There’s a lot going on in the world of retail technology when it comes to artificial intelligence (“AI”) enablement.
RSR’s research has shown that retailers have very high expectations of AI; in a recent benchmark study, 41% of retailers declared that AI “will transform every part of our business. AI is transforming technology solutions too, and Symphony RetailAI has gone all-in, even to the point of putting “AI” into the company’s name.
The company is using AI to modernize – even revolutionize – core retail processes. CEO Dr. Pallab Chatterjee spoke at the recent Xcelerate forum in Dallas about the company’s plans to revolutionize the category management process. The data scientists at Symphony RetailAI are infusing the process with insights derived from AI models to not only help planners choose the best products, prices, promotions, and product placements, but also to be able to model how changes in one product in a category can affect other areas of the store and even impact the store’s overall performance.
Retailers should expect announcements in the coming months about these new capabilities. But right now, let’s focus on how AI is being applied to the demand forecasting process. (See viewpoint on “Is AI and Machine Learning the future for demand forecasting”.)
Retailers are anxious to modernize their forecasting capabilities for several reasons, but it all boils down to two things: they recognize the need to more accurately match assortments to consumer demand, and they know their response to changes in demand has to be faster and more focused. The good news is first that the data to help retailers and their partners to be more accurate and more responsive is available, and secondly, the technology to glean insights from that new data is proven and in the marketplace right now.
The AI-enabled Forecasting Platform
At the Dallas conference, Patrick Buellet, Symphony RetailAI’s Chief Strategy Officer, shared an update on the company’s demand forecasting product. Buellet started out by identifying the “elephant in the room” problem with forecasting today: bad data. According to the technologist, “bad data is 90% of the problem” with traditional forecasting. For the record, RSR’s own research shows the same thing; in our March 2019 study Mastering The Art Of Merchandising In The Technology Age, retailers identified “data is not clean: pricing, inventory, customer or POS” as the top organizational inhibitor standing in the way of integrated merchandising processes. Patrick stated that one outcome of dirty data and the resultant forecast is that 65% of orders end up requiring manual intervention.
According to Patrick, the difference between traditional forecasting vs. Symphony RetailAI’s AI-enabled forecasting platform is that the new solution is “self-cleansing, self-aggregating, self-learning, <uses> automated models, <and requires> no parameter settings.” When it comes to “dirty data”, the forecasting system uses AI analysis to clean that data and “complete” it.
When it comes to grouping data, the AI platform helps retailers cluster stores, identify seasonal families of products, and identify and maintain parameters and thresholds. Additionally, the system can help retailers handle difficult situations such as slow movers, erratic sales, very short lifecycle products (such as fresh foods), and new product introductions.
Perhaps most importantly, the new platform analyses the effect of external factors (weather, seasonality, holidays, promotions, competition, and geo-demographic data) on the forecast. What this all means is that unlike traditional forecasting, which uses historical data and rules (what Buellet called the “program”) to produce a forecast, Symphony RetailAI’s forecasting platform analyzes those external factors plus “cleansed” historical data to produce a “program” – the AI platform establishes the rules associated with aggregation and chooses the best forecasting algorithm. In other words, the program varies based on the context, to produce a better forecast.
To put a number to the value of AI-enablement, Symphony RetailAI performed an A/B test with a client. According to the technologist, the retailer already had processes in place to address data quality issues, and as a result was able to achieve an 80% forecast accuracy. Using the Symphony RetailAI forecasting engine, the retailer was able to increase the forecast accuracy for all items regardless of their velocity and category to 95%.
AI In Context
RSR’s 2019 research showed that over-performers (“Retail Winners”) are thinking opportunistically about how new data and AI analytics will help them to:
- Have a consistent, accurate and detailed demand forecasting platform
- Better incorporate customer segmentation & preferences into the planning process
- Tailor assortments to customer preferences
- Improve their ability to adjust to deviations from sales forecasts
- Shift to a holistic pricing, assortment and promotion decision-making process
- Integrate planning with cross-functional teams
As it is with any new and promising foundational technology, aggressively adopting new capabilities and getting viable solutions out into the marketplace creates competitive advantage. That is clearly Symphony RetailAI’s strategy with it comes to infusing solutions with AI-derived insights, and according to the retailers who attended the Dallas conference, it’s working.