Building a data & analytics platform for retailers to drive value with AI

Retailers have big hopes for artificial intelligence. AI promises to personalize and enhance the shopping experience in new ways. Using AI, retailers can offer personalized product offerings or discounts to consumers, minimize waste by using complex algorithms to forecast demand, adjust inventory based on how consumer preferences are trending on social media, optimize logistics, figure out pricing markdowns, and more. The advantages are endless. 

But identifying impactful use cases isn’t the hardest part about AI. Where many companies run into challenges is in building the data and analytics platform needed to support AI. Here are some of the most common challenges that I have seen in my interactions with retailers around the world.

Focus on value

The foundation of any AI project is data. Companies start their AI journey by identifying the data needed to develop the algorithms. Let’s say a retailer wants to improve pricing markdowns using data analysis. To do so, the company needs to assemble and organize the relevant data sets such as real-world data on current and historical price and stock levels, promotions and other campaigns and their impact on product sell-through rates.  

Organizing the data and building the platforms to support AI is a big project and companies sometimes lose sight of the business objectives. We occasionally see retail companies buying large data platforms without understanding what they want to accomplish with it or how they can use it to create a competitive advantage. Data in and of itself doesn’t yield any value. 

To avoid this, data scientists and business managers need to work closely with one another to ensure AI initiatives are aligned with concrete business outcomes and company strategy. 

Building ‘unicorn’ use cases

Companies that are just getting started in AI and data analysis should begin with a unicorn use case – that’s one with the highest value creation opportunity. Starting with a small opportunity doesn’t have the same kind of wow factor that can be critical in winning over management. As companies see success and gain confidence, they can expand into more AI opportunities. 

The highest value creation opportunity varies depending on the type of retailer. In grocery, it’s often merchandise and assortment planning. In fashion, it might be optimizing the markdown pricing structures. 

Don’t buy a solution that locks you in 

Some data platforms integrate nicely with other IT systems while others don’t. Retailers need to be cautious with their data and make sure it’s not locked into a single system unless that system can serve all their needs.

Most retailers have ERP and merchandise systems from vendors such as SAP, JDA, Oracle, Aptos, Cegid, or Microsoft. These providers have some offerings for analytics, but retailers that plan to do a great deal of customization may run into limitations. Similarly, retailers that are storing CRM data in Salesforce, using WeChat in China (for the scan and go mini program) or Alibaba’s Alicloud stack face limits if they want to freely use customer data to, for instance, create an optimized store assortment based on customer profiles.

While giant retailers are building their own data & analytics platforms, this isn’t practical or necessary for every retailer. But companies that are looking to incorporate AI and analytics into their business need to consider where and how their data is collected and stored.

Knowing what and when a customer buys something, whether or not they take advantage of promotional offers, and other sales data is vital to optimizing algorithms. But that data can’t be used if it’s locked into a proprietary CRM system. Companies should ensure that cross-organization data can be combined.

One solution that’s become quite popular is open cloud platforms like Google Cloud, Microsoft Azure, or Amazon Web Services or cloud solutions from Oracle or IBM. Companies using these platforms or solutions are well positioned to adopt open standards or add on functionalities.

Growing the platform

Once a company has chosen a use case, identified the data needed, and taken the initial steps to build the data and analytics platform, there’s the task of growing the platform. As retailers expand AI initiatives, they also need to learn how to fund waves of improvements.

To make the most of their initial investment, companies should plan several initial use cases that can share the same data and analytical capabilities. Retailers typically keep tech operating costs relatively low, and many are using cloud-based data & analytics platforms, which require very low or no capex to expand.

Companies do need to build expertise within the organization to extract data efficiently from their data sources. Companies need to ensure their tech teams have the capabilities to develop and operate these data & analytics platforms. Managing AI platforms requires a different skill set so companies will likely need to reskill, train, recruit and possibly add temporary external expertise to build up their tech teams.

As the data & analytics platform grows and the team matures, companies then need to turn their attention to the industrialization of the platform. That is, retailers need to find ways to automate management of code and operations, monitor and control data quality, and understand the further requirements of the analytics use-cases and cost development.

#pushingboundaries

AI promises to transform retail and other industries. Instead of continue to rely on experience and human observations in making key decisions, companies are combining their old inputs with data and algorithms. This shift requires pushing boundaries - a sustained investment of time and effort, as well as alignment between technology and business strategy.