Personalization: a guessing game or a data-driven decision?

14 januari 2019

Are you stuck in Personalization? Here are 3 examples on how to move on.

Maybe you are one of the marketers who has worked hard to establish data driven marketing automation systems and can’t reach the point you were always aiming for. You wanted to make it possible to provide your customers with personal relevant messages, at the right time for engagement and through the right channel, as part of an integrated cross-channel experience.

You are not alone. Several of our clients experience the same, even after having worked hard and systematic for more than 10 years. That made us investigate what AI and Machine Learning could do to make it possible for our clients to achieve their goals.

To make a long story short, a new service called ONE Prediction was developed, which integrates into our clients’ existing marketing automation solution and utilizes the full potential of the data our clients already have available.

By using AI and Machine Learning, ONE Prediction automatically builds state-of-the-art prediction models for all kinds of customer responses, which make it possible for our clients to implement Extreme Personalization in their in- and outbound marketing activities.

If you are a marketer, it’s the real practical examples that you’re looking for, to understand how to get started, and not technical descriptions. Beneath is a brief presentation of three different real client cases solving personalization issues.

Case 1: Finding the right bank customers to call

Our client, a Danish bank, wanted to engage, cross- and upsell customers who have a potential, but whom they do not have a close relationship to. Throughout the years the bank has created a rule-based point system founded on experience, common sense and assumptions using demographic- and business data.

In a test, the rule-based point system was replaced with a ONE Prediction treatment impact model in order to predict the likelihood of a positive response if the customer is called by a bank advisor.

The model was built on the exact same demographic- and business data as the rule-based point model made by the bank employees.

When the bank used the ONE Prediction model output to sort and call customers, they got 3.5 times more customers per call than before. 63% of the customers called following the prediction booked a meeting or bought a product. In comparison, only 18% of the customers called following the rule-based point system booked a meeting or bought a product.

Implementing ONE Prediction instead of the rule-based point system not only increased the conversion rate but also makes it now possible to automate the selection of leads, while getting rid of maintenance of a complex set of rules, which saves a lot of working hours.

Case 2: Match the right concert to a subscriber

DR Concert Hall in Copenhagen has a very diverse program of more than 400 different concerts a year, spanning over sophisticated classic and jazz music, all kinds of rhythmic music, pop and entertainment.

Looking at each individual subscriber, very few of the 400 concerts is of interest to that person. In this context, communicating everything to each person or even concerts within a main genre could result in low engagement and risk of missing relevant concerts or unsubscribing.

ONE Prediction has a feature where the content (text and images) a person has responded to gets converted into a personal cognitive profile. This cognitive profile can be used to match the relevance of new content to each subscribers’ individual cognitive profile and thereby determinate a range of content relevance parameters which can be added to an offer relevance or content relevance prediction model.

ONE Prediction is integrated into the DR Concert Hall’s marketing automation system and is used to select relevant subscribers for each concert. These relevance predictions are used to decide who should be part of the target group, as well as which concert should they be exposed to in e-mails. In the near future, the relevance predictions will also be used to determine which concerts are shown on the front page of the website and advertised on Facebook and Instagram, to align communication cross-channel.

With the new models, the number of e-mails per concert is reduced to less than 10% of the subscribers, still reaching out to around 85% of the purchasers. Hereby, more than 90% of the subscribers are spared of irrelevant communication.

There will always be a dilemma between reducing spam and reaching all customers. The models can find the optimal level of number of subscribers to communicate to, based on available data and the models will automatically improve over time based on registered historical results.

Case 3: Identify customers who are ready for repurchase

Our client, a webshop selling electronic equipment, wished to be able to identify the right time to engage customers with an individual targeted approach. To be able to do this, it is necessary to predict when the customer is likely to proceed with the next purchase.

Using available data about customers and their purchase history, a ONE Prediction readiness to engage prediction model was created. The model calculates the likelihood of repurchase for each customer every day and our client can use the outcome of the model to select audiences for different marketing activities.

The model has proven to be very accurate in identifying customers with very high likelihood to repurchase and customers with very low likelihood to repurchase.

In the top 10% segment, the model accuracy came up to more than 90%, meaning that 9 out of 10 of these selected customers are going to repurchase.

Studies show that top 1% of customers are 18 times worth an average customer, so the ability to single out customers with a high likelihood of repurchase and concentrate a targeted marketing approach to this group can be very valuable.

I hope you enjoyed the reading and found it helpful.

ONE Prediction is meant to help marketers perform better, by generating sharper results, in a timely manner.

If you’re stuck in personalization, have a question or want more information, then feel free to contact me.

Kim Gregersen
Marketing Robot Developer


Mats Gustafsson

(+46)8 - 642 21 90


Maria Green

CRM ansvarig, Preem

”Certified CRM & Loyalty Manager är en heltäckande utbildning inom CRM med ett trevligt upplägg och grymma föreläsare. På kort tid kom jag till insikter som annars skulle tagit mig åratal”

Gabriella Elias


”Var på en mycket inspirerande kurs om kundresor (alltså hur man förbättrar kundupplevelsen genom kultur, arbetssätt, processer, mindset etc.). Kommer tillbaka till Skandia med förnyad energi och motiverad att fortsatt driva CX framåt!”

Jackie Kothbauer


"Att få utbilda och föreläsa hos Wednesday Relations är en kvalitetsmarkör som jag sätter stort värde på. Det bygger mitt personliga varumärke. Det ger mig värdefulla kontakter och en snabb återkoppling på nya koncept och idéer. Wednesday Relations har byggt en av Sveriges viktigaste arenor för kompetensutveckling. Det är ett nöje att få bidra till det."

Real Time Web Analytics