US retailer Kohl's develops prototype to predict shopping trolley abandonment

American department store chain Kohl's is testing "event-based architecture" which can track online or in-store movements so it can intervene when customers are about to abandon a purchase.

Using information gathered from customer devices, websites and in store, Kohl's can develop algorithms to predict when a customer is about to change their mind about a purchase and send relevant offers to tug at their purse strings.

Kohl's, which has 1160 stores in 49 states, recently hired Starbucks' CCO, Michelle Gass, to boost revenue. This year, the c-level told the company that it wanted to increase its US$19 billion revenue to $21 billion within three years, Gautam Kotwal, Johl's big data engineering director told Apigee's 'I love APIs' conference in San Francisco this month.

The business now realises that to do this it must understand its customers better, he said.

This push will be underpinned by Kohl's digital and data strategy - which means extra funding for new technologies and supporting infrastructure and IT skills.

"We are investing significantly in APIs and store infrastructure which means all stores are Wi-Fi enabled; we are increasing data pipes in our stores," he said.

The firm has built its first version of a "data-lake" with a Hadoop-based platform which will ingest data and send summarised, usable information into Kohl's existing data warehouse. Kohl's is currently using this to drive customer recommendations on its website.

Kohl's has developed 78 algorithms to correlate its customer data, Kotwal said.

"But we want to do more to leverage this data-lake to understand customer journeys and get insight that will inform our strategies," he said.

"We have just started our journey to predictive offers and recommendations."

Related: Shoppers more willing to share data for a better omni-channel retail experience
Why customer engagement isn't just about data or digital at Ozsale
How chat is allowing CostumeBox to tap into new customer data

Kohl's are using various off-the-shelf tools like Scoop and Apache Flume with Hadoop. It is also working with API management company, Apigee, to use its API platform.

Despite these advances toward predicting customer behaviour, the 52-year-old firm is up against its legacy IT. This includes clickstream (every click a customer has made on Kohl's channels), sales data, reviews and ratings data and in-store wifi data, to name a few, Kotwal said.

"Our pricing and inventory sits in our mainframe system so that is the first challenge. Mainframes - yes I have to deal with them," Kotwal said.

Kohl's is currently piloting a smart fitting room in some of its stores. It is also piloting real-time trending product recommendations based on in-store and online purchases, shares and "pins" on social media on its website.