If retailers want to get serious about managing disruptive trends, then using traditional approaches won't work. They've got to aim at experience and timely incentives - which means deep investments in efficient tech that enables both.
Don't believe me? Ask Jeff Bezos how it's working out for him. Then ask whoever is the current CEO at Sears what they might've been done differently in the past 5 to 10 years.
Recommendation Engines - The Old Way
Providing relevant, timely, and fast product recommendations is a must-have for retailers that want to increase customer loyalty and the average purchase amount. If you're not delivering it now, you are opening the door to competition that has already made the digital transformation.
It goes well beyond providing a basic categorization-based recommendation. It's not enough just suggest a product pairing that makes sense from a this-goes-with-that because two products are in the same product hierarchy or family. That is the type of approach is something Amazon was pioneering and abandoned years ago.
It even goes beyond retailers simply understanding how their customers browse or how like to pair products in their purchases. Truly intelligent recommendation engines provide an understanding that is customized to the retailer and thinking critically about which questions can help, such as when to suggest the pairing, which customer persona is right for a certain recommendation, and what the external trends can be mapped to product inventory - just to name a few key attributes.
“If we have data, let’s look at data. If all we have are opinions, let’s go with mine."
The legacy approaches of building recommendation based on data available through a warehouse - like in relational database systems - have not been successful because of lack of data. Many large retailers have had customer loyalty programs in place for decades. These same retailers have compiled years of transactional history. If your data warehouse is simply a copy of your transactions and customers, then it can only answer questions like "who purchased what in the past".
No, the old approach does not work because it is missing data - but because it is missing relationships between the data. In addition, the old way does not work because it imposes three limits that will always remain when based on an inherited model:
- the speed at which new information can be added to the model
- the speed at which ever-growing volumes of new information can be processed
- the speed at which the model can modified to incorporate new relationships
The last point is key. Using a data warehouse to do real-time recommendations is like continually throwing boxes in a room and expecting them to magically grow, form new shapes while they convince the room to meet new capacity.
Data teams need approaches and supporting technology that will allow them to remove these limits. Sticking with the old approach will always create a gap between the speed at which teams think they can adapt and how fast they can actually adapt. Consumer-focused companies that want to grow the average ticket size must first connect all of the relevant data at their disposal, start measuring relationship patterns and incorporating results in real-time.
Moving to a Graph-based Aproach
As George Manas, president of Omnicom’s Resolution Media performance marketing division, states in Market Watch "Amazon’s critical difference is that they own what I call the purchase graph”. Amazon goes beyond understanding purchasing trends and customer personas, which are important in retail intelligence, but only a few pieces of the puzzle. Incorporating graphs provide a 360-degree view of data that can be updated and analyzed in real time – and run up to 1000 times faster than legacy systems.
While graph databases are also newer technology, they are proven in the retail space, especially when wanting to deliver on-target and successful recommendations. McKinsey points out that "35 percent of what consumers purchase on Amazon and 75 percent of what they watch on Netflix come from product recommendations". Retailers can use graph systems to quickly join together complex, connected and varied data that provide vastly improved personalization for customers.
The flexibility and change-resilient nature of graphs allows for intelligent, adaptive understanding of information in order to see the most valuable patterns within your data. Using Graph-based search systems, retailers can go far beyond standard word or phrase matching to deliver precise, intelligent results based on connected content, customer interaction and rich metadata, such as location, time and events.
From recommendation engines to personalization, graph-based systems create a better experience for your customers and unlock cross-promotion and upsell opportunities not possible with legacy systems.