get what they are looking for. A shop without a working recommendation system stays behind its competitors" Mateusz Gordon sums up.

How does a personalized recommendation system work?

• A recommendation system enables a shop to tailor an offer to concrete user needs on any location (i.e. the homepage, product, product category or cart or thank-you page). Taking the form of a recommendation frame, it displays those products that the visitor may find interesting. This kind of profiling is possible thanks to the fact that the system can ‘see’ particular types of customer behaviour: which price range they go for, which categories they click on, what they buy, but what they store in their carts, what their paths on the website are. Based on such information, the algorithm used in the recommendation system proposes the client such products that may be deemed supplementary to their shopping choices made so far. 

• Depending on whether the user is a regular client, or if they are registered or maybe they are first time visitors, the system will base on four types of recommendation: alternative, complementary, generic and personalized. The first  involve product of the same category or type, but of different brand. Complementary recommendation, in turn, puts forward items that correspond with the product the client has chosen and that increase the value of purchase, e.g. a remote to go with a TV set. Generic ones appear on a website when a user visits the store for the first time. It features bestseller proposals, products bought by other users or promoted by the seller.

The most complex recommendations are the personalized ones, suited to a customer taste based on his/her concrete behaviours on the website: posting comments, liking, adding to cart or simply browsing through the stock. These four types of recommendation may be combined.

What is more, generic recommendations are no longer generic upon logging in, as - based on a user visit pattern - the system will retrieve his/her interests and then, in form of a frame, it will propose whatever may potentially suit their needs. This is possible even when the customer deleted the cookie files but then re-logged. But if the cookies were not removed, the system will remember the preferences even if the last visit took place a month before.

An e-store not taking advantage of recommendation systems records a lower conversion rate (or the indicator showing the proportion of effected transactions to the number of website visits). Even if every additional product exposition may rise the rate, it is the increase generated thanks to presenting it in form of a recommendation frame that brings far greater benefits. Such shop also has a shorter user path (2-3 clicks) and lower CTR, as an implemented recommendation system will extend it and provide a larger number of clicks.