Footasylum steps up artificial intelligence to drive customer centricity

Footwear retailer Footasylum is looking at evolving its use of cloud-based artificial intelligence (AI) and machine learning (ML) tools as part of its omnichannel and internal efficiency plans.

The company started looking into the technologies in January 2018, and has since introduced an AI-driven marketing platform. The company then built a recommendation engine late last year and is now carrying out trials around AI and ML to enhance areas such as inventory management.

“We recognised that there is a sort of bifurcation happening in retail at the moment where stores are heading in one direction, and the web is pulling in another,” Tom Summerfield, head of e-commerce at Footasylum, told Computer Weekly at the AWS Summit in London.

“For example, we realised that we have customers on our email database that don’t shop on the website, so we needed to be serving them better and, ultimately, removing friction from their consumer journey.

“So we started a process of digital transformation, whereby we want to see how the momentum that was being created in the stores could help the web, super-powering existing trading strategies that we have.”

The software delivers a predicted customer view of people most likely to be engaged with and buy Footasylum’s products. Customer segment profiles are then used to target similar audiences via social media.

As well as helping the retailer understand online transactional data, the product links up with what’s happening at a customer level in stores and systems such as the firm’s loyalty scheme. This, Summerfield said, has enabled the creation of a single customer view across the business.

Footasylum is already reaping results from introducing the AI marketing engine supplied by Peak into its social media advertising campaigns.

According to the retailer, use of the system led to a 28% increase in email revenue and a 8,400% return on ad spend, 30 times more than industry average and 10 times higher than its standard marketing return on advertising spend.

“By using the data from consumers, we’re learning when they are in market with high propensity to purchase and then we tune up the message for those customers. Likewise, if they’re not in market, we will warm them up with a different type of message,” said Summerfield.

The company’s aim is to send less, but better, marketing to its customers, the executive said. That is all under the hypothesis that if the relevant products are displayed at the right time to the right consumers, average order value and lifetime value metrics related to the consumer base will tend to improve across all channels.

In addition to a recommendation engine for online products, also underpinned by a Peak product, the company is working with the supplier to see how it can extend the AI and ML capabilities to enhance stock management.

“We’re entertaining the idea of how these tools can positively affect our inventory, how stock moves around the business, as well as demand forecasting, where we can potentially use AI to positively impact demand in a certain store or regionally, or on the web,” Summerfield added.

“What the immediate future holds is automation alongside hyper personalisation. The web can then feed stores with information that is valuable to them as part of this digital-first, customer facing culture,” he added.

According to Summerfield, the fact that the Peak systems are hosted in the AWS cloud is helpful since it allows for scale.

Footasylum ingests data into Peak’s system, and the data is stored in AWS’s Simple Storage Service (S3) and processed using Apache Spark running on Amazon EMR. Peak uses a number of AWS services to power its AI offering, including Amazon Elastic Compute Cloud (Amazon EC2), Amazon Simple Queue Service (SQS), AWS Lambda, AWS Fargate and Amazon SageMaker.

“We want to have a one-on-one relationship with our customers and drive these projects around personalisation across our entire database – so, potentially millions of customers. This would require a huge [infrastructure] resource, whereas cloud-based AI services can scale across our entire ecosystem,” said Summerfield.

When it comes to challenges involved in the project so far, Summerfield singled out culture as the most intricate aspect of ramping up use of AI and other related technologies.

“Challenges were more culture-related rather than technical, related to new ways of doing things, which we were able to quickly overcome as we educated ourselves around how [the new systems] embed into our existing efforts,” he added.

AI adoption at industry-level faces challenges such as lack of understanding of what the technology can do for retailers, as well as difficulties around integration and access to resources needed to fully embrace it, said Summerfield.

“No doubt, there will be some CIOs and CTOs who would be confident of building their own AI capability. But to support initiatives like the ones we are working on, we would need to employ a squad of data scientists, which is really expensive and hard, so engaging a third party is very useful,” he added.

When it comes to increasing use of emergent technologies such as AI in retail, Summerfield pointed out leaders in the sector have to let go of some common misconceptions.

“AI isn’t just an e-commerce play. This is a bit of a fallacy among certain retailers, but this might be the key to breathing some life back into the high street – so I would say, don’t hide away from it,” he said.