Humans may claim to be rational beings, but our purchasing decisions are almost always driven by emotional impulses. Traditionally, product teams and market researchers have relied on user interviews, focus groups, surveys, or customer analytics to derive behavioral insights. However, predicting user or consumer emotions from the gathered data is tricky and not always accurate. What if there is a better way?
What if brands could not only gather what their users or consumers say but also understand how they feel – their true emotions – hopes, dreams, fears, and needs? With advancements in machine learning and artificial intelligence, it is now possible. And Entropik, a Bengaluru-based multimodal Emotion AI startup, has figured out a way to do it at scale.
Entropik has made significant breakthroughs in affective computing and has taught its AI engine to decode human emotions by reading eye gaze patterns, facial expressions, and voice tonality without using any fancy hardware tech. Their clients, P&G, Tata, Nestle, and ICICI, to name a few, are using Entropik’s multimodal Emotion AI solutions to make better product decisions and get actionable marketing insights.
At ProdWrks, we were curious to learn how Entropik built its Emotion AI and how brands can use it to get better user and consumer insights. Entropik’s cofounder and CPO, Lava Kumar, was more than happy to indulge us with the tricks of his trade and how they make it happen.
This is the story of Entropik.
Inception, Problem Statements, and Solution
Entropik was founded by Ranjan Kumar (CEO), Lava Kumar (CPO) and Bharat Shekhawat (Head of Engineering). Ranjan and Lava Kumar are veterans in the adtech and martech domains, with two decades of experience between them. Before starting Entropik, Ranjan had his own adtech startup, Red Castle, and Lava Kumar led product initiatives at cutting-edge martech ventures in the US, like Yahoo Gemini and Operative, to name a few.
The spark which led them to start Entropik began with a simple yet profound question: What if ads could be personalized based on users’ emotions? – Their first problem statement.
"Typically, any on-screen ad that you see is based on your browsing data, cookies, demographics, and interests. Our underlying problem statement was, what if we could personalize ads based on users' emotions as they engage with the content? Can we make ads contextually sensitive to both the emotion of the viewer and the content?"
To tackle this challenge, Entropik embarked on a mission to measure and understand human emotions. The initial exploration led them to EEG (electroencephalogram) – a medical device that helps derive people’s emotions by measuring brainwave activity. The EEG allowed them to quickly and accurately measure the users’ cognitive state (attention and engagement) and predict affective states (emotions being exhibited) while watching ads.
However, scalability was a hurdle. You cannot ship the gaudy-looking EEG to your users and expect them to wear them while watching ads, can you?
So, Entropik got their second problem statement – How do you detect user emotions at scale?
To solve for scaling, Entropik pivoted to human expression tracking (facial coding) and eye gaze tracking by harnessing standard cameras in laptops and mobile phones. It could also be cameras already deployed in public infrastructure in retail stores and other environments where ads are displayed. This enabled them to tell exactly what people are seeing and their emotional state, at scale.
“Let's say Deepika Padukone is in an ad. By combining eye tracking and facial coding, I can tell you whether she driving the emotion of the users or if the overall content is driving the emotion, or if it is the brand’s key tagline on the creative that's driving emotion. And with this data, marketers can do a bunch of different optimizations to drive home the message.”
With the core premise of eye gaze tracking and facial coding technology sorted, Entropik started its product journey in 2016, bringing behavioral insights at scale to brands with its multimodal emotion-sensing AI solutions.
Affect Lab and Affect UX: Entropik's First Products
The journey from figuring out the technology to applying it to real-world use cases led to the creation of Affect Lab, Entropik’s first product that provided brands with media and shopper insights.
Lava Kumar disclosed to us that he narrowed in on the two verticals of media insights and shopper insights after speaking to 80+ brands across the globe, including ITC, Tata, and P&G, to figure out how Entropik’s emotion-sensing technology could add value to them.
“We were very clear from the onset that measuring emotions just for the sake of measuring is not going to add value to our clients if they are not getting ROI out of it or if they are not able to make actionable decisions. So, we dug deeper by conducting user interviews and zeroing in on two specific verticals within market research and consumer research - one was to provide media-related insights, and the second was to provide shopper-related insights.”
Media Insights is all about helping brands test ads on real users and how it plays on their emotions, assisting marketers do a creative analysis and get to an ideal ad state where engagement rate is high from start to finish.
Shopper insights help consumer brands with everything from getting their package design to placement right so that it promotes positive customer perception of the brand thereby increasing brand loyalty and brand equity.
Seeing the success of AffectLab, Entropik added another product offering – Affect UX, which uses the same eye tracking and facial coding technology but is designed to help product teams build engaging user experiences for digital products using Emotion AI.
Product Evolution - Decode and Quatalyst
This year, Entropik underwent a sea change, transitioning its product portfolio to serve as a full-scale self-service SaaS platform that market researchers can use themselves. Affect Lab has now evolved into Decode, a DIY consumer research platform, and AffectUX is now Quatalyst, a DIY user research platform.
“Decode and Quatalyst are completely self-service and very intuitive. We keep the overall user experience as simple as possible because that's our forte, and that's what we believe the technology is for. So, we apply those same principles when we build our own products as well.”
Explaining about Decode, Lava Kumar says, “It is a pure consumer research and market research platform. Anybody who wants to understand consumers at scale, primarily CPG companies, are our major customer. We work with Unilever, P&G, ITC, and pretty much all major CPG brands across the globe, helping them understand consumer behaviour at scale.”
And when it comes to Quatalyst, it benefits digital-first product companies the most as it helps them deeply understand their users emotional states while using their products. It works across verticals – BFSI sector, the hospitality industry, travel booking sites, or anybody who does primary business in a D2C model through websites or apps.
“Quatalyst also benefits companies that build applications for their partners or for internal consumers where user experience matters a lot. These companies use our platform to test the prototypes, mock-ups, or live websites to make sure whatever they're building is in line with what their customers want and what their end users expect it to be,” adds Lava Kumar.
The Underlying Tech Behind Quatalyst and Decode
From a product-technology standpoint, Quatalyst and Decode share a common underlying platform, which supports all the core technologies that Entropik has built over the years. This includes facial coding, eye gaze tracking, and along the way, Entropik has also added another vital technology layer that adds more depth to emotion sensing – Voice tonality.
“Voice tonality technology tracks your word choices, tone, frequency, pitch, and 30 other variables to do a text sentiment analysis and determine your emotion. Combined with eye gaze tracking and facial coding, the three technologies together give a holistic outlook of users’ emotional states when they see an ad, use an app, or browse a website.”
These integrated multimodal emotion detection technologies (voice tonality, facial coding, and eye gaze tracking) are available natively on Quatalyst and Decode. On top of emotion sensing functionality, Quatalyst and Decode also incorporate GenAI features.
“We have trained our own large language model based open source which allows brands to ask any questions, and get responses in milliseconds based on historical research that they have done. For instance, they can ask how their brand is performing, questions about the last six months’ trends for their brand, what they should be working on, and what they should not be working on.”
Building a Deep-Tech Moat and Finding Niche Use Cases
Explaining their decision to focus on multimodal emotion sensing AI as a niche, Lava Kumar says that it was rooted in the logic of building Entropik as a deep-tech, IP-led company that could stand the test of time and one day cater to diverse verticals beyond advertising. And they have 17 patent claims filed, with a couple of them approved in USPTO to make a strong case.
"We didn't want to take a regular problem statement and solve for it. We wanted to make meaningful deep-tech solutions from a multi-vertical standpoint. We want our multimodal technologies applied uniquely to solve problems across industry verticals ranging from edtech to healthcare and beyond.”
While Entropik primarily focuses on consumer and user research today, Lava Kumar’s vision of seeing Entropik’s multimodel technologies used in other industries is already taking shape. The company is actively collaborating with a US-based agency on projects related to postpartum depression measurement using voice tonality.
“There's no real way to measure postpartum depression today. During counselling they ask preset questions to determine if the mother has postpartum depression. But statistically, it’s not accurate. We are working with a couple of hospital networks in the US to measure postpartum depression using our technology so that no mother, right after pregnancy, has to suffer without knowing whether they actually have it or not,” says Lava kumar
He adds that there are other niche use cases outside of consumer research that Entropik sometimes dabble with as it involves leveraging technology for the greater good.
As we dig deeper into the product philosophy of Lava Kumar it became clear to us that Entropik’s core driver is to make emotion-sensing capabilities mainstream and make deep personalizations possible for brands and their users – a clear differentiator in their growth trajectory amongst competitors who are venturing into the affective computing space.
“There is no real personalization unless you get to know your users at a deeper level. And there is nothing more personal than really knowing what people's emotions are at any point in time when they engage with your brand. And with that, we want to build a long-lasting, deep tech, IP-led company so that we can explore more cross-platform applications as we scale up and as our company and our idea become successful,” declares Lava Kumar.
He also adds, “So, that basically is our key driver. At the end of the day, we are building a transparent layer between a consumer and the brand, enabling brands to see consumers without any filtration, without any sort of bias, and without asking any questions – more implicitly and more naturally.”
Entropik's Vision for the Future
As Entropik continues its journey at the forefront of emotion-driven research, the company envisions expanding its impact across multiple verticals.
“What we are working on is very cutting edge… bleeding edge. Affective computing as a space didn't exist ten years ago. The term itself didn't exist. It's an upcoming technology where emotionally and behaviorally AI will be ingrained into the overall cycle of products and brands.”
To fastrack the adoption of Emotion AI, Entropik plans to make it’s technology available to developers to find new use cases, like how OpenAI mainstreamed GenAI technologies through its APIs and SDKs.
“We are planning to make our technology SDKs available in the public domain for a larger community of developers so that they can build different apps. This technology is completely disruptive in terms of what you could do with it. So, we're doubling down to package it and put it out to the public domain for anybody and everybody to pick up this technology and build some unique stuff. Let's see how that scales up.”
Despite Entropik’s tremendous achievements and breakthroughs in the field of affective computing, which is getting them growing list of customers, Lava Kumar says that they have hardly scratched the surface.
“There's a substantial market out there. Companies have now started to focus on their digital presence aggressively. You just cannot put a website out anymore and expect people to be happy with it. You need to provide a differentiated and streamlined experience while making it as simple as possible. If you're unable to provide that experience, customers will switch to somebody who actually provides it because the switching costs are really, really low.”
In a world where emotional connection defines brand loyalty and user engagement, Entropik stands as a testament to the transformative power of technology in understanding and harnessing human emotions for a more personalized, meaningful future.
