When Thematic Coding is a Delight: The Impact of Easy, Accurate and Adaptive Text Analytics
Editor's note: This article is an automated speech-to-text transcription, edited lightly for clarity.
Fathom was one of eight organizations to sponsor a session during the January 30, 2025, Quirk’s Virtual Sessions – AI and Innovation series. The session the organization presented was on the impact good text analytics can have on insights.
Tovah Paglaro, Co-Founder, Fathom and Eric Knoben, president, Emerald Research Group teamed up to discuss text analytics effect on open-ends. Paglaro shared a big about Fathom’s platform works and why it is so effective. Then Knoben shared ways in which Emerald Research Group was able to use the platform to better serve their clients.
Session transcript
Joe Rydholm
Hi everybody and welcome to our session, “When Thematic Coding is a Delight: The Impact of Easy, Accurate and Adaptive Text Analytics.”
I’m Quirk’s editor, Joe Rydholm and before we get started let’s quickly go over the ways you can participate in today’s discussion. You can use the chat tab to interact with other attendees during the session and you can use the Q&A tab to submit questions for the presenters during the session and we will answer as many questions as we have time for during the Q&A portion.
Our session today is presented by Fathom. Enjoy the presentation!
Tovah Paglaro
Hi, my name is Tovah Paglaro. I'm one of the founders of Fathom. I'm joined by Eric Knoben, who's a partner with Emerald Research Group.
If you are looking to join a session about what is possible when thematic coding is a delight and to understand the impact of easy, accurate and adaptive text analytics, you are in the right place.
What we're going to do today is start with introductions. I'll introduce Fathom. Eric's going to tell you a bit about Emerald Research Group, and then we'll talk about three tiers of value that Emerald has been able to unlock with really high-quality thematic coding.
So, the first, as you would imagine, is saving time and gaining efficiency for the organization. We'll talk about how that translates into real value.
Then we'll talk about what happens when you start to build trust and confidence in that and scale your use of open-ends across the organization and how that unlocks even more value from open-ends that you're already using.
Finally, that third tier super exciting stuff, and I'm glad we've got Eric here to talk through this when you've accomplished both of those things and are able to start creating new products, new types of insights, layered on top of that really high-quality text analytics the kind of net new value that you can unlock for your stakeholders, clients and partners.
Of course, we'll explore how all of that rests on a foundation of really high-quality text analytics and that can sound a little hand wavy. We'll get right into that and start to define what we mean when we say high-quality text analytics.
Let's get into it.
Fathom is a platform used to understand people through their own words. We use text analytics to make that possible. We have processed more than 30 million open-ended responses for leading brands and agencies. Folks like Emerald Research Group as well as brand side teams and all of that unlocking the value of unstructured data with really high-quality text analytics. That means that teams are able to save time and resources, stop coding and put your time into the high value strategic stuff.
It's possible because Fathom enables nuanced custom code frames. What we mean by that is that they're hierarchical. That they align with your research goals. That they're nuanced and detailed and accurately reflect what's going on in your data.
All that comes together in a delightful analytics interface that allows you to analyze your open-ended data. Not just summarize it but interrogate it and analyze it with really detailed binary coding that enables streamlined reporting.
We know folks want to get to those summary insights and all those kinds of things, and we'll talk about how we get there.
We'll also talk about how all this rests on being a “human-in-the-loop" solution. On yes, of course we are a powerful AI company and also human-centric, human decision, human guidance, human validation every step of the way.
So, that's sort of how we work at Fathom. I'm going to kick it over to Eric to tell you a little bit about Emerald.
Eric Knoben
Yeah, thanks.
We created Emerald to serve tech clients and tech adjacent clients who I think have some really unique needs and challenges.
We have about 25 client facing researchers, but we're backed by Stagwell. They give us a lot of resources to go invest and buy some really cool new toys. As a researcher that's so fun to play in that space.
We're proud of our partner led engagement. We think pound for pound that you're going to talk to and engage with someone who's really smart and knows your product.
I love human focused design because some research design can get so complex you haven't set your customer up to give feedback that you can take action on. It's really thinking through ‘how do we make simple activities to engage in, but get awesome insights out of it.’
Then lastly, I'm all about chasing edges. We don't have any branded gizmos that we're going to push. Everything is bespoke and custom. We look at the unique business problem and what can we do to build something that's appropriate for that business challenge.
So, even taking ideas as basic as open-ends, we've learned there's so much more you can do with them than just reading through open-ends and telling customer stories.
Tovah Paglaro
I love that.
Let's talk about what you can do with them.
And that's really what we're going to be all about today is how using really high-quality text analytics creates value allows you to do more with those open-ends and save time.
When Emerald Research Group started working with Fathom, your team was really after efficiency. You needed to save time. Not have the team code and be able to focus on those more important things.
But what you shared with me is that once you had that in place, you were able to discover that with the kind of nuance detail, accuracy and ease of use that unlocked for you, you were able to unlock so much more value for your clients.
We started to really think about that as three tiers of value that I'm excited to share with you all today.
The first is really increasing efficiency, saving time, saving money and freeing the team to focus on strategy, not coding. I'm saying it in a flippant way, but obviously that's all very valuable and a great starting place with that foundation in place.
Then being able to expand insight. So, higher quality insights, your open-ends, more frequent use of open-ends in your survey instruments, better analysis of open-ends, driving additional value for your stakeholders. So, still working with the same structure of open-ends, but better analysis using bigger sample sizes, a lot of scalability there.
Then finally, with the confidence of both of those things in place, being able to unlock net new value. So, with accurate adaptive and delightful open-ended analysis that leads then to process and product innovation and creativity on the team, which unlocks net new insights for your clients, for your stakeholders, for your brands really connected to business value.
We're going to talk through case studies as we move through the rest of this presentation, bringing each of those to light.
Let's start with increased efficiency.
When we talk about increased efficiency, what we mean specifically, in a quantifiable way, is cutting open-ended analysis time by at least 75% while at least maintaining but often increasing the quality of coding that you're working with.
That's really what you experienced when you started working with us. Tell us a little bit more about that.
Eric Knoben
Yeah, exactly.
Look, when you're picking an AI partner, especially in open-ended coding and unstructured data, I really encourage you to do that side-by-side comparison and run those trials, have your humans code it versus Fathom. What we experienced is that Fathom does it better.
What's really important here is that Fathom is adaptive to our context and goals, meaning it's not just looking at the open-ended question, but we can feed it our business objectives and what we're trying to learn. Of course, you can even feed it a code frame for example, if you have to align to an old tracker, the code frames are nuanced and detailed.
Our tech clients have complex products, and your OE code frame should not be dumb down. It needs to adapt to that complex product. The themes, and especially sub themes, are able to capture in-product experiences that I haven't seen done elsewhere with other OE tools.
Then most importantly, it's easy to use. My entire team can access a self-service dashboard and run their own projects. It's not something where we had to invest hours and stress for how do we train and upskill our workforce to take advantage of this tool.
I'll also add that it's accurate across niche and language. So, not just different geos, but it's just as good at looking at feedback from a cloud architect, for example.
We're really impressed with its ability to handle complex B2B business challenges beyond just coding unaided awareness.
Tovah Paglaro
Love that. When the Emerald team is working with Fathom or any team is working Fathom, once that subscription's in place, ease of use is really important.
Anybody from the team with an active account can upload data in CSV or an SPSS, whether that is for a bespoke use case, brand new research, create the code frame from scratch or tracking data with an established code frame that you're tracking in Fathom, input that strategic guidance, provide a little information about research goal, business goal, any key themes. Then hit go and go work on other things.
You get to go focus on other things and in the background, the AI is doing what it's doing and our team of coding supervisors is working with the AI to ensure that it adheres to that guidance, that it reflects the research needs and the research goals that you uploaded, and also that there's no hallucinations or bias creeping in.
It’s really important to have that human review in place when you're working with an AI tool. What that means is that you kind of get the best of both.
Humans are really good at context, uniquely good at context, and we don't want to take that out of an automated solution.
AI is really good at scale, at accuracy, at detail so you're able to combine those two things and that's how we are able to help you unlock net new insight, brand new insight.
Let's talk about what that looks like. What happens when you start to be able to scale your open-ended insights using more open ends, more surveys, more detailed analysis, all of that at scale and things start to get pretty exciting when we're talking about that.
Obviously, it's about value. It's about increasing the value that you're getting from your open-ends. You're able to do that with more analytics that have nuance, detail and accuracy across them.
When we talk about high quality, think about coding. That's exactly what we mean. Nuance and detail context, adaptive hierarchy and themes and sub-themes and really accurate labels, not just summaries, not jumping ahead to the high-level summary instead really labeling the detail. That's what enables the rigorous analysis that is so valuable for a researcher who wants to dive into that data and analyze it.
You're able to work with accurately quantified themes because that binary coding is in place, really accurate, really quantifiable. You're able to start to do things like comparative analysis, comparing behavioral segments, demographic segments, other data that you might have in place to be able to intersect with your open-ended themes you can track over time.
We've talked about that a little bit, but do that with the kind of nuance that is able to drive business decisions and you're also able to start layering in sentiment and emotive code frames, unique code frames that might be really important to your business outcome, to your research goals. All of that needs to come together obviously into an interface that's easy to use.
You talked a little bit about that, Eric, the importance of that for your team. When it comes together in the interface, some really important things happen. You're able to start to segment that really quantified data by attribute or by closed-ended responses that comes right from your upload. Anything that the Emerald team puts into fathom, it's there for analysis.
You can also create segments by combining custom segments and create new things and really follow that data the way you want to as a researcher. You can also apply statistical significance tests to be able to quantify that variance and make sure that you really understand what's going on in the open-ends. It can get really complicated when you've got big samples and lots of themes. You see a difference.
Does that matter? The platform is going to help you to identify where it's statistically significant, where you should be following that thread. Of course some of the basics, being able to click through themes to the raw verbatims and find those colorful responses that you want to pull into reporting and presentations or even just read for yourself as a researcher to really verify and validate what you're finding.
Eric Knoben
We're going to talk through our journey, and it started with, let's just master the basics. We need to code some OEs. I'll talk about where that went.
We have a tech client who had to take this product meant for adults and bring it to children, and they wanted to invest in some third-party IP collaborations and partnerships, but they didn't really know where to start.
They had a lot of quant data from closed ended-choices that told them basically pick a Marvel hero. Just pick one and go with that whole world. The challenge of that though is they just feel like, what if we're missing key characters that would unlock maybe a much broader audience?
Through our quantitative research speaking to both parents and their children, over 2,000 of them, we were able to say, not only is Spider-Man the one that's driving purchase intent, passion and joy for children, but through our other advanced math work, we were able to see that no, you actually need to pair Spidey with Louie. The combination of those two maximizes the broad appeal of your product.
Think of it as having two skews, right? I have a skew that's for Bluey and a skew that's for Spidey. That combination really unlocked this “aha moment” for our client that their close-ended data.
They've been so focused on just what wins the top, but we were able to show, through open-ended data that, ‘oh gosh, don't forget about this younger audience. They're all about bluey and they have a large passion from their parents to engage with your product.’
Tovah Paglaro
I love that. Such a feel-good story.
Eric Knoben
Yeah. Our next step in our OE journey and coding is that we worked with a B2B SaaS app that had a real challenge where their license uptake had stalled and they were hearing a lot of feedback from their field team and sales force that their product was just too expensive. They were frustrated by research and actually thinking of doing far less research because they were just tired of hearing that feedback over and over.
What we were able to do with Fathom’s sub themes is say all this feedback that you're coding or you're thinking as a price or too expensive challenge, it's actually a value challenge.
Customers were complaining that it's too expensive for the performance. What they really meant was, I have a willingness to pay the price, but I need better performance for that price. Now instead of having frustrating finance or business planning conversations where they can't figure out how to get their cogs low enough, we're actually saying, no, go fix the performance.
This client just felt so empowered and happy with that research outcome; they can take action on that. That's a problem they know how to solve.
Unlocking that really took some detailed sub themes to quantify the amount of people who were really talking about performance value, not how expensive a product is.
Tovah Paglaro
I love that. That's such a great example of how a high-level theme price might not be the level of detail that is necessary to drive business outcomes. It really helps to connect what we're talking about when we say that with the sub themes and the binary coding and the ability to interrogate at that level, not just move to summaries, not just move to high level topics, is what's necessary for you to meet the needs of those really discerning teams that you work with.
We really think about that as meaning that we're able to accurately label every occurrence within the Fathom platform. That you can really interrogate it. That their transparency really provides trust to be able to make those kinds of recommendations, business decisions that are going to drive a lot of value and that have some weight to them.
You need to be able to trust that data to stand on it. You need the transparency. They're able to analyze really rigorously because of the binary coding and with that human-in-the-loop, you're able to bring domain expertise. You talk there about this being a B2B SaaS product, really important to have the domain expertise in place.
Finally then you can streamline the reporting with editable summaries and all of those sorts of things. Once we've got that foundation in place, we go from there to the part that I'm most excited about, which is the way that Emerald has then been able to unlock net new value.
We are talking about, okay, we save time, awesome. Then we were able to scale OE, do more OEs across more use cases and across more questions and in more surveys and analyze them better. Awesome.
Then from there, we're able to start to unlock net new value, do things that our team wasn't even thinking about. We got so excited as your team, Eric, the Emerald team, started to share with us the ways that they were now building new products, net new insights and entirely new kinds of value on top of that really high-quality coding.
Eric Knoben
In hindsight, I'm almost frustrated with myself that it took that long to go, ‘aha, here's the other things we can be doing with this.’
And our first big moment where we took it a step further, we worked with a client in gaming and their CMO had the challenging directive for the team that said, ‘your KPI for the year is to drive brand love and fandom for our brand.’ Of course, the challenge of that is that advertisers can't launch an ad campaign that says, “please love us.”
What is love? That's a challenging emotion to drive. So, we have to look at it from a causal standpoint, what are the other related emotions and experiences that drive the outcome of fandom and love that we're being measured against? What we were able to do with the same set of open-ended feedback from their customers is actually code it both ways.
I thought this was really cool where Fathom adds a lot of value, same set of data, same question.
First, we said we're going to code this based on the emotion being expressed and felt in this open-end. We have that. Now we're also going to code it by the end product experience that drove that feeling.
What was happening when they felt this feeling?
Then through our causal work, we were able to see which emotions lead to brand love and what are the specific in product experiences that create that brand love. What the product team was able to do is say, ‘Hey, we have this feature that we didn't even realize was a hero feature, but it ought to be because people who experience it express far more love and fandom for our brand.’
It gave this team focus, right? They were able to know, ‘Hey, we need to highlight this product and show it in this context the feelings that it evokes.’ And that's how they were able to hit their KPIs for the year.
Tovah Paglaro
I love that. So quantifying emotion, quantifying experience and intersecting the two. Super fun.
Eric Knoben
Yeah. Then speaking of quantifying, I often observe that senior stakeholders or leaders can chase the firetruck. Sometimes they might see something on TikTok that just catches their eye and it's about their product and they think, ‘oh no, we have a big problem.’ And it spurs up all this research. Quite honestly, sometimes chaos, but you don't know if this customer moment is one of one or a one of many, and how big of a problem do we have?
So, we were working with a B2C SaaS app and they had a churn problem, but they wanted to take it a step further and actually match it with their customer data and understand not only what predicts churn, but which experiences, and poor feedback are churn worthy. Because if you have a lot of noise coming at you, not all of it leads to churn. Sometimes it could be a squeaky wheel.
That's okay. What we care about though is a squeaky engine, right? Your car can't stop.
What we are able to do here is look at through open-ended work with Fathom, what are the in-product experiences that predict churn? And when customers give you bad feedback, which feedback actually results in churn down the road? And it really helped the engineering team prioritize what are the things we have to fix?
It wasn't that TikTok that started the whole thing where a person was expressing poor feedback about the product. Certainly, those are stories that we want to fix down the road, but the larger problem was actually in these quiet quitters, and they were able to see their product experiences.
That's what we have to fix because we were able to attach that and predict churn with those experiences.
Tovah Paglaro
I love that. I also just heard a persona segmentation that happened to the quiet quitters and the ability to identify them.
In summary, building exactly from Fathom’s high-quality text analytics unlocks efficiency. Absolutely. Of course, you want that when you start working with a text analytics solution, but also so much more value for Emerald Research and for the other clients that we work with.
All of that, of course sits on having uniquely powerful text analytics, with adaptive code frames, with nuanced and detailed coding with really highly accurate coding that you can trust and that you can interrogate with an interface that's easy to use and that is always human supervised.
You're able to unlock all three tiers of value, the efficiency of saving time, so the team can focus on strategy with better coding, underfoot scaling, the use of open-ends in all of the places that you're curious about why all of those times you felt held back, gosh, we wish we could understand why.
Now we can add an open-end to that survey and we can start to lean into why and be able to quantify that. Then net new value innovating to create new offerings, new kinds of insights that drive value for your stakeholders, for your clients, for the folks who you need to deliver for all of that sitting on that text analytics.
That's the end of our presentation today.
The QR code here will let you connect with Eric if you are looking for an amazing research partner to help drive the kinds of outcomes that he just described with Emerald Research Group. Also with our team at Fathom, if you were looking for a text analytics solution that is going to deliver that kind of quality and unlock that value for you either in your agency or on a brand size team. Either way, we would love to talk with you.
Now we will move over to our Q&A period. Thanks so much.
Eric Knoben
Awesome. Thanks everyone.