Date published

Is your in-store data giving you the complete picture?

By CK Tan, Associate Director, Product Marketing at Trax

Your shelf-based strategy is only as effective as the completeness with which Computer Vision solutions collect in-store data.

We discussed the importance of accuracy in shelf data and how a rigorous data quality enforcement helps Trax provide high image recognition accuracy at the retail shelf. But in an ever-changing retail landscape flooded with new products and package redesigns, accuracy is only part of the story. If you’re using Computer Vision (CV)-powered solutions for retail execution or shelf measurement, the completeness of your data is every bit as important to staying up to date and on top of your competition.

Why Completeness Matters

Coca-Cola ran a campaign during the recently concluded FIFA World Cup, releasing limited edition cans with numbers 0 through 9 printed on them, for fans to share score predictions on social media.  A global event like this is every brand manager’s dream, but it can also turn out to be their worst nightmare. Without a system to track whether the new products made it to the shelves in the most important stores, it is hard for brands to drive optimal ROI on these expensive campaigns.

Even with Computer Vision powered store data collection, the challenge of recognizing new items on shelves is a hard one. Besides new products, incomplete data can also result from a host of other anomalies, like poor image quality, fraudulent store visit data entry, duplicate images and so on. Incomplete data can lead to inaccurate results during the recognition and digitization process, and in turn affect the reporting of shelf metrics.

Ensuring Completeness in an Ever-changing Shelf 

  1. Automated detection of new or redesigned products

New product launches are the lifeblood of many a team in a CPG company and are an important way for a brand to remain relevant. At the same time, new designs of existing products get introduced. For solution providers, an outdated SKU database is the main cause of degeneration in the performance of the recognition engine, leading to inconsistent data and incomplete insight.

To prevent this, Trax has developed an ongoing monitoring service based on a well-designed algorithm and running on an active learning engine. So, every time a new product or design is detected by the system, the identity of this product is inferred using reference data, validated and then updated on the database.

The result? Users can keep track of new packaging designs and new products and stay on top of new launches from competitors in the market.

 

  1. Automatic anomaly detection   

Image recognition has simplified in-store data collection, but to be truly effective, retail execution solutions must be immune to anomalies related to image capturing. Field users may capture the same section of the shelf multiple times, or sometimes fraudulently upload the wrong images as visual proof of activation. To avoid such anomalies and ensure the data collected is complete, Trax has developed an unsupervised learning engine that monitors every session, identifies specific flaws, detects incomplete and duplicate captures and even cross-checks the photo’s GPS with the known store location.

  1. Augmented reality for a comprehensive shelf capture

With shoppers walking by, shopping carts being pushed around and other distractions, it can still be challenging to capture every aisle from end to end. Manual audits in the store are also prone to human errors like double counting. To address these problems, Trax leverages the immersive power of an augmented reality feature which flags areas on the shelf that users may have missed capturing and guides them in resuming the audit from the exact point they were interrupted.

A Complete Picture of the Store

To make confident business decisions, the data captured and collected through image recognition has to be clean, complete and free of anomalies. Enforcing rigorous data quality enforcement checks with completeness as a key pillar in the image capture is fundamental for the proper calculation and reporting of shelf metrics such as on-shelf availability, shelf share and more.

Watch this video to see how Trax enforces data quality in our Computer Vision platform.