The retail experience is certainly changing in the face of the global pandemic. A Rip Van Winkle who might have fallen asleep in January 2020 and woken up in September 2020 would find their retail experience to be a surreal experience with shoppers wearing masks, markings on the floor separating folks from one another by six feet, and plexiglass screens by registers in checkout aisles. 

The online shopping experience has changed in many ways as well, with some items that had previously been taken for granted such as toilet paper, inflatable pools, and other commodities now being scarce commodities. Online retail is changing in other profound ways as consumers change their buying patterns and behaviors, with the shift to work-from-home and school-at-home changing the way people live, work, and socialize. Retail establishments that had previously counted on big Fourth of July and Labor Day celebrations, back-to-school specials, large social gatherings, and practically the whole travel and hospitality industry have had to throw out their usual sales, marketing, and supply chain practices and rethink their fundamental business strategies. 

All this is making the focus on data and machine learning even more essential than ever. Previous process and program approaches have been challenged, resulting in organizations realizing the importance of data and data-driven decision-making. At the recent Data for AI 2020 conference, Khalifeh Al Jadda shared deep insights into how The Home Depot is tackling these existential retail issues and provided in-depth insights into the core of the company’s e-commerce systems. On a follow-up AI Today podcast, he shared insights into the changing data science organization and its increasingly strategic role in retail operations. In this article, he shares further insights into how major retailers like The Home Depot are approaching AI and data science.

What are some of the challenges retail operations face when it comes to AI adoption?

Khalifeh: There are many challenges facing AI adoption in retail. The most important one is building the data science organization with the right talents given the shortage in data science leadership in the job market. Also, the placement of data science is another challenge since retail companies are not technical companies and as such they tend to not have R&D organizations where they can place data science. Sometimes the data science group becomes part of an existing IT org and they try to manage the data science team with the same strategy they use to manage the other IT teams but that is not right. The other challenge they face in adopting AI is the mindset of the business leaders that don’t necessarily believe in automation and machine learning. Many people in retail companies will feel threatened by AI and thus they will push back in any initiative or opportunity which the data science teams may present.  Organizations need the ability to create a research and discovery culture which is essential for the success of any data science organization.

Recommended For You

How is Home Depot solving challenging E-commerce problems using the power of AI and Data Science?

Khalifeh: Home Depot has a mature data science organization with world-class data scientists that came from top schools and research labs. This organization leverages different aspects of data science to solve challenging problems like search relevancy, query understanding, personalized recommendations,  and other applications of data science. One of the areas where [Home Depot] leveraged data science is in automation of collection recommendation. The customer pain point was to find all the products that form a collection when they shop for bathroom renovation or kitchen renovation or patio furniture. The customer can find one product, such as a faucet, which they like and wants to complete the bathroom set with shower head, towel bar, towel ring, and other coordinated items which have the same style, color, color finishing, and brand. At this point the customer has to conduct a separate search for each of the other products to find them in our catalog which is a time consuming and frustrating experience. Our deep learning multi-modal algorithm was designed to automate the process of finding all the products in our catalog that form a collection and provide those as recommendations whenever the customer lands on the product page. This work was published in the ACM RecSys 2019 and we have many other use cases which you can read about in these published research papers.

What are some of the unique opportunities retail operations face when it comes to AI adoption?

Khalifeh: Retail operations have unique opportunities leveraging AI. some of these areas include: 

  • Better pricing based on customers behavior and real-time analysis.
  • More accurate demand forecasting.
  • Anomaly detection to protect customers and business.
  • Personalized search and recommendations.
  • Voice and Image based search.

How is The Home Depot leveraging data science to gain insights and feedback on products?

Khalifeh: Home Depot has built a state-of-the-art sentiment analysis system which automates the process of understanding customers complaints as well as the features that customers like about our products. This system helps our customers quickly understand what other customers liked or disliked about a product without a need to read thousands of reviews.

How is The Home Depot using AI to provide better recommendations for related products?

Khalifeh: Home Depot has invested in building personalized recommendation engines leveraging cutting-edge techniques such as deep learning, active learning, and graph mining. Our AI-based engine uses different modals like text, images, click-stream, and profiles data to match our customers with the most relevant recommendations that match their intent and interest.

What are some examples of how The Home Depot has leveraged different aspects of AI to solve challenging e-commerce problems?

Khalifeh: We used Statistical Analysis and Association Rules to discover the relationship between different categories. We used NLP and NLU to understand customer reviews and extract the pros and cons of the products. 

We have seen lift in the engagement and conversion rates after deploying these advanced data-driven techniques especially when those techniques enable personalized experience.

What recommendations can you provide on how to form and manage data science teams?

Khalifeh: Data science needs to be treated as an R&D organization so managing the data science teams should not follow the agile process and the 2 weeks sprint framework. The data science teams need to take their time to conduct research and discovery in a culture that encourages innovation and creativity. Moreover, data science teams need to work closely with product managers and engineering teams so their placement in the organization is crucial to set them up for success. Some companies place data science teams under IT which helps make them better connected with their engineering partners, but they should not be managed as a typical software engineering team, instead they should be managed as an R&D team. Other companies place data science teams under business unit to be closer to the product management team which help in making them closer to their product partners but that usually creates problem in their relationship with their engineering partners so the success of the data science team in that case rely on the ability of the product managers to strengthen the relationship between data science and engineering teams. 

How do you see AI use evolving over the next few years in ecommerce?

Khalifeh: AI will keep transforming e-commerce and with more NLP, NLU and Computer Vision capabilities I expect the e-commerce to get more personalized. Also, Conversational AI will transform the way we shop by making shopping more interactive. For example, I can go to the Home Depot app and ask the app, “I need a new vanity”, and then the app then will ask me if I would like one with a dual sink or single sink. Upon my answer the app then will discuss with me different options and guide me through the purchase process in an interactive way.

What’s one AI technology that you’re most excited to see come mainstream in the coming years?

Khalifeh: I’m excited about AI applications in Healthcare and one exciting thing which I’m looking forward to is the AI for health care through wearable devices. Imagine a smart wearable device that can predict a heart attack before it happens and recommend actions/medications which you can take to prevent any adverse consequences? Imagine your smart wearable device predicting potential kidney failure before it happens and recommending to you the actions/medications to prevent that? AI in healthcare will enable all of that and more, and so I’m very passionate about it.