Digital Signage

Anonymous Video Analytics

The AVA Study MaxMedia Retail Lab, with Tayler Jordan, Behavior Analyst, and Addison Lange, Marketing Technology Coordinator, analyzing data.

AVA accuracy brings ROI to the retail integration package.

The role of the systems integrator in retail covers more ground than other verticals, where audiovisual systems are largely focused on communication. In retail, the integrator’s purpose is to deliver a comprehensive ecosystem that engages consumers and ultimately drives sales. There are few, if any other, verticals where return on investment and revenue generation are as significant as the quality of the technology and its content.

Retail Architecture

Although the hardware element of this ecosystem delivers the engaging visual and audio content presented to shoppers, the systems integrator’s retail architecture needs to leverage software-defined technologies that reinforce the revenue-generating value. This includes the incorporation of big data and analytics into the IT backbone. As a key software integration strategy, these important components will ultimately help the retailer quantify the effect that in-store technology plays in driving revenue.

Online retailers have always had the advantage of understanding shopper data. However, new advances in anonymous video analytics, or AVA, allow the integrator to deliver a reporting service that rivals ecommerce within their overall scope of services. AVA gathers and evaluates key data from video as it relates to the shopper’s path, dwell time and engagement inside a store. It combines hardware and software solutions to provide other key demographic data, such as gender, age and race. AVA will soon capture and interpret other nonverbal cues, such as smiling or frowning, to determine shopper emotion, as well.

For the retailer, acting on the right information gleaned from AVA can create proactive changes that positively affect operations, marketing and messaging efforts. Ultimately, these analytics can boost their bottom line. This is a win-win situation for the integrator and retailer alike: The systems integrator is providing a service the retailer needs, instead of forcing the retailer to look to third parties for a separate managed service.

Naturally, many of these technologies are just emerging and not yet part of the everyday lexicon. To better understand these advancements in AVA, and how they could help their own clients be more successful in the retail space, MaxMedia conducted a study on the accuracy of three AVA suppliers.

Setting Up The Study

The study checked the accuracy of three AVA suppliers across the following major recognition categories: headcount, gender, age, distance, gaze and emotion. To create a reliable study in the lab, Behavior Analyst Tayler Jordan and her team simulated the inside of a retail environment. Three cameras (Logitech C615 HD webcams recording in 1080p resolution) were set side by side above a bank of digital signs.

With the digital signs in place, cameras mounted and software in place, passerby and gaze traffic was set at incremental distances to simulate general traffic patterns, including the various distances of shoppers as they walk by and consume digital content. To simulate real-life shoppers, individual men and women, groups of men only, groups of women only and mixed gender groups passed into video frame.

The suppliers of AVA for MaxMedia Retail Labs’ study were Quividi, Emotient and IMRSV from Kairos.

Here are some of the results, by category:

Headcount Accuracy

Headcount Accuracy chart
Headcount Accuracy chart

It’s always good to get an idea of foot traffic in a store. So, the first recognition category tested was headcount accuracy. In this situation, first individuals then groups passed before the signage. Quividi was the most accurate in both tests, followed by consistent Emotient and IMRSV results.

To introduce a variable, some participants then wore eyeglasses in an attempt to challenge the eye-recognition technology that registers people. In all cases, there was a 12.5% reduction in accuracy because fewer people were “seen” wearing eyeglasses.

Gender

Gender Accuracy chart
Gender Accuracy chart

Determining whether men or women are engaging with digital signage has a significant bearing on how a retailer communicates. To determine gender accuracy, groups of women, groups of men and mixed groups passed the cameras. Again, the three AVA suppliers all noted gender difference fairly accurately.

Many AVA suppliers evaluate hair length, face size, height and other characteristics to make their gender assessment. To add a variable to this case, women put their hair up in an attempt to fool the systems. The trick worked. There was a 12.5% reduction in gender accuracy: Women were counted as men more often when their hair was tied back.

Age

Age Accuracy chart
Age Accuracy chart

Plenty of products and services ultimately segment by the age of the consumer. That’s why the study sought AVA age accuracy, as well. Most age recognition technology looks for key facial determinants, particularly around the eyes, to provide an estimate or age range.

In the study, only Quividi and IMRSV could tell the age of a shopper accurately. These two systems were able to determine that recognition category with 96% and 95% accuracy, respectively.

Distance

Distance Accuracy chart
Distance Accuracy chart

To tell if people were recognized at various distances, walking paths were established in increments of 5, 10, 15 and 20 feet. At close range, all three AVA suppliers did well to detect people. However, at greater distances, only Quidivi and IMRSV could do so. In a scenario where different people were set at different distances, the results were similar.

Gaze Time

Gaze Time Accuracy chart
Gaze Time Accuracy chart

Knowing how long consumers look at a communication can be just as vital as knowing who is looking at that communication. The study had simulated shoppers glance at the digital signage as they passed by and then, during a separate test, had them hold their gaze in front of the signage for 30 to 45 seconds.

Quividi was the only AVA supplier that detected glances with 75% accuracy. And Quividi and IMRSV detected those who held their gaze with 100% and 75% accuracy, respectively.

Emotion

Emotion Detection Accuracy chart
Emotion Detection Accuracy chart

AVA suppliers that capture emotion love when shoppers wear their hearts on their sleeves. Test shoppers put on their happy, neutral, angry and sad faces, and went before the cameras.

There’s a reason it’s called Emotient. When shoppers were a few feet from the camera, this AVA supplier detected happy and neutral reactions at 100% accuracy, angry ones at 90% and sad expressions with 60% accuracy. At the time of the study, Emotient was the only supplier of the three tested that offered emotion detection.

Thanks to AVA technology, retailers are learning more and more about consumers while they shop. In this test, suppliers Quividi and IMRSV provided accurate demographic data when information about traffic numbers, age, gender, distance and gaze time was desired. Emotient proved to be a strong AVA supplier for retailers looking to get a firm read on consumers’ emotion because it’s an outstanding tool for emotion recognition at very close range.

The information gained from these analytics is valuable to the overall return on investment, accelerating payback for the cost of implementation for the system and its upkeep. This makes for a happy retail client, potentially opening the doors for return business for the integrator as system needs scale moving forward.

The takeaway for the systems integrator is that the in-store experience promises to be more intuitive, efficient and profitable for retailers using AVA, especially as the technology advances, accuracy improves and new functionality is added to this quickly-evolving platform.

Previous ArticleNext Article

Send this to friend