In this Q&A, Martin Smethurst (pictured below), MD EMEA, Trax, discusses with RetailTechNews the work Trax are doing with retailers, why the industry is adopting computer vision technology, and what retailers need to do to make the most of this tech.
RetailTechNews: Can you explain what Trax does and how the technology works?
Martin Smethurst: Trax is a world leader in computer vision solutions for retail. We work with major CPG (Computer Packaged Goods) companies such as Coca-Cola, AB InBev, and Heineken, as well as major retailers, to map out and analyse what is happening on physical shelves.
Our computer vision technology uses a combination of artificial intelligence, fine-grained image recognition, and machine learning engines to mimic the human visual system. It recognises images and then contextualises them to convert store images into actionable shelf insights. For instance, we’re able to recognise exact products on the shelf and provide insights into how that product is performing compared to its competitor on the shelf below, or even the larger-sized version of itself.
At the core of our technology are neural networks, a game-changing innovation in computer vision that allows us to excel in detecting objects within shelf images, and classifying these images with unparalleled accuracy. Our deep-learning architecture is trained on billions of shelf images, so it can now recognise fine-grained differences, such as a large and small version of the same product, and can overcome challenges like poor light, reflection, background clutter, and partial obstruction.
Since images are taken from multiple viewpoints, we’ve designed a stitching engine that mosaics shelf images and projects them on to a single reference frame. This visual reconstruction produces an optimal 3D structure that solves the problem of duplicate or removed products, ultimately digitally representing the actual shelf in the store.
What are the benefits to retailers of computer vision technology?
Quite simply, it allows retailers to digitise the physical store and compete against the rise of e-commerce. Computer vision allows retailers to make use of data insights, which have been so important in the rise of online shopping, to gain a full view of the ‘health’ of each physical shelf.
In retail today, it’s common practice for CPG companies to send products to retail shops every single day of the week, sometimes even several times a day, but they often have no idea what their shelf looks like. This is so important, because the way a shelf looks really dictates how much a company is going to sell. By utilising technological innovation, like computer vision, this is where retailers can give an organisation like Coca-Cola or Nestle a picture of what’s happening in stores across the world, so these companies can limit out of stocks or achieve the most profitable displays.
We’re working with clients to find out the perfect amount of shelf space for a brand to pay for. Some products are very ‘elastic’, so the more space you give to a particular product the more it will sell. Then there are other products that are ‘inelastic’. It doesn’t matter how much space you give them, they will still sell the exact same amount. In fact, if you reduce their amount of shelf space, they may still sell the same amount.
Using our technology, and using some really clever data science, we’re able to start to identify the products that are elastic and increase their share of that shelf; and at the same time we can also identify products that are inelastic and can reduce their shelf space so they can save money without reducing profits.
Are retailers currently making the most of computer vision technology? What are the challenges holding adoption back and how can these be overcome?
Some of the leading firms have embraced this technology and have seen immense benefits, but we want to see this become the standard in physical retail as brick-and-mortar stores fight back against the massive online retail players.
In terms of challenges, the industry has been stagnant for decades, with retailers and CPGs sticking to traditional methods, like manual audits, to see how products look on the shelves, how they compared to their peers, and whether out-of-stock items are being replenished. The key thing is to highlight to the companies how this process is not only time consuming and outdated, it’s also prone to error, something that computer vision can dramatically reduce – our technology boasts 96%+ accuracy.
Furthermore, it’s about highlighting the benefits of this technology, such as allowing CPGs to make actionable decisions. It won’t just tell you your share of the shelf isn’t as good as it should be, but offer insights on how to improve this. Using computer vision and deep learning methods, the solution can identify products and provide fully digitised shelf data. It can give companies real insight into the store so that they can see where, and how, they can improve.
Once retailers understand how their products are performing on the shelf, how can they optimise against this data?
Trax gives companies a fully digitised shelf and breakdown of performance so once optimised, the data can allow companies to identify the best product assortment for a store, understand the best location for a product and the best displays, as well as what pricing and promotion strategies are likely to succeed.
Once all of this these things are realised, retailers and CPGs can make the necessary changes and, ultimately, improve profitability. They can reduce hours spent on auditing, reduce errors, increase productivity of sales reps and, therefore, increase sales.
What’s next for Trax?
Having now worked with some of the biggest names in retail, Trax is now working on expanding the platform into additional geographies, covering more time during the day and adding additional frequencies and granularity of data, and moving into additional forms of trade.
This will solidify Trax’s position as the ‘Bloomberg of retail’, the ‘one-stop-shop’ for shelf data, and CPGs will have access to data on any store in any category.
Where we can take this one step further is to monitor and pre-empt trends very quickly, for example the flavours of drink that are dominating the market. This could help retailers and CPGs understand their target markets better and give retailers the opportunity to get ahead of these trends and market accordingly.