AB InBev trials drones and AI in store to improve inventory visibility

Through its innovation unit ZX Ventures, AB InBev is investing in the technology and trialing how it can work in a real life situation to improve inventory visibility.   

Small drones navigate the store autonomously, scanning shelves and codes to report on stockouts, planograms and more, giving ‘unprecedented levels of strategic visibility’. The drone system can collect hourly and daily data on out-of-stocks and real share-of-shelf, with 98% accuracy.

Pensa Systems has just raised $5m in additional funding to accelerate deployment of the system, bringing its total to $7.2m, with investors including ZX Ventures.

A 'one trillion dollar blind spot'

Out-of-stocks could be costing retailers nearly $1 trillion, according to research from IHL Group. Such out-of-stock situations include empty shelves or instances where consumers cannot find what they’re looking for.

Inventory visibility is an age-old problem for brands and retailers, which is only getting worse as e-commerce and physical store shopping combine. Industry studies report nearly one in 10 items as missing on the shelf or misplaced, leading to lost revenue, dissatisfied customers or excess inventory.

Aimed at tackling this ‘one trillion-dollar blind spot' around in-store inventory, Pensa’s system is designed to inform retailers and brands of what is currently on shelves – across all stores, everywhere, at any point in time.

Watch the drones in action here: The Pensa System​​

The technology uses computer vision to 'see,' patent-pending artificial intelligence to learn, as well as agile, autonomous drones as roving eyes to automate high value visual tasks for the first time.

Pensa says it is the first company to deliver a broadly scalable system that provides complete, real-time and actionable data and insights, enabling CPG brands and retailers to automatically and systematically track in-store inventory.

AB InBev's Montreal trial​

ZX Ventures - AB InBev’s incubator, operation and venture capital team - first invested with Pensa in its initial series funding round which completed in May 2018. The two players have now completed a successful pilot the new inventory visibility system in an IGA Extra Beck store in Montreal, Canada.

With less than 90 minutes of set-up time, Pensa’s drone-based system collected hourly and daily data on out-of-stocks and real share-of-shelf within the beer section.

Over a period of two weeks, Pensa’s system scanned dry shelves and coolers with multiple product types (such as cans, bottles, packs), capturing 15.9M SKU images during 200 flights (totaling 2 miles of travel), with its continuously learning accuracy already reaching 98% for out-of-stock detection.

“Retailers and brands throughout this trial (as well as other trials) reported positively on the system’s tangible benefits—such as improving stockouts, ensuring planogram compliance, protecting shelf integrity—as well as how smoothly it deployed and operated in a store environment,” ​Richard Schwartz, CEO, Pensa Systems, told BeverageDaily.

“Shoppers were for the most part oblivious to the drone when it was 10-15ft away, but when they noticed the drone at work they were overwhelmingly positive and wanted to learn more. ​

“Overall, based on our experience with our live store trials, the Pensa system can be up and running in a couple days with no operational requirements to integrate into retailer infrastructure or software interfaces. First data can be collected in as little as 90 minutes on a per store basis.​

“Also, this approach is highly accurate (98+%) for out-of-stock detection. We believe that is a significant leap over the capabilities of other systems on the market.  This is due partly to the autonomous vision capture via the drone, which in effect is as if there were up to 100 fixed cameras focused on every product on every shelf in every location, and partly due to the cognitive AI processing on the backend in the cloud.”​

Schwartz says the technology is ‘uniquely effective’ in understanding real product and shelf conditions – and can do so at a lower costs per location than any other manual or automated approaches.

“Pensa breaks through the complexity and overcomes the limitations of prior attempts by marrying a lightweight, low-cost drone with powerful, patent-pending AI in the cloud to process the input and learn just like a person would,"​ said Schwartz. ​

"We call this combination an 'autonomous perception system,' and it’s the core of how we are able to achieve breakthrough results.​

“Specifically, our system is able to see every individual product, recognizes missing items and spots misplacements and other planogram misalignments. It can also learn, evolve and get more accurate as it goes.  ​

“As a result, we provide complete, real-time and actionable data on in-store inventory and conditions, including out-of-stocks (current and predicted), misplacements and planogram misalignments and real share-of-shelf.​

Wider potential

Furthermore, once the technology is set up in a store, operators can start training the system to highlight other issues - end caps, shelf performance of a new product placed relative to others, shelf performance of a major brand against a competitor for stockouts, share of shelf and other shelf conditions impacting revenue.

"In other words, we gives retailers and brands a common, zero-lag view of what is actually happening on shelves – across all stores, any product category, at any point in time,"​ said Schwartz. ​

"Because of that, we think our system is uniquely qualified to eliminate the $1trillion retail inventory visibility blind spot, so brands like AB InBev and retailers can minimize stockouts, optimize product planning and increase revenue.”​