Kullfi Kumarr Bajewala, a prime-time musical series on StarPlus gained massive popularity within a few months since it first aired. And, many might say, it hits the audience with just the right emotions. So, what worked for the show that has recently completed a year of airing?
Before launching Kullfi Kumarr Bajewala, Star TV team was pondering over how they could create an impactful promo during the launch, and if they could optimise media spends. That’s when a Bengaluru-based start-up, Entropik, came into the picture — a company that built the country’s first Emotional Artificial Intelligence technology — AI that understands human emotions, according to the company.
For the show’s promotional campaign, the start-up mapped brainwaves of respondents that represented the target audience. With the help of this technology, they could study how the subjects perceived each character in the promo videos. While the video was played, emotions and expressions of the respondents were tracked, which helped in understanding the ‘high’ and ‘low’ points in the video.
Entropik’s Emotional Efficiency engine scored each promo based on the emotional response, predicting their chances of clicking with the audience. Based on the scores, Star TV could cut down 30% of its marketing expenses. And, the launch campaign went viral! According to the start-up’s co-founder, Ranjan Kumar, his client also saw 3x returns on their media spend of around Rs.50 million.
“We are the first ones to build such a platform in India. We essentially build technologies that track consumer emotions at their points of origin,” explains Kumar.
What’s on your mind?
Entropik has built three key technologies for its platform — first is brainwave mapping with the help of an easy-to-wear headset. It can track the respondent’s state of mind such as tension and emotional level, constantly. Second is facial expression analysis, which tracks emotions with a camera. Third is tracking eye movements using cameras. All three technologies are patent-pending.
Entropik’s systems are trained on a database of 14 million unique user emotions, which is embedded in its consumer-behaviour research platform, Affect Lab. The start-up took one year to perfect the emotional AI technology. Entropik’s clientele includes media houses, advertisement firms, consumer and automotive brands. It helps its clients optimise their customer support experience. “We provide this as an analytics tool for them to really know their end users at a competitive level,” says Kumar.
Making the leap
An employee of Citrus Pay (acquired by PayU in 2018), Kumar moved out of the company in 2016, wondering what could be the problem statement for the next 5-10 years that was worth solving. The chemical engineering graduate from IIT-Kharagpur co-founded RedCastle, a media communications start-up, prior to his stint at Citrus. That’s when he discovered a large part of marketing spends were ineffective.
He explains, “For example, advertisement conversion rates are 3%. That is ridiculous, and I felt that it was happening because people’s attention span was reducing, but the amount of content kept rising. And, all brands are trying to push their ads to the fore. So, how does a consumer-centric business create experience which is optimum and not too pushy?”
Kumar found that several consumer-led businesses were adopting technology at length. However, they weren’t always successful, as the consumer experience on offer was not always built to suit the customer.
In the case of media, its strengths – money and time – were wasted if a campaign did not strike a chord with the viewer. Kumar felt that intelligence should not just be ‘artificial’, but must have an emotional perspective. “You have to build a technology that completely resonates with the consumer. With that as the core motto, we started this company. It was a futuristic plan, and a very ambitious project for us,” notes Kumar.
Along with Bharat Singh Shekhawat, co-founder and head engineer of Entropik, Kumar started his company in 2016; co-founder and product chief Lava Kumar, joined later in 2017. The first year was dedicated to developing the technology from scratch. In early 2016, the start-up raised a seed round of $200,000 from angel investors Dileep Bhatt, president of downstream operations, JSW Steel, and Milind Chaudhary, director of Sea Global Services.
Scaling the business and identifying partners to reach a product-market fit were the pursuits in the second year. The start-up has been a part of a string of accelerator programmes ever since — the San Francisco-based Accenture Ventures Cohort (January 2017), Plug and Play (August 2017), Oracle Cohort (June 2018), and Intel’s AI Builder programme (November 2018). In the media segment, Entropik is part of Viacom’s VStEP (start-up engagement) programme. The association has helped them partner with Viacom 18’s business units such as Colors, Nickelodeon and Voot.
Ditch the questionnaire
Entropik mainly targets segments such as media and advertisement, and user-experience testing (or UX testing). In the media space, brands resort to conventional surveys or feedbacks to collect consumer insights. Even the technological advancements are limited to get insights from buying and browsing patterns. “Instead of spending millions of dollars on questionnaires to collect consumer insights, we have been able to quantify actionable outcomes,” Kumar points out.
According to Kumar, the questionnaires completely miss out on the subconscious element. Plus, there’s also a research bias while framing questions; whether the respondents understood the question the way the research team intended to ask. Was the audience able to quantitatively articulate their responses, and were the responses of a few influenced by the responses of their peers?
Entropik’s technology overcomes the human errors and biases, as the subjects are constantly monitored with the help of an electroencephalogram (EEG) headgear. The device monitors brain’s electrical activity and generates readings that can be translated into emotions felt by the wearer.
For any study, the base sample size is calculated with the help of technology that calibrates statistics on minimum respondents. More than 95% trends get stabilised by a specific value, and any more respondents don’t yield better results. Hence, the start-up studies around 50 respondents per target group for emotional analysis, and according to the team, this would suffice to get clear emotion analysis data. For brainwave mapping, the minimum sample size is 40; for facial analysis the base head-count goes up to 100; and for eye-tracking, the team needs at least 35 respondents.
Facial expression and eye movement tracking can be genuinely handy tools for marketers, ad folks, or even office spaces for that matter. “It can be in a retail environment where a customer’s facial expression could tell what his satisfaction level is; or in human resources, where video resume applications are picking up. We create a psychometric profile of the candidate and allow recruiters to screen resumes based on such profiles,” says Kumar.
Television Audience Measurement (TAM), a media research firm, is one of Entropik’s clients, and its CEO believes the tech has added value to their research. LV Krishnan says, “Earlier, we used data to find out how viewers responded the way they do to our content. But we could never find out why they respond that way. Now that we can, it’s a big step forward for marketing promotions. Entropik helps probe audience behaviour using technology.”
He adds, “For instance, one of the use cases is for OTT (over-the-top entertainment services) platforms to analyse if their promo content is good enough to create recall value in the minds of the customer.” Kumar claims that the typical customer sees a viewership jump between 120-450% with the help of Entropik’s services.
Testing 1-2-3
In case of user experience (UX) testing, Kumar explains the possibilities of Entropik’s technology, noting the company’s research for Accenture’s BFSI clients. The technology could identify for the UX user where the frustration levels are high, where the service’s app is asking too many questions or where the user is getting distracted and not navigating through.
“A classic example is when you get an OTP number. You have to leave the app, get the OTP number from the SMS, copy it, come back to the app and paste it. Your mental effort is much higher in such instances, and we can deliver insights to optimise this experience for the user. Our clients have been able to improvise accordingly and get 3-4x returns on the platform itself,” notes Kumar.
He explains, “If the UX of an app or chatbot is being tested, 100-200 respondents as per the target definition are invited and allowed to go through the whole UX experience with a brainwave mapping headset on.” Screen recordings of all such sessions are analysed and actionable insights are provided such as content being too cluttered or repetition of information.
“We are working with an insurance company that was asking customers to enter PAN card details, phone number and so on. But many users are not comfortable sharing those details on an app. A lot of these data are already collected somewhere else, for example your phone number. You can automatically capture such details and figure that part out. Asking the user to enter those details is leading to the ‘fear’ emotion set. In a five-minute session, the user becomes fearful by the third or fourth minute. That was a huge insight for the insurance company,” says Kumar.
In the automotive space, technology is expected to bring in high-end consumer experience even in low-end cars through live analytic solutions. Using facial recognition, the technology would be able to measure every second if the driver is fatigued, frustrated, restless, stressed, unwell, and create alerts accordingly.
Entropik claims that this tech will not drive up the vehicle cost significantly, and has submitted a PoC (proof of concept) to Daimler. The team is considering its applications in driverless cars, where it would monitor how the ambient conditions inside the car could adapt to the driver’s and passengers’ emotions.
The 22-member start-up currently boasts around 55 clients across India, US and Singapore, which include names such as Star TV, Viacom 18 and Sony. Other clients are serviced through major partner IT firms such as Accenture, SAP and Oracle.
“These brands are driving the digital transition for their consumer business clients. We enable them to measure the behaviour and emotions of their consumers in a scalable way,” he says.
The company also has use cases in the education tech industry, where they work with Byju’s. Entropik helps such brands position content to the end user. Kumar believes, in EdTech, one of the key challenges is to measure Cognitive Learning Index, the true reflection of a student’s learning process while they consume content.
Entropik partnered with Byju’s at an early stage to grade the app’s learning content. It tested the curriculum on a number of students. Today, Byju’s content is highly adaptive to the student’s learning capabilities, and Kumar claims this feature is to their credit.
Now, the question is, who is leveraging such technology? Kumar estimates, “70% of our business comes from media and advertisement, 20% from UX and the rest from others.” A SaaS (software as a service) platform, Entropik follows a usage-based revenue model. The start-up gets paid based on the number of creatives tested, or the number of respondents the client has tested.
“We charge on the basis of per video minute of testing, and then aggregate it across monthly billing,” The average ticket size varies depending on the vertical. “Media advertisement use cases are typically of Rs.300,000 per project a month. UX projects cost slightly higher — around Rs.500,000 to Rs.600,000,” he adds. The company recorded revenue of $1 million for FY18-19. Kumar expects the company to grow at 100% quarter on quarter owing to high demand and lack of rivals.
With ambitions to scale up operations, Entropik is focused on winning the media and ad vertical in Singapore and other south-east Asian countries. Europe is also on its radar, but crucially, it wants to keep a tab on more use cases where its AI engine can be leveraged; and south-east Asian nations offer fast moving consumer brands with high growth. Entropik is also plans on boosting headcount. R&D makes up nearly three-fourth of its employees. “Over the next year, we expect to be twice the size we are currently,” Kumar says.
He adds that Entropik is a cash flow-positive business right now, and they are targeting to scale up 4-5x in the next 18 months. The business raised $200,000 in its seed round (early 2016), and another $1.1 million in its pre-series A (mid-2018), led by BIF and Parampara Capital. The second infusion was a bid to boost R&D efforts and boost activities.
Mind, the new Gold mine
What made Entropik an attractive investment proposition? Ashwin Raghuraman, founding partner at Bharat Innovation Fund and one of the investors, says, “It is a very different venture. Brainwave mapping has been used in fields such as forensics for some time. But, it wasn’t being used in an innovative manner to analyse and decode customer response to visuals.”
Raghuraman believed they were investing in something that could disrupt the media space, and the creation and distribution of videos significantly, in terms of quality. “That’s because we all know video is going to be an important channel of communication. They were hitting a new-age solution to market and help people improve their content,” he adds.
They realised that the technology could be useful through a social perspective as well, and not solely in media management. “Facial mapping for car drivers could be crucial in saving lives. These guys are solving the problems of today and tomorrow,” he says.
According to a study by Reuters, the global emotion detection and recognition market is estimated to witness a CAGR of 32.7% over 2018- 2023. Kumar expects the market to reach $24.74 billion by 2020, with the presence of a few companies globally. For example, Nielsen has been offering some solutions with the help of brain-mapping, while Boston-based Affectiva is into facial coding based on emotional analysis. There are also a few companies that provide eye tracking. While Kumar acknowledges these players, he points out that they have been offering fragmented solutions, while Entropik integrates all these technologies. He considers Entropik’s huge data set of unique user emotions its USP. But, where did all this data come from?
“Currently, we have a repository of over 11 million unique human emotion samples. We have recorded over six million of these samples through our training and data collection exercise over the past two years. The remaining five million samples have been acquired through customers,” he explains.
Kumar assures its data mining process does not breach security issues, at a time when data and cyber security is one of the pressing global concerns. Entropik claims to be GDPR (General Data Protection Regulation) compliant, and doesn’t store any personal information; but only the derived intelligence in an encrypted way. “We have 100% rights to derive collective intelligence from this data, given it doesn’t have any Public Interest Litigation information,” he adds.
Kumar believes ‘data’ is the next ‘oil’, and categorises data further into ‘commodity’ and ‘niche’ data. “We are playing with niche data, which is very powerful. For example, a lot of companies will have consumer clickstream behaviour or data relating to phone calls. We are playing with data that is very fundamental and pure, and that is human emotion. It has to be captured from the source,” he notes.
So, will Entropik emerge successful? It’s too early to say, but Kumar is confident — “This is a $40 billion (according to a Reuters study) opportunity, and the market has three players. The industry is bound to attract more players. But the victor will be the one who has access to the largest amount of purest data.”
