More satisfied customers with artificial intelligence

No one likes irrelevant advertising, but customizing advertising and product recommendations for each individual customer can be tricky. To help POWER take advantage of their vast amount of data, Computas has developed a new recommendation engine with artificial intelligence that has provided customers with more precise product recommendations. By using machine learning, the recommendation engine has enabled POWER to cut significant costs, which has been important in a highly marginalized industry.

Expertise in machine learning

For POWER, it was important to explore the possibilities that lie in the data they have, in order to make the shopping experience better for their customers. And this is where Computas’ expertise in machine learning came in handy. The answer for POWER lay in a machine learning model that could automatically be trained to become more intelligent. While POWER previously used solutions where, among other things, the product managers were involved, they have now become fully automated and can rely entirely on the machines. This gives better relevance to the customer while freeing up the manual work of the employees.

Computas has been a good sparring partner for our highly competent developer team, and their expertise in machine learning has been particularly important

Head of Digital in Power, Kåre André Jevanord

Customers in focus

Together with POWER, Computas developed a machine learning model that captures the customer’s movement pattern and provides relevant product recommendations based on the products the customer is looking at. With the continuous collection of such data, the model can thus automatically be trained to become more intelligent, and this is what sets the recommendation engine apart from other similar systems. Previously, POWER’s product recommendations were largely driven by business needs, based on pure vision or inventory status. With the new system, the recommendations have become much more customer-centric. Now, the focus is on the user benefits and the customer will see much more tailor-made content and targeted product recommendations. Simply because the recommendations are data-driven and there are facts that determine.

– The new recommendation engine we have developed together with Computas has managed to match and at the same time outperform expensive enterprise solutions we have tested. The fact that we can cut expensive licensing costs, while also having better control over internal data, makes us more competitive. As the solution provides better recommendations, the satisfaction of our customers also increases, says Kåre Andrè Jevanord, Head of Digital at POWER.

Omnichannel strategy

The solution has given POWER better control over their data and enabled them to use the results in an omnichannel strategy. This means that they can also take advantage of the results in physical stores since the system can provide the same recommendations to customers across all channels. All information about a customer obtained online should also be available in stores.

The cooperation between POWER and Computas did not stop the development of the recommendation engine. Today, POWER has an ongoing collaboration with Computas to discuss other opportunities in machine learning that can solve everyday challenges. Among other things, efforts are being made to include external data sets such as weather forecasts and geodata so that the product recommendations can be further improved. For example, there are big differences in when they want to promote barbecue products and heaters.

Technology we used

  • BigQuery
  • Cloud Storage
  • Cloud Machine Learning Engine
  • App Engine
  • Cloud Composer

Tools and algorithms

  • Python
  • TensorFlow
  • Apache Airflow
  • Collaborative filtering

Would you like to know more?

Send us an email and we will contact you.

Rune Hagbartsen