Ep. 520 – IA Annual 2022 Conference Series: Erin Kelly, President and CEO of Advanced Symbolics Inc.

Find Erin Online:

Find Jamin Online:

Find Us Online: 

Music: 

This Episode is Sponsored by:

The Michigan State University’s Master of Science in Marketing Research Program delivers the #1 ranked insights and analytics graduate degree in three formats: 

  • Full-time on campus 
  • Full-time online 
  • Part-time online

NEW FOR 2022: 

If you can’t commit to their full degree program, simply begin with one of their 3-course certificates: Insights Design or Insights Analysis. 

In addition to the certification, all the courses you complete will build toward your graduation.

If you are looking to achieve your full potential, check out MSMU’s programs at: broad.msu.edu/marketing


HubUX is a research operation platform for private panel management, qualitative automation including video audition questions, and surveys. 

For a limited time, user seats are free. If you’d like to learn more or create your own account, visit hubux.com


[00:00:00]

Jamin Brazil: Hey everybody. I am live at the Insights Association’s national event. We’re on the exhibit floor. I’m sitting right now with Erin Kelly. She’s the CEO of Advanced Symbolics. Erin, how are you?

[00:00:15]

Erin Kelly: I’m great. Thank you.

[00:00:16]

Jamin Brazil: Pretty excited to be back in person for me. But I understand you’re an old hat at being in person.

[00:00:20]

Erin Kelly: I’m an old hat. I was ready to be back in person a while back. So I actually, January 2nd at the height of Omicron, I went to Vegas, the CES show 10,000 people. We had a booth there. I went to the Cirque du Soleil. I went to restaurants and then five days later I went to New York, rode the subway. So yeah and came back unscathed from both events. And so I’m looking at, this is like three is a charm. But I am very happy that we are having an in-person event. I’m definitely Zoomed out. I can’t stand the idea of doing another virtual conference.

[00:00:52]

Jamin Brazil: 100 percent no more virtual please. Over virtual. Tell me about Advanced Symbolics?

[00:00:57]

Erin Kelly: Sure. So our big innovation if you will is we have found a way to create randomized controlled samples, the gold standard in market research, the randomized controlled sample on social media. So this took ten years to develop. It gives you a full randomized controlled sample of any platform. So Twitter, Facebook, Reddit what have you. We’re platform agnostic. Randomized controlled sample the population. Huge sample sizes like 300,000 people perfectly balanced to match the stationary distribution of the population you’re trying to measure. And it gives you full audience insight. So if we’re looking at the trucker protests in Ottawa, we can tell you what the truckers thought, what moms with young kids thought, what people who lived downtown thought, what people who lived in the suburbs thought. What the police thought, what the politicians thought. So that’s the big innovation. And because we have these huge sample sizes, we can also do really good predictive analytics. So for example in the last Canadian election in 2019, we’re a Canadian company. We got the seat count down to two seats. So the, we said the liberals would win 155 seats and they won 157 when no other pollster could call it. And in fact a documentary film crew followed us during that election. And Polly is the name of our AI, was by far and away the most accurate pollster.

[00:02:21]

Jamin Brazil: Absolutely amazing. You started the business?

[00:02:24]

Erin Kelly: I co-founded it with my partner Kenton White.

[00:02:27]

Jamin Brazil: Very cool. What was your background?

[00:02:30]

Erin Kelly: My background is a business background. I’m an accountant and an MBA. Kenton is a physicist. And he was an adjunct professor at the University of Ottawa at the time. And he developed this when he was at the University of Ottawa. So it was developed at the University for University Grade Research, and then we moved the IP and everything over to this company.

[00:02:49]

Jamin Brazil: Very cool. I totally agree with the thesis that it’s factual. You’ve got a tremendous portion of the population that’s on social media, and the rest of the population that isn’t, they’re not picking up the phones so you can’t, it’s really hard to reach that other non-digital segment nowadays. And I would argue maybe, not to demean it, but the truth of the matter is the importance of that segment and from a buying capacity etcetera continues to go down. So when you think about, I’m really curious when you think about the, how you balance the sample in a social framework. Because you have population differences by TikTok which skews younger, 100 million monthly active users in US alone. Versus Instagram and versus Facebook which is now skewing older from a usage perspective. So how do you handle the balance, sample balancing in those different platforms?

[00:03:47]

Erin Kelly: So that’s exactly why you have to sample properly. You don’t wanna be doing keyword searches etcetera because then you will be overrepresented in young people depending on the platform you’re using. So the way we do it, our AI is able to, she uses an algorithm called conditional independence coupling. So she starts, let’s say we can do geographically specific samples as well. So let’s say we wanna measure the population in Philadelphia. Well we wanna start with somebody in Philadelphia because people in Philadelphia will tend to hang out with other people in Philadelphia. So we start with say Frank in Philadelphia and we move out from there. Now we don’t want Frank and all his friends in our sample. So if you’re a first network connection to Frank, if we’re taking Frank, then we don’t wanna take his friends and family. We go one or two degrees out. Somebody who still lives in Philadelphia but doesn’t know Frank, is not tied to him, and has a different demographic profile because we don’t want all black people in the sample. So we make sure that we get that variation and age and everything. So if we’re, so the cool thing about if you’re to do this say using a Markov chain algorithm, it would take you a year to develop that sample.

[00:05:01]

Jamin Brazil: It’d be very hard?

[00:05:02]

Erin Kelly: Yeah. But the cool thing about conditional and pens coupling is you can make sure you’re not double counting. You’re not getting all people who know each other, and you can do it a brand new fresh sample in 30 days.

[00:05:13]

Jamin Brazil: Wow.

[00:05:14]

Erin Kelly: So it goes out and it figures out I’ve got enough people in their 20s in this sample. This person is 65, I’ll take him into several –

[00:05:22]

Jamin Brazil: Brand new people too, not professional participants?

[00:05:23]

Erin Kelly: Yes. That’s right. And it’s randomly selected.

[00:05:25]

Jamin Brazil: Wow. Amazing. Erin Kelly, thank you very much for joining me on the podcast today.

[00:05:30]

Erin Kelly: Thanks Jamin.

[00:05:31]

Jamin Brazil: If somebody wants to get in contact with you, how would they do that?

[00:05:33]

Erin Kelly: Info@advancedsymbolics.com.

[00:05:36]

Jamin Brazil: Got it. And of course that information will be in the show notes. Be sure to check it out, screen capture, share it. I will send you a t-shirt. Have a great rest of your day.