New Qualtrics integration to streamline analysis of open ends!

7 Questions with our CEO and Founder Jared Feldman on Happy Market Research Podcast

16 min read

| December 2, 2020

Canvs CEO and Founder Jared Feldman recently joined Jamin Brazil on his podcast, Happy Market Research.  The two discussed Jared’s past,  how to navigate a successful career in market research, and the future of market research. Learn interesting tidbits about Jared as well as how Canvs built a platform that enables anyone to take their survey results from any provider and run them through the Canvs engine for immediate insights.

Below is an abbreviated transcript of the discussion, listen to the full recording here: https://happymr.com/jared-feldman/ 

Jamin Brazil: It’s interesting how you have, I think it was Steve Jobs, right, and his famous dots. You don’t know how the dots connect, but when you look backward, you see it so clearly. And it’s interesting how the upbringing, where you had fast friends on a regular basis, I imagine those – That kind of a context would create the need for you to quickly access the emotional status of people and really over-index on the EQ side of things so that you could fit into the – Because moving is hard. Moving states is really hard. It’s funny how now, all of a sudden, you started a successful emotions management company. Hilarious.

Jared Feldman: Yeah, absolutely. I mean, it did highlight for me, naturally, the value of EQ and just understanding intrinsically how emotions change our behavior, it changes how we feel about a situation and our perspective, and then ultimately what we do about it. And, unfortunately, in the business world, folks really don’t care how you feel, they care what you do because of it. And, so, there then became this interesting opportunity to figure out how to connect dots, sort of as you mentioned looking backward, how do we understand how people are feeling, and then, ultimately, what decisions those things drive. And, for me, I became very interested in people and understanding sort of complex emotional systems and was exposed to a lot of different people in different regions. The friends I made in Virginia versus New Jersey versus Texas versus Connecticut versus New York, got a sort of a good sample size, so to speak, and was exposed to a lot of different dimensions of humanity. And this is really the underpinning, is the emotion.

Jamin Brazil: In fact, I just had an interview with the Head of Customer Experience at Disney Parks. And in that interview, she talks about how important it is to connect at an emotional level, but then also, in the US, how diverse we are from our points of view. And, so, just going to the sample size that you just – Or the sample frames that you just mentioned, you get a lot of differing opinions based on a lot – All sorts of stuff. And because we as Americans have such a diverse point of view, it becomes even more important that companies are able to measure emotion. And getting that information quickly is, I mean, now it’s the speed of light, right? So that you can make correct, informed decisions.

Jared Feldman: It’s a really great point that emotion is transient by nature, it’s impermanent. And its relevance has a deadline. It’s important right now, but it declines in value over time because, well, it’s likely to change. And one of the things you just touched on implicitly is this idea of quantification of emotion. Like, how do we actually quantify and put a percentage – What percentage of folks are feeling this way? And, then, how is that trend moving? Is it – Are more people beginning to feel this or less? And it’s a really key thing in this day and age, especially as our expressions of emotion have become so – It’s like everybody has a voice. Expression is totally democratized. And just in the snap of a finger, you can have new waves and new sort of types of emotions that are really driving all sorts of powerful movements. And one of the quotes that I heard from Adam Bain back when he was at Twitter was that the entire business, the monetization model of Twitter is to monetize emotion. It’s a quote that I’m paraphrasing slightly, but it’s a quote he made in a Wall Street Journal article. And it really struck me because, also, Twitter was the very first dataset that we started to look at, and it was all the public and unsolicited tweets about content and integrations and advertising and trying to understand people’s emotions. And when he framed it that way, it’s almost all monetization is the monetization of emotions, if you think about it. And it becomes really important to not just sort of anecdotally understand using gut or art or previous kinds of methods that aren’t necessarily scientific. You now have this ability to be able to at scale and in real-time quantify how populations are feeling. Then it becomes really exciting as the researcher, as a CX professional, like, “What do I get to do about this? What does this mean for me? How do I make more empathetic decisions? And that, how does that feedback cycle accelerate such that we’re able to move and stay current with culture?” And all those things open up with things like Twitter at our disposal in today’s day and age.

Jamin Brazil: I think it was Jeffery Moore’s book, Crossing the Chasm, where he talked about the only asset a company really has is the relationships with its customers. It’s the products or the services that are really just a euphemism for wanting – People wanting to do business with you. And I think that’s – What you’re saying is exactly right. That just hits me right in the bull’s eye, which is we have to be – At any kind of a company, even if it’s a B2B, emotion plays a big part in the overall transaction. And that customer relationship is really the – Is the centerpiece or is the core asset. And I think it’s – I would like to unpack a little bit, give us some context for Canvs AI. How are you guys enabling companies to assess and diagnose their customers or constituent’s emotional states?

Jared Feldman: Sure. So Canvs’s mission more broadly is to make the world more empathetic. And we do this through patented AI and machine learning technology that ingests short-form text, public unsolicited dialogues happening on Twitter and Facebook and YouTube, or private and solicited, for example, open-ended survey responses that can at scale understand and detect what we term unnatural language. This is kind of the crux of the technology that we’ve developed, is we know how to deal with all the strange ways in which people express themselves in the written form. Unfortunately, none of us speak properly, certainly on the Internet, certainly when we’re having conversations quickly and in rapid-fire fashion. And, so, what the technology does is it’s really good at dealing with modern dialogue, with strange phrases and misspellings and emojis and all that just – Especially Millennials and Gen Z. It’s the first system of its kind that really with precision understands how people feel. So, Jamin, you say, “That ad, it was freaking hilarious.” And I say, “That ad, with just four crying laughing eye emojis.” And you and I are basically saying the same thing. And you use some strange language and maybe kind of hyphened something or misspelled something, and I didn’t actually explicitly say anything, really. I just noted the ad and used some non-textual cues like emojis. And the exercise for a researcher is to understand that those pieces of text share a common idea. And, so, what we’ve done is we’ve developed a framework where we have 42 core emotions that represent the most common ways that people talk about their feelings. And this was done by studying tens of billions of signals on social media. This is a little bit of a different path that most research companies take, where we started with social data, and it’s what trained our systems to be really, really competent at modern dialogue and all the crazy ways in which people express themselves. And, so, we started very verticalized with using social data and creating a feedback loop largely for the media, entertainment, and advertising industry. It’s the number one spender on research in America. And it turned out that every single network on the planet knew their operational data cold. They knew exactly how many viewers watched their show the night before, but they didn’t know why it was happening. And, in fact, it would end up taking days and running a survey or a focus group or doing a lot of manual digging. And, so, Canvs developed a standardized measurement of every single TV show. And TV is really just sort of a pronoun for content right now. So it’s whether it’s a linear show on NBC or an OTT program on Netflix, Canvs is quantifying how passionate the audience is, how funny the comedy was or how boring, how much love there was for the main character or actress, etc. And doing it at scale every day in a syndicated fashion, so that Netflix can benchmark their OTT programs against Hulu and vice versa. And that’s where we started using social data. And that expanded to measuring campaigns on social and things of the sort. And what we’ve been really focused on over the last 18 months is the application of this technology in open-ended survey responses. And this has been a huge need state that we’ve identified that almost every researcher and every CX professional will nod their head and tell you how important open-ended survey responses are, but when you ask them, “How do you use it?” They tend to roll their eyes. It’s such a frustrating experience for them because open ends are insanely valuable, but they’re just impossible to quantify without hand coding. And most researchers, most of the time, are spending hours in Excel or an equivalent, trying to hand-code open-ended survey responses to figure out, here’s what Jamin said. Here’s what Jared said. What are the common ideas and responses? And, so, we built a platform that enables anyone to take their survey results from any provider and just run them through the Canvs engine, and it will in a few seconds replace the hand-coding. It basically organizes the conversation, weeds out spam, it tells you how people are feeling, it helps you understand the main topics of discussion in a very flexible, easy to use interface that lends confidence to the researcher. And we found that our partners in media, like Disney, love it, but also outside of media in gaming and CPG, and we’re learning that there’s a really big opportunity to help researchers everywhere sort of get out of the weeds and really focus on storytelling. And, so, that’s our focus. That’s all that we do, is we focus on taking people’s voices that they give to you on Twitter or that they feed to you via a survey response, and try to help you understand how they’re feeling and why. And that, ultimately, we believe should help enable more empathetic decision making by organizations.

Jamin Brazil: Me being a survey guy, is a survey is really just a conversation at scale. So if I own a corner market or, let’s say a small restaurant, a family-owned restaurant, I don’t need to do an NPS survey because I’m talking to my customers probably every day. I’m seeing what’s happening on the floor. However, if I open up multiple locations, now it’s impossible for me to keep my finger on the pulse of the consumer. And, so, now it’s required of me that I start doing some sort of a mechanism to gauge how well my customers are connecting. And that’s why we, traditionally, have used surveys, because open-ended questions are just too complex and take too much time for me to analyze. The really interesting thing about what you’re doing is, in a lot of ways, it could become a better survey.

Jared Feldman: Yep. All of a sudden, Jamin, you can ask less closed-ended questions. You can hear more of the customer’s voice. Instead of saying, “Do you agree or disagree that the service was OK? Do you agree or disagree that you like this character?” However, you would sort of end up arriving, and, “What are your favorite things about this? What are your least favorite things about it?” Just, like, “How was your experience?” And letting people give you their honest truths. And enabling researchers with that technology is awesome because then – It’s like a superpower. Because, before, in order to get that sort of unvarnished truth, you would have to have very non-scalable, one-on-one conversations. But as you said, if the survey is this conversation at scale, then the conversation should scale. And that’s really what the technology is trying to accomplish.

Jamin Brazil: So how is Canvs AI qualifying the emotional score that you’re attributing to the brand or to the participant? And the other thing I’m really interested in is, how are you dealing with any sort of biases that may be embedded in the algorithms?

Jared Feldman: Both of those are very important questions. So let’s talk about quantifying emotion and how you do that. Because the exercise is, well, first, one, Canvs doesn’t actually quantify emotion. We quantify the expression of emotion. And as part of that, what we’ve developed over the last six years now is, what is – What we believe to be – And it’s hard to sort of prove this, but it’s the world’s largest ontology of language. We have the trillions of expressions that our system is looking for, which, to us, means that that user is feeling something. And this is – Was painstakingly built by hand, and then supplemented with the machine learning algorithms, and then scaled exponentially over the last couple of years. And in part was designed to update every 24 hours because our founding thesis is that not everyone really has a voice if you can’t understand everybody. And language is just changing so quickly. So the first answer to your question is having a comprehensive understanding of what language people use when they’re expressing something emotional. Now, we talked about the open end semantic problem a little bit, it’s actually not just an emotion problem. The survey problem, by definition, is a question and answer problem. It’s like I have this question and the exercise is, “How can you help the researcher arrive at the answer in as few clicks as possible?” That’s the whole game as we see it. And, so, to that end, there are additional types of analysis that we need to do, like topical analysis and more theme-based analysis, there are classification problems like Jamin and Jared are both people, and so what percentage of folks are talking about people or something else? But, broadly, from an emotion standpoint, which is foundational and is a unique value proposition for candidates, we’re effectively reading the open ends. And we’re looking for clues that this person is feeling something, and then we’re trying to understand why they’re feeling that way, and doing all sorts of qualification around it because, Jamin, I think you’re pretty cool, man. But, also, the weather is cool outside. And the same language, if about something in particular, could mean something totally different. And, so, there’s a lot of sophisticated edge cases that have to be considered as you think about dealing with modern language. But it’s just something that we’ve been focused on wholeheartedly for six years. And this is the recall problem. Basically, if you give me 100 responses and 80 of them have some sort of emotion in it, how good is a system at detecting all 80 of those things, regardless of what the emotions were, but is this person emotional, yes or no? And this is the first KPI. It’s a measure of passion. We call it the reaction rate, basically, how emotionally charged is a group of respondents or a group or a given population? And this is kind of – If you’re doing ad testing, for example, or if you’re trying to pilot a new product or get people’s perspective on something, this is the first order of business, is to get people to feel something. The next order of business is, well, understanding what those feelings are. And this is the precision problem. Basically, once you have an understanding that someone’s feeling something, then you have to categorize it. Is Jamin expressing love or is he just saying it’s interesting? Are we laughing here or are we angry or upset or bored? I will say though unequivocally that this is not a sentiment problem. And I’m gonna draw a quick distinction because most researchers have seen green and red dials that say positive or negative scores. And positive and negative is an unequivocally bad way to think about language. Effectively, if you’re a really funny character on your sitcom as people are laughing, then that’s really amazing. But if you’re a super-serious political candidate and people are laughing, maybe that’s not so good. And, so, we actually encourage our researchers to think about positive negatives, not in the experiential data sense, but instead in the operational data sense. Is this good or bad for my business, or the KPIs, ultimately, that the CFO or the shareholders care about. And that’s what makes it good or bad. Positive and negative is a conclusion. But, so, once we’ve detected an emotion, the next order of operations is to understand how people feel. And, so, we have this framework of 42 core emotions that is the most nuanced framework we know to exist, but you also have this interoperability with more academic framework, so if you want to leverage Paul Eckman’s six core emotions or Robert Plutchik’s eight-core emotions with the different degrees of intensity, we’re giving the researcher total control over the technology to decide how they want to classify this. And, so, that’s basically the process. We get a piece of text, we’re figuring out, is there an emotion present? Yes or no. And, then, once it’s present, how can we group it together with other emotions that are very similar, and then provide really clear explanations as to what makes this up, what are examples, and why are they feeling that way? What’s effectively driving it? So that’s how we’re quantifying our emotional score, and it becomes, as opposed to a word cloud which will just tell you generally people said the word funny a lot, we’re attempting to summarize the emotion humor and the millions of ways in which people could be expressing that. And that’s what the system is really good at. Now, you asked a really important question about bias, and having bias make its way into algorithms and training set is a huge consideration and important point. One of the things that we’re proud of is how our system’s been trained on public and unsolicited data going back six years. It’s something that makes it unique in its ability to understand Black Twitter and how – Also how politicians speak and how we talk about our favorite programming versus our health concerns and brands. The system’s been trained to cross 26 different categories and has been trained across a very, very wide population of inputs. And it’s through that diversification and just making sure that as we refine the systems, we’re calibrating it to how the population actually looks and the market we’re actually trying to understand. That’s how we ground ourselves in approaching that sort of problem.

Jamin Brazil: It sounds very comprehensive, which is exciting from my vantage point. We’ve seen major companies make big mistakes in the past, not being conscious of the fact that not everybody is a white male or whoever is doing the programming. And, in fact, even recently, one of the top three market research agencies had a global report on the state of COVID as it relates to minority groups. And they incorrectly identified African American as being optimistic through this time, but the framework was not – It was taking into account from a white person’s perspective and not a historical perspective, which blacks historically are seen more – Or voice more optimism, which is really interesting from a cultural perspective, right? I think it’s fascinating how you’re incorporating the diversity in the points of view, especially in the context of our world today.

Jared Feldman: Yeah, absolutely. I mean, you raised the important point that this is a massively complex issue, but it’s just as important as it is complex. And, so, it’s just why semantic analysis generally shouldn’t be a box that’s checked or sort of a side project done by one of these platforms. It takes dedicated years of research and understanding to really get underneath the nuances and just the idiosyncratic details of our language and what it means about us when we express it that way. And that’s also – It’s why I made the distinction that we’re not necessarily measuring emotion. We’re not hooked up to people’s brains, per se, this isn’t neurological or physiological in that way, but it is the expression of emotion. And I just – There’s just so much importance in what people are willing to express. And, also, as long as you’re able to dimensionalize it properly. Who is this coming from? What is the history there? And how do we make sure that when we make statements, when we classify things, how do we make sure that this is representative of what they meant and how they felt?

Jamin Brazil: As you look forward, how do you think market research is gonna be different in the next five years?

Jared Feldman: When I think about what’s happening and the changes across enterprises – And we work mostly with large Fortune 1,000 companies, it becomes obvious that the researcher’s role is changing quite a bit with the introduction of new technology. I think it’s clear that researchers broadly, their role will evolve. And it’s not scary, actually. I think it’s a really exciting opportunity. Researchers will become storytellers for the organization. There’s too much time being wasted by researchers being like the PA on a movie set. And I really like this concept of the researcher becoming the director, where with the technologies that are being introduced at the enterprise, at every level, at every single point of the survey value chain, for example. How data gets cleaned, how surveys get designed, how we think about prepping data, all the work that has to happen before you can even sort of start to think about connecting dots, and what does this mean for the organization? And, then, ultimately delivering it. What is the mechanism by which – Researchers spend hours just, even after they have a story, putting together the PowerPoint or putting together the report that then goes out and then has to be presented. And I think that the entire – All of that friction in the value chain is gonna be systematically alleviated. And it’s gonna free up researchers to do what they love to do, which is storytelling, which is using their domain expertise to connect dots and to craft insights, which is really just the shortest story you can tell between a data point and, ultimately, an action that the organization takes. I mean, this is really fundamentally why researchers love what they do, broadly, is to make evidence-based recommendations or decisions based on their understanding of why people are behaving the way that they are. And I think that there’s a really cool set of technologies – And not just in the analysis space where we are, but just broadly how surveys get designed, how they get distributed, how they get cleaned and prepped, how blockchain will ultimately play a role. How business intelligence tools like Tableau, etc., will play a role in the distribution and, as you noted, the democratization of insights. I think that this swirl of technologies is no longer just a sort of idea, but you’re seeing them be implemented in different ways. And researchers who have spent a long time being the PA on the movie set are gonna start to be the director of the data. And I think that all of these tools at our disposal are gonna be – Are gonna make them heroes. Are gonna make them feel like superheroes and get back to doing things that are powerful, and especially in uncertain, volatile environments, like where we are now. We’re in an extraordinary volatile time in the world, in our culture with the pandemic and other things. And I find that organizations that recognize that begin to really double down on the data. They start to say, “Well, we can’t just throw things at the wall. We don’t have extra budget or excess experiments that we can be doing. But, instead, let’s found all of our assumptions in the data. Let’s really prove out what we think should work and really understand our customers.” And I just see that accelerating because the amount of venture capital and really, really smart people dedicating themselves to the systematic friction points because it’s a really enormous market. There isn’t a single brand on the planet that wouldn’t benefit from being more empathetic, from really better understanding the customer in a systematically less friction-filled way, and being able to connect those dots more quickly. The movies that we make today are just so much more exciting and amazing, there’s so many more tools at your disposal than what was even possible 50 years ago. And, so, it just – It’s a very exciting moment, I think. And I – And especially in periods of volatility or economic distress, it’s my experience that researchers really are able to step to the plate and start to make some really important decisions and empower more empathetic decisions across organizations.

Jamin Brazil: Last question. What is your personal motto?

Jared Feldman: I’ve had a couple of quotes that I think about often. I don’t know what constitutes a motto, per se, but just some – There’s an idea that I try to ground myself in quite a bit. There’s a stoic philosopher, Marcus Aurelius, who wrote meditations, which likely a lot of folks that listen to this are familiar with. But one of his quotes, he says, “Always bear this in mind, that very little indeed is necessary for living a happy life.” “Always bear this in mind that very little indeed is necessary for living a happy life.” And, for me, this is about humility and gratitude and perspective and trying to find moments throughout the day and ways in which to view the world that makes me hopeful and happy. And it’s a – It’s just a thing that keeps me grounded and, generally, is a helpful principle and just kind of motto to adopt that the world is tough and there’s really hard things going on for a lot of people, and to just find ways to be grateful and to remember that every day can be a good day. And that’s something that, on a personal level, helps me deal with the euphoria and also the panic that comes along with running a start-up, especially in a pandemic, and trying to be an empathetic leader and citizen and just general human, that you don’t need that much to live a happy life. And you can find those things if you look for them.

Canvs_Logo_Black_RGB-01

See Canvs in action

Schedule a demo to see how Canvs automates text analysis, saves time, and improves insights from open-ended text!

Canvs_Logo_Black_RGB-01

Schedule a demo

See how Canvs automates text analysis, saves time, and improves insights from open-ended text!

Canvs_Logo_Black_RGB-01

Almost there...

Please complete the information below to access this content.

Canvs_Logo_Black_RGB-01

See Canvs in action

Schedule a demo to see how Canvs automates text analysis, saves time, and improves insights from open-ended text!