Empik loyalty program

The first personalized loyalty program in Poland with 1 million customers within the first 100 days.

Empik loyalty program


Laurens Coster created and implemented Empik’s in-house loyalty program called Mój Empik and launched it in August 2016. We based it on three principles which we followed throughout the creation - the loyalty program must: be functional (no spam), be useful (easy to use) and offer great user experience (members are appreciated).









Data Analytics



1 mln downloads of the app within the first 100 days.

About the client

Empik Group is a leader in the Polish market for the distribution of cultural goods and a dynamically growing brand on the Ukrainian market. In more than 250 stores on Polish territory, Empik offers its customers a mix of cultural products from knowledge and entertainment categories like books, music, film, games and multimedia programs, press, products, toys, board games, gadgets with a unique design, and tickets for cultural events. A significant part of the Group is the empik.com online store, which offers over 1.5 million products - both from the categories traditionally associated with Empik and new ones (i.e. Electronics, Home and Garden, Kids or Health and Beauty).

In 2016, Empik wanted to quit the Payback program and introduce their own loyalty program.

Empik needed a complex marketing personalization strategy for loyalty program members, that they were missing at that time, along with a reliable data warehouse for customer, transactional and behavioral data. Four years later, there were over 5 million members of Mój Empik and a long list of automated marketing campaigns.


Loyalty program

The first personalized loyalty program in Poland
  • 1 million customers within the first 100 days of Mój Empik

Laurens Coster created and implemented Empik’s in-house loyalty program called Mój Empik and launched it in August 2016. We based it on three principles which we followed throughout the creation - the loyalty program must: be functional (no spam), be useful (easy to use) and offer great user experience (members are appreciated). From the very start, it was truly revolutionary on the Polish market as each member was treated individually by being offered promotions linked to their purchase history and interests. And customers definitely felt it as within the first 100 days there were already 1 million customers who signed up, and 2 million within 7 months from launch! Customers could sign up for the program in Empik stores, but also through the Empik website and mobile app as the app was fully compatible with the physical plastic loyalty card. It allowed loyalty program members to only use their mobiles for all purchases while still making use of all personalized offers. And as all their receipts were saved on their app, there was no need to carry the paper receipts around just to make a return. Members could also easily check their purchase history, scan barcodes or check whether the product they wanted to buy is available. Just two years after the launch, 85% of Empik mobile app users were also members of Mój Empik, and the average basket value of loyalty program members shopping offline was 27% higher than of those who were not signed up. Additionally, members had double the frequency of shopping than non-members. But for the loyalty program to work this well, there was a need for a data warehouse which enabled successful marketing automation.


Data Warehouse

Data warehouse
  • Old Payback IDs linked to new Mój Empik members
  • Data warehouse integrated with marketing automation platform

Along with the new loyalty program, Empik needed a data warehouse. Together, we analyzed their business needs to design a data warehouse that would essentially meet them all. We then integrated it with Selligent to automate our marketing efforts and increase the chance for success (and what a success it was!). In short, the data warehouse we built gathers three types of data necessary for marketing automation: personal, transactional and behavioral data. The second type includes preferred types of products, frequency of purchases, whether the purchase was made online or offline, customer’s favorite shop (based on frequency of purchases made) and the time of day a customer bought something. All this allowed us to not only focus on transactions, but truly tailor our communication with customers based on their patterns. Empik’s new data warehouse uses REST API to gather data about online transactions, customer data and updates regarding personal information about customers. It also provides the data warehouse with any withdrawals of consent needed for updating marketing communication. For offline transactions, however, it uses SFTP. We then take all offline and online purchase data and merge it into a single SalesFact as well as merge two accounts (if a customer had one account they signed up for online, and one offline with a different email) into one single row. As a result, we know more about customers and are able to customize our communication with them even more! And as we had no previous transactional data linked to customer personal information, we suggested expanding the loyalty program registration form for customers to share their old Payback ID. Of course it was not obligatory, but, if a customer chose to fill this part out, we got historical data on transactions for that particular customer joining Mój Empik. This enabled us to gain insight into patterns from before the customer joined the new loyalty program, which, in turn, increased the accuracy of our awesome recommendations.

Recommended system
  • ALS machine learning algorithm for recommendations

Successful communication with customers is key. So to make sure it’s the best it can possibly be, we decided to develop a recommendation engine using a machine learning algorithm which recommends relevant products to particular customers. We decided on collaborative filtering with the ALS algorithm (Alternating Least Square). It analyzes customer behavior, preferences and interests to then select products to recommend based on the data about the users it has. We developed it on AWS Elastic MapReduce (EMR) platform and it’s a Spark job in Scala that uses Hadoop to calculate the algorithm. And the process is fairly simple: data about new promotions and available products is uploaded to Amazon S3. Then, API Gateway runs AWS Lambda which launches a Spark job in a transient EMR cluster. The results that are generated are sent back to S3. Then, we use them for flawless recommendations for loyalty program members who signed up for newsletters.


Marketing Automation & Analytics

Marketing automation
Personalized marketing including SMS campaigns, newsletters and targeted promotions

We worked hard together with Empik to devise an effective strategy to increase loyalty and, in turn, increase basket size and frequency. We started from scratch by choosing the right database management system, configuring and implementing the best marketing automation tool – Selligent – and, finally, supervising the campaigns on a daily basis. We have built tens of triggered campaigns, sent endless newsletters and text messages to loyalty program members.

  1. SMS and Newsletters
  • Recommendations in newsletters based on purchase history

Together with Empik, we developed a strategy for newsletters and SMS campaigns to get the best out of both. All newsletters were sent every two weeks and contained two different types of information. The first one was standard newsletter information about upcoming events or promotions. The second one, however, was personalized to each member. So what did it look like? Every two weeks we received a list of products on sale and by analyzing loyalty program members’ purchases (including product categories, book categories, book authors they were interested in) through our recommender system, we matched ongoing promotions to customers to create a personalized offer inside the newsletter. What’s more, SMS campaigns that were built were based on customer segmentation that Laurens Coster prepared. And the combination of personalized SMS campaigns and newsletters with personalized recommendations worked wonders! The sales from these were a crucial part of Mój Empik’s success.

  1. Trigger marketing
  • Many trigger campaigns including one based on average perfume use - remind customers to buy a new bottle before they run out of their fragrance

To automate our marketing efforts even further, we meticulously planned and executed a large number of triggers with the goal to increase sales and brand loyalty.

One of them is called PRESS - if a customer had bought at least one issue of a particular magazine in the last 12 months, they would receive an email once the new issue is available.

‍Another one is SCENT ON and it’s for customers who bought a bottle of perfume using Mój Empik card. Based on the number of milliliters the bottle had, we calculated, based on average milliliters of perfume used per day per person, the perfect time to remind the customer to purchase a new bottle before they run out of their favorite fragrance.

A similar trigger is used for other products, e.g. diapers. We calculated the average use of diapers per day, and depending on the number of diapers in a pack, a customer received an email encouraging them to buy another pack before they’ve used all of them. But here the situation is a little bit more difficult. If two weeks ago a particular size had been selected by a customer, we cannot be sure if the same size would apply today. So the communication focused on the last size bought as well as a size up - we want to think ahead and recommend the right products from the start.

Data analytics
  • RFM, churn rate and CLV analyses, among others

Once we had all the data not only about the customer transactions but also their behavioral data including interaction with marketing communication, we were able to go deep into data and what it all means for Empik. Throughout the years, we have done numerous analyses including RFM (to group customers based on the recency, frequency and monetary total of their transactions), calculating CLV (Customer Lifetime Value = Customer Value * Average Customer Lifespan) as well as churn rate. We have used Welch t-test to test for statistical significance regarding differences in an average basket value for loyalty program members and non-members. We have also calculated the odds ratio to examine the impact of marketing communication on a customer making a purchase in an Empik store. All of our analyses have offered valuable insight for both Empik and Laurens Coster to alter our approach to get even better results and to monitor our success.