#{LINKS}!
#!LOGIN!#
Today's customers expect personalization. To keep up with the pace of customer expectations marketers have to call in help from machine learning experts.
The upside of using machine learning for personalization can be huge. Think about it: delivering relevant one-to-one experiences, using dynamic content, and optimizing customer experience continually. Getting done more, with less: machine learning can optimize your marketing budget as it optimizes user engagement, loyalty and retention.
Spotify your experience
The normal (old-school) way of personalization in marketing is rule-based, where customers are grouped into segments by predefined criteria. Today we are a step ahead: we now have the know-how to create highly personalized segments of one. Great. But as personalization increases, so does the amount of data. Here’s where machine learning comes in. As M.L. is programmed to learn everything it can about the customer, it doesn’t use rules. Based on our clicks and preferences, it selects the most attractive user experience. The customer’s response to this experience is then used for the next and so on… Improving the customer’s experience all time, any time.
Think about Spotify, who now offers their listeners personalized playlists based on their unique taste. Spotify found that this way, users were much more likely to listen longer. What’s more, the number of listeners seeking out a track on their own after discovering it through their personalized editorial playlist, went up by 80%. For the techies among us: learn more about how Spotify & Google are making music special for everyone by watching the video below:
Pick and choose
Women getting ads about men’s shoes. Someone being targeted for a book that he already bought. Your five-your old son using your Spotify account. There are many examples where personalization has gone wrong. Machine learning can do a lot on its own, but human input is still needed. Marketeers have an important role in choosing the right data to be used by A.I. to maximise relevance.
There are still some choices to make for the marketeer. First of all, you have to pick the right algorithm. Choosing a ‘trending’ algorithm could for example show the most popular item on the website at that moment when a customer lands on the thank you page. Furthermore, filters, boosters and variations need to be chosen, as they set preferences for the choice of ads. For the thank you page this would mean that some items are preferred (boosters), kept from showing because the customer just bought it (filters) or shown only a limited amount of times (variations).
Did it work?
Equally important to using the right tech and choosing the right data is the post mortem analysis. Whether a customer bites for a personalized offer or not, you need to know what made the person decide. The results of the personalization need to be analysed in order to make better choices in the future and adjust the machine learning model accordingly. This analysis will boost the success of machine learning in personalization and can in the end positively influence your sales.