Machine learning software enables Mars Petcare to measure how appealing pet food images are to online consumers.

Mars Petcare understands what shoppers like in store, says Roman Vorobiev, director, design and artwork management.
“But we had very little understanding of what types of images entice online shoppers to buy, he says.
The COVID-19 pandemic shifted the shopper experience online. Demand for Mars Petcare’s pet food products online grew “exponentially,” Vorobiev says. “We realized we needed to find out what drives the effectiveness of content online.”
“We doubled down on e-retail,” Vorobiev says. That included investing in Vizit artificial intelligence image analytics software.
Mars Petcare sells its products through Amazon brand stores including Cesar and Pedigree. Retailers worldwide sell its products in stores and online. Retailers include Walmart Inc., PetSmart, Target Inc., and others.

AI image analytics show what people like to see on screen

Mars Petcare began using Vizit in early 2021. It first tested how its existing images scored. 
Roman Vorobiev, director, design and artwork management. Mars Petcare

Roman Vorobiev, director, design and artwork management. Mars Petcare

Vizit’s software produces a score on a scale of 0-100. The higher the rating, the more appealing an image is to a group of online consumers. Vorobiev wanted to link how these scores could lead to higher conversion rates.

From May 2022 to October 2022, images with a Vizit score of 60%-70% achieved 30% higher conversion.
The other part is understanding what consumers engage in. Vizit’s machining learning software collects data points. It predicts and recreates the visual perspective of a target audience.
The result? Consumer preferences varied depending on where in the world they shopped.

Online shoppers prefer different details

Mars Petcare can toggle between different audiences/consumer groups in Vizit. Voroviev can see how the same image appeals to different audiences.
For example, consumers in Mexico prefer yellow and green on packaging.
Meanwhile, U.S. shoppers prefer flat 2D images. Images show the front of the dog food packaging featuring a single dog. But, in Japan, online shoppers prefer a 3D view of the food packaging with many dogs.
“Something like 20 dogs on the packaging,” he says. Also, Japanese consumers prefer to see the side panel.


In Mexico, online shoppers prefer green and yellow colored packaging for Pedigree dog food.

In Mexico, online shoppers prefer green and yellow colored packaging for Pedigree dog food.

Using AI in design

Mars Petcare designers now incorporate the software into their creative process. Vizit scores help Vorobiev decide details, including how many dogs or cats to place on packaging and which breeds are most appealing.
Vorobiev reviews Vizit’s benchmarks to measure how customers are responding to various images. It helps him understand how engaging Mars Petcare’s content is through images.
“We see this tool [Vizit] as almost like a visual spellcheck,” Vorobiev says. “Think about how many decisions are being made by a group of people designing this content. Decisions like, should text be bold or plain text? Should we photograph food from the top or should it be three-quarters down? There are a ton of decisions, and my job is to reduce the noise in decision making design.”

Harvesting images

Before using Vizit, “It was me and a couple of other people harvesting images by hand,” Vorobiev says. “It took us a lot of time and the integrity of the metadata was not necessarily the best.”
Metadata includes information such as aperture, resolution, shutter speed, camera brand/model, date and time the image was created, and GPS location.
Within the last year, Mars Petcare started automating harvesting images. “We got more clarity and accuracy from the metadata,” he says.
Mars Petcare can also separate images into different categories, Vorobiev says.
“Differentiating between cat and dog photos is easy for an AI sorter to do,” he says. Differentiating between ingredients or the back of packaging can become complicated, he says.

Quick changes using AI

Mars Petcare tests different image setups during a photoshoot. Designers can alter images in real-time as they photograph them.
“One of my content managers was able to make improvements almost monthly,” Vorobiev says. “Or he met some of the retailers weekly to see what we should change.”
Over time, the algorithm learns more. Right now, Voroviev tests images through Vizit every few months. He says he plans to cut back to once a year.

What is image analysis in AI?

Vizit has collected millions of images that specific groups of people have been exposed to online. It is data based on their shopping behavior and preferences. The software uses syndicated data. Merchants buy data collected by market research firms.
“We can sort of reverse engineer their visual experience, including the things that they’re likely to have seen,” says Jehan Hamedi, founder and CEO of Vizit.
Triggers, or what appeals to consumers when viewing images, are important. There are an infinite amount of ways to photograph an item — different angles, lighting, positioning, coloring. It is impossible to create a rule-based system of combinations that will “always work,” he adds.
Vizit’s software is an annual subscription service scaled on the volume of content. 
The product images are reproduced with permission of Mars, Incorporated. All rights reserved.

Sign up

Stay on top of the latest developments in the ecommerce industry. Sign up for a complimentary subscription to Digital Commerce 360 Retail News.


Follow us on LinkedInTwitter and Facebook. Be the first to know when Digital Commerce 360 publishes news content.