Where Retail meets Tech

November 22, 2018

8 Key Questions to Ask When Choosing an Image Recognition Vendor (Part 1/2)

Written by Christophe Blot

Image Recognition (IR) is a perfect example of how Consumer-Packaged Goods (CPG) manufacturers can leverage digital transformation to win in retail. A 2017 Gartner report highlights why this is the case. When sales representatives are in stores, they can optimize retail execution activities in one click through photo-taking, using Image Recognition capabilities to monitor planogram compliance faster and more accurately.

Thanks to four main levers — speed, coverage extension, accuracy, and measurement consistency — Image Recognition can help CPG manufacturers maximize profitability by driving retail execution costs down and sales up. But as with any complex technology, it’s important to know what to look for when picking a solution.

There are 8 questions CPG manufacturers should ask themselves when selecting an IR vendor, which we will cover in our next two blog posts. Here are the first 4:

 

1.     What Do I Want to Achieve with Image Recognition?

business-case-for-image-recognition-for-retailA key business objective should motivate every new investment. Do you need IR to reduce costs, or would its primary goal be to uplift sales? Knowing how IR achieves this goal can provide a framework to measure its success. Assessing its total costs and benefits will then help determine precisely where and how your firm gets value for its money.

If you misjudge what the solution can do for you, it will likely cause frustration and prove counterproductive. Just as you wouldn’t bring a jackhammer to an archaeological dig, your IR solution needs to be consistent with the needs of your business.

Identify the business case for your IR solution, then use it as a reference point throughout the project. If your primary goal is to reduce costs, make a list of associated retail execution costs and determine how these will be mitigated - for instance, with lower auditing costs, or a reduced need for third-party data. The best IR vendors will work with you to build a solid business case around your exact needs and will help you keep track of your progress as you move along with Image Recognition.

 

2.    Does the Image Recognition Solution Deliver Highly Accurate Results?

accuracy-image-recognition-for-retail

A key motivator for CPG companies to adopt IR is its promise of greater accuracy and, as a result, more informed business decisions. We humans have our perks, but we’re also the sole perpetrators of human error. (Hell, it’s even named after us.) Over the last decade of reporting for our clients, we’ve seen that manual shelf measurements are usually off by 15% to 40%.

One of the main reasons for human error is that people tend to think on their feet and adapt to their immediate environment — even when it negatively impacts overall operations. If one of your sales reps feels that shelf KPIs don’t accurately reflect the reality of what’s happening in store, they might take liberty and change them accordingly. For example, if a product is Out Of Stock (OOS) due to late delivery, the rep might mark it as present or include it in the number of counted facings not to impact their client’s global compliance score. Since sales reps can go several weeks without visiting the same store, altering reality might make sense to them in this type of situation. But for CG firms, an OOS can mean heavy sales losses, up to 30% of total revenue according to research funded by P&G.

In addition to an immediate impact on sales, such actions can create a disconnect between what is believed to be happening in-store and the reality of the field, resulting in unreliable analyses and incoherent business decisions.

In other cases, sales reps might intentionally skew the numbers for their own self-interest — to achieve a sales bonus, for instance. Natural cognitive bias is another source of error in manual shelf measurements, as our brains tend to take shortcuts when processing a lot of information. This leads to a rapid, yet frequently faulty, evaluation of reality.

Image Recognition overcomes human error with deep learning. Deep learning algorithms train models to identify common attributes in a large set of images, and then recognize the same objects in new images and environments. This is the most effective IR technique for the retail environment: once the model is trained, it can quickly and accurately identify SKUs in any store. Leading IR solutions offer more than 98% accuracy, 12 to 38 points more than manual measurements.

But IR projects can often be relatively complex. As experts like Gartner rightfully note, they are part of a change dynamic within the organization, which is why they should always be assessed against the vendor's expertise and the value gained from the business case driving the investment. For example, leading IR vendors such as Planorama, who pioneered the use of Image Recognition for retail, have developed models that are already familiar with a large set of products, making the initial set up faster and smoother.

 

3.      How Fast Can Sales Reps Turn Insights into Actions?

speed-image-recognition-for-retailFor your IR solution to be cost-effective, reps need to be able to take corrective actions as soon as possible. This means that in addition to fast turnaround time for data analysis — leading vendors give results within 5 to 10 minutes — your solution should be easy to use. Thanks to extensive user research and testing, best-in-class IR applications offer sales reps an experience which is rooted in their own context to guarantee adoption and increase productivity.

For instance, leading IR vendors use the picture upload on the fly feature to ensure that every new photo taken in-store is readable and analyzable by the solution from the moment it appears in the app. Image recognition relies on pictures to digitize shelves, so ensuring a photo’s compliance by sending reps live feedback on photo readability guarantees that the right categories and products will indeed be assessed by the solution. This step is critical to not only produce accurate actionable results but also to increase efficiency, as it enables sales reps to retake non-compliant photos on the spot to optimize image recognition results. For brands and their field forces, it ultimately means faster store checks. Reps gain in efficiency and can spend more time on value-added activities such as optimizing visual displays or ordering top-performing SKUs. They can also increase their coverage capabilities.

 

4.      Can the Solution Adapt to My Specific Needs Today and Tomorrow?

Best-in-class IR solutions are dynamic and versatile enough to accommodate the changing needs of CG firms. For instance, they can adapt to complex store situations — narrow aisles, low lighting, etc. — and still provide accurate recognition results. This guarantees consistent insights across channels.

But complex store situations can mean connectivity issues, too. We don’t live in a perfect world with 24/7 internet access available in any location. Just like you may have trouble loading a webpage in certain areas, some stores have poor connectivity. Sales reps will inevitably find themselves in such situations. A great IR vendor will listen to your requirements and adapt to your challenges to ensure you have the most appropriate solution in every type of environment, no matter how specific.   

flexibility-image-recognition-for-retail

Your IR solution should also allow for extensive customization according to your KPIs, regardless of their complexity. A reliable vendor should be able to tailor and compute your KPIs the way you want them to be measured. Every company has their own way of measuring success. For example, some brands will closely monitor share of shelf while others will evaluate their performance based on core SKU eye-level placements, adjacencies to main competitors, or quality and quantity of secondary placements. Your supplier’s faculty to understand what you want to achieve is therefore critical  to offer you the best way to track progress.  

Lastly, the solution should serve the various goals and needs of its end-users: in store, offering reps simplified immediate insights to enhance their experience and empower them to take quick actions. At the HQ level, expanding and detailing these insights for reporting and decision-making purposes.

 

Would you like to know more? Read second and last part of this blog series or contact us directly! We’ll be happy to answer any question you have.

 


Topics: image recognition, retail execution