shelfwatch

The present planogram compliance assessments are often time-consuming and difficult to manage around work-peak. Meticulous manual input is required to guarantee that products on the shelves match with the planogram. The lack of visibility and up-to-date data prevents consumer goods organizations from tackling problems proactively. During a crucial sales period, a lack of data can lead to suboptimal decisions.

According to a study, “As many as 81% of companies reported that they’re unsatisfied with their ability to execute at retail. Another 86% said they’re not satisfied with their trade promotion efforts”.

But with ShelfWatch, all these redundancies can be tackled quite easily. A powerful and hassle-free tool, Shelfwatch is capable of running on a wide spectrum of retail channels. In this blog, we walk you through all the aspects of Shelfwatch that make it stand out amongst the existing Image recognition software solution in retail.

1. Real-Time, Offline Image Quality feedback

shelfwatch

Image Quality is an important criterion to ensure high accuracy of Image Recognition. SKU level recognition or price tag compliance is only possible when the image is not blurry and free from glare. ShelfWatch’s mobile app, known as StreetSnap, has a real-time image quality algorithm that can detect poor quality images and instruct the sales rep to retake photos. This detection works on the device and therefore, this functionality of the StreetSnap app is available in offline mode.

The reps can easily take high-quality images even in a no-internet zone and the images are automatically uploaded whenever an internet connection is available. In our experience of working with manufacturers, we have found that before using ShelfWatch, 15-20% of the images collected in the field were too low quality to be analyzed by AI or in many cases, by humans as well. This often leads to unnecessary delays and incomplete analysis. Existing Image Recognition vendors put the blame of reps in case of blurry or glary photos and put the onus on manufacturers to train their busy reps.

An ideal data collection app should be robust and smart to ensure high-quality photos are collected without any additional training for the reps.

2. De-duplication

shelfwatch

Very often it happens that while collecting data, the reps take multiple images of the same shelf from various angles. This is a serious problem as it may lead to double counting of shelf metrics (such as share of shelf) which in turn affects insights. ShelfWatch masters this problem very efficiently. Shelfwatch’s de-duplication algorithm improves data quality by detecting duplicate images and ensuring the metrics are not double-counted.

We also leveraged this algorithm to detect fraud in regular audits of retail execution for a tobacco company. The field auditors would often submit an old image to indicate that they have completed the audit. Using the de-duplication algorithm, we were able to surface such instances and reduce the possibilities of fraud in field audits. Within three months of integrating ShelfWatch, there was a 90% improvement in the data quality leading to reliable insights.

3. Integration with SFA and DMS apps

shelfwatch

While ShelfWatch provides its own app known as StreetSnap for capturing data in the field, we do understand that sales rep are already using handheld provided by the Salesforce automation vendors and will find it cumbersome to switch between multiple apps in the field.

We have integrated our solution with multiple SFA vendors and all the features of StreetSnap such as real-time image quality check and real-time shelf insights work in the integrated solution as well.

4. Fast to train and setup ShelfWatch

shelfwatch

Under the hood, most of the Image Recognition engine runs a neural network to detect SKUs and POS Materials in retail stores. However, neural networks, especially deep neural networks are notorious for needing a large amount of data to train them and get 90% and above accuracy.
Also, the training data needs to be manually annotated before it can be fed to the neural network. An example of an annotated images is shown below

shelfwatch

However, a large manufacturer will have 200 – 300 SKUs across multiple categories of their own brands and another 100 – 200 SKUs that they may want to track for their competitors. Generating a manually annotated dataset that covers 300 – 500 SKUs is a tedious and very expensive task.

Most Image Recognition vendors will take 90 – 120 days setup time during which they collect and manually annotate data. As you can imagine, this is an expensive and time-consuming process and does not scale well for new product launches or during peak promotions time.

Setting up Shelfwatch is a simple, two-step straightforward process. First, you need to share only one image of the SKUs that you want to track. And second, ask your field reps to take images of the retail outlet’s shelves using our mobile application. Shelfwatch’s algorithm is trained in such a manner that it automatically analyzes the images to give out a competitive analysis like share-of-shelf, and planogram compliance.

5. Cost-Effective

shelfwatch

ShelfWatch has been made with state-of-the-art technology to give out optimum results without having to spend a lot of money. With our superior technology, we support low operational costs because of the fewer resources required in setting up ShelfWatch. Our algorithm controls the data quality at the collection level to bring out the standard, objective analysis

Found this blog useful? Read this blog to know more about how Karna products provide effective solutions to traditional retail execution methods to improve brand presence and visibility.

Want to see how your own brand is performing on the shelves? Click here to schedule a free demo.

Share

Any thoughts?