10 concerns about image recognition in retail and how to address them – Part 1 – Trax

10 concerns about image recognition in retail and how to address them – Part 1

Trax Retail
Trax Retail Trax Retail

10 concerns about image recognition in retail and how to address them – Part 1

Image recognition technology is disrupting the retail industry. But how do you tell one solution from another? And what are the essential attributes of the best image recognition solution in the market?

Here we’ll explain everything you need to know before you buy.

The CPG industry is flooded with mobile solutions aimed at improving the transactional piece of field sales.  Order entry and maintaining customer records is done largely electronically, and in real time. This has reduced operational costs and improved the supply chain process, but analysts agree that it’s just the tip of the iceberg. The ability to manage retail execution better and unlock truly impactful insights based on shelf truth is what’s next. Gartner notes that, “as the retail execution and monitoring market matures, the focus is shifting from transactional capabilities to those that can help a field sales force to sell more and do so more consistently”.

The coming revolution is spearheaded by image recognition and AI technologies that not only automate or digitize in-store data capture, but equip companies to convert data into sales. However, a recent Frost & Sullivan analysis reveals that a key challenge for users of image recognition technology is finding the proper solution that will address their specific consumer goods retail applications.

For consumer goods companies willing to make this step change, they need to carefully weigh in multiple purchase parameters before opting for an image recognition solution.

Buyer’s guide: Product attributes to look for

1. Match to needs: Is the product’s design and positioning directly influenced and inspired by your needs?

There are many applications of Computer Vision out there today in the market – people tracking, gesture recognition, driver assistance, medical imaging and facial recognition to name a few. However, to observe store conditions, the underlying deep learning architecture of the image recognition solution should be trained on vast volumes of retail product images. Trax, for instance, is designed for retail and consumer goods, and has the world’s largest retail SKU database, processing 40,000 images per hour. Further, the ideal product should be significantly more effective than manual store data collection, and have the flexibility to seamlessly integrate into existing processes, be it sales force automation or data analysis.

2. Reliability: Does the product meet your expectations for consistent performance during its entire life cycle?

You must be able to rely on the data generated by image recognition to make confident business decisions. Two key drivers here are accuracy and performance. Check for accuracy in image capture, recognition, calculation, and reporting of vital metrics like OSA, shelf share, pricing compliance and so on. Then, findings from the audit need to be actioned at the point of sale very quickly. For example, from the time new shelf photos are uploaded to the Trax cloud, real-time actionable results are delivered to sales reps, store employees, managers, and executives every 5 minutes, on average.

3. Quality: Does the product offer best-in-class quality, with a full complement of features and functionalities?

Choose solutions that can recognize and distinguish multiple products that are nearly identical in appearance. The right solution should overcome obscure and reflective packaging, identify empty spaces, products in poor visual conditions, and partially obstructed products. The overarching algorithms should also detect changes in the product life cycle like new design versions, or anomalies like incomplete shelf capture.

4. Positioning: Does the product serve your unique, unmet need in a way that’s better than other offerings in the market?

An enduring struggle in the CPG industry has been the ability to track how their brand and competitors are performing at the shelf. For image recognition to really solve this, granularity is key. The ideal solution should, at a bare minimum, identify dozens of unique, product SKUs, its size, type, shelf location and adjacent competitor products. The AI behind the Trax computer vision engine, for instance, comprises of numerous engines spanning across recognition, stitching, geometry and quality scoring – all designed to give you details down to the last SKU.

5. Design: Is the product built on design principles that supports ease of use and visual appeal?

The typical IT landscape of a consumer goods company is already complex. So, look for a solution that accepts retail photos from any digital source, and also integrates shelf KPIs into existing business intelligence (BI) tools. Another important design consideration is the provision of role-based dashboards and analytics. Shelf metrics can be used by teams beyond just the field forces; for example, as causal factors to help explain promotion results, which in turn can lead to better promotions in future business cycles. Gartner believes that this benefit alone will be big enough to justify image recognition investment for many highly promotional brands.

Read our next post to add 5 other critical considerations to your image recognition evaluation checklist.

Related posts

ARTICLE
by David Gottlieb

Now is the time to rethink retail merchandising

Retail merchandising is an essential tool for positive shopper experiences. To regain loyalty that may have eroded due to the pandemic, brands must ensure shoppers can easily find the products they are looking for and feel incentivized to purchase again.

Read more
Back to top