On the latest episode of the Change Agents podcast, Alter Agents Chief Strategy Officer Devora Rogers is joined by Julia Berk, Senior Director of Global Insights, Analytics and Research at Smartly, and a longtime collaborator of nearly a decade. Julia shares the Snapchat-era research finding that reshaped how the company talked about its product, how she’s earned research a seat at the corporate table, what it really means to be a “data wrangler” in the age of AI, and why she still can’t choose between quant and semiotics as her favorite method.

What is the insight that changed everything for you as a researcher?

For me, the insight that really changed everything was back at Snapchat, as we were establishing our thought leadership research practice on insights related to Gen Z and Millennials, as well as the Snapchat audience specifically. We did research looking at how people were communicating and what the trends were, knowing that Snapchat was really at the forefront of this new evolution in how people connected with friends and family online.

What we found was that people preferred to communicate with pictures, images and video over words. That was something we sort of instinctually knew at Snapchat, because we could see how everyone was using the platform — but we’d never had the data to back it up. The numbers were astounding, and the preference was near universal. It marked such a shift, because historically people have communicated through the written word to convey their thoughts, feelings and opinions. This was back in 2020, and when you look at social media and communication trends today, it’s all photo, video, rich media first, versus text. That was the insight that gave us the data to back up our strategy and double down on how we talked about our product and our community.

You and Devora also worked together on a research project about friendship, early in the pandemic. What stood out to you about that work?

I loved that project. It touched on so many different aspects of research — including academic interviews to understand how people who study relationships saw friendship evolving at a moment when everyone was apart and in isolation, and how people were choosing to spend their time. We were seeing all sorts of data on the loneliness epidemic that was emerging, and we wanted to understand how technology could be a way to connect people virtually and make them feel together even when they were physically apart.

What we found about how relationships can so meaningfully enrich your life really stuck with me — especially for a technology-based company where people aren’t necessarily together when they’re using the product, but can still communicate in a way that makes them feel together when they’re actually apart. The researchers and professors we talked to showed us the real, scientific benefits you get from relationships with friends, and how you don’t necessarily need to be physically together to experience those benefits.

What are you working on now, and how has your role evolved since then?

My role has evolved in a lot of ways since that work. A lot of what I’m focused on today is using data we have access to — whether that’s our own, from custom research, or from other sources — to find the right insights to help drive strategy and use them in the right context to inform decision making.

The way I see research today, as a function and a critical function for a business, is that you’re almost like another brain for the business. You have access to all sorts of information that, without the right skill set — someone who intimately knows and understands the data and the context in which it was collected — can’t be put to good use. When you have that skill set, you can provide the right context to meaningfully move a business forward, whether that’s designing the strategy for a new brand campaign, putting together content for a compelling blog post narrative, or informing go-to-market or company objectives. It’s about having the right insight, collected and measured in the right way, so the business can be accelerated through the investment in research.

You’ve been very good at getting a seat at the corporate table — something researchers sometimes struggle with. How have you done that?

It really comes from understanding your stakeholders at various levels within an organization — what are their objectives, what are they trying to learn, what are they working toward or shifting from, and how can research support them in making those decisions. It’s having stakeholder buy-in at the right levels and the right time, figuring out how to amplify something that looks really positive, and how to message it appropriately and manage expectations when the research comes back with something less than what you’d hoped to see, before you’re in front of the larger group.

It’s a bit of an art and a science — building those strong relationships so that when you come to a stakeholder with research, or you’re scoping a new project together, you have a really good understanding of what everyone’s looking for. I’m usually running one research project to support many stakeholders, so it’s about hearing from each of them what they need from a project so that everyone ends up with what they’re looking for to do their job. I see research as an enabling function — whether that’s enabling marketing, revenue or strategy — and I think you earn that seat at the table when you come to it from the perspective that we’re all on the same team. We all want to win. How do we use this as an information point to help us make the right decisions to move the business forward?

You’ve described part of your role as being a “data wrangler.” What does that mean to the uninitiated?

I think it means that you can get a lot of data — more today than ever before — but the question becomes: where did it come from? It could be your own source, it could be synthetic or AI-generated. All of those have the right time and place to be used, but you need to know where it comes from, why and how it was collected, and what it really means. Just because you have a metric that tells you one number about the business, how is it calculated? That’s so much of what my day looks like, because we have so much data available to us, but when you dig deeper, you sometimes find you need to understand what series of events or analyses led to the data point you’re working with today.

So that’s what I mean by wrangling data — I’m working within my organization, across the various functions on my team and cross-functionally across the company, to get as many inputs as we can. With AI, we can synthesize data in a way that’s much more scalable and faster than ever before, and I love being able to be efficient — but the output is only valuable if the inputs are high quality. That’s what I spend a lot of my time doing: making sure we have the right inputs so we can move quickly to outputs that actually make a difference. For example, prioritizing which types of accounts or opportunities to look at first in a book of business — if the inputs for that model aren’t there, the outputs won’t actually help you prioritize the way you need to.

It’s become a bit of a cliché to say garbage in, garbage out, but it’s even more pronounced right now in the era of AI. You can get more data than ever, which is great, but unless you put it together in the right way, you end up with garbage — you end up with AI slop. It’s so easy to find “research” anywhere, but is that source reliable? Doing that data interrogation — just because something fits the story you want to tell doesn’t mean it’s the right data point to use — is part of the different level of ethics you have to apply to your work as a researcher. Others will take a data point and run with it, but they’re able to do that because you, as the researcher, have blessed it, because you’ve been diligent and thorough in collecting the information.

Speaking of AI — how are you thinking about it in your research? What’s on the table, and what’s off?

I’m open to everything with AI. I take synthetic audiences with a bit of a grain of salt — I’m just starting to vet solutions that have a synthetic audience component, and it really matters to me who’s building these models, what their inputs are, and how many different models are working together, because you don’t want bias. We’ve already seen that AI has inherent biases, so it’s important to plan how to counteract that in your strategy.

Where I use AI a ton is for synthesis. We’re constantly going through a lot of different reports, or going back to research we’ve run previously — like brand health data — and I want to pull it together for different audiences. I can use AI tools to help point to the high-level themes that are emerging, which I then go and dive into myself in the data. It helps you pick the right direction to go in. If I’m doing qualitative interviews, I’ll use it to synthesize notes from the conversation. These tools help you make sense of large amounts of information quickly and get to the outline of your report faster — but you still need to go back into the data to find the right, or the best, data point, because AI does a great job of summarizing, but I often feel it lacks the nuance or the ability to transform a story with strategy the way a human can.

It really helps that I’ve had a smaller research team at Smartly to date, because I’m able to use AI tools to be more efficient. I talk about having an “AI Julia” who helps me get a lot of my work done, so I can really focus on strategy and application. AI helps me get to where the meaningful information is, faster.

One obstacle I dealt with earlier in my research career — though not so much recently — is that you do a research study, launch it, socialize it, share your findings, and move on. With AI, it’s much easier to go back to research that’s still relevant, even if it wasn’t done in the last six or twelve months. I don’t think research goes stale — it has a good shelf life — but historically, we haven’t done a great job of using and reusing the research we already have. With AI, you can create a real-time library to access it, so rather than thinking you need to go run more research or source something new to support a perspective, you can go into what you already have and use it. It’s more sustainable, and it means that when you do go and run new research, you can really focus on the new learnings instead of repeating something you’d already done simply because you didn’t know where it was.

That’s actually something we’re building right now — an insights “gem” that connects to different research we have across our Google Drive folders. You can ask it a question, and it will only look at those folders. If you ask it something like, “what are marketers’ top challenges with adapting creative for CTV,” it will respond with links to the relevant NotebookLM notebooks — all within the Google suite of AI tools. From there, you can go have a deeper conversation, pull out the data, or quickly build presentations or audio components to make sense of the research. It’s really helping unlock insights at scale for the entire company.

What’s your favorite research method?

My first reaction is that I love quant work — I love a great, strong data point, and I love being creative in thinking about how you segment respondents in a survey to get at a really interesting consumer insight. Something like: people say they like fruit, but when you dig into food and eating habits, you discover fruit eaters are more likely to do certain things, and something different is true of people who prefer vegetables. That’s really fun with quant.

But saying quant feels a little expected, so I’ll add that my other favorite — one I don’t get to do as often as I’d like — is semiotics research. It’s hard to do really well and make actionable, but I had the experience of doing it at Snapchat, looking at emerging trends and using semiotic analysis to help identify trends and the communities where they were happening. With the right semiotician, it can be an amazing way to inspire stakeholders with new insights to inform marketing or go-to-market strategy. It makes the data come alive, followed up with quant, in a way that isn’t as possible without it. I love the different layers of texture you get from images, sounds or quotes, and what they mean — it’s not just what’s happening, but how it feels, in a way you don’t get through an interview or a survey. I think it’s a powerful way to inform what you do next, and it’s something I wish we got to use more. It’s very creative — AI could help aggregate it, but you really need the human element to interpret the findings and point to how to use them to inform the rest of your research questions.

You love quant, but rich qualitative research — like a semiotics study — can be expensive and hard to scale. How do you think about that trade-off?

You usually turn to research because you have a fundamental business question that needs answering, and you need to be able to extrapolate it to an entire customer or consumer base. Talking to a smaller group of people is really nice, but the hard part is knowing whether you talked to the right people. When you can talk to a thousand or ten thousand people, you’re going to get there.

New tools and technology — like mobile research platforms — have made some of that richer qualitative work more accessible than it used to be. How do you think about embracing new tools versus sticking with tried-and-true methods?

Technology is a tool that helps us do more, and rather than run away from it, it’s about embracing the unknown or the new challenge and seeing what else you can get from it — how you can make it a superpower, or how it can give you superpowers, rather than deciding it’s just not for you because another way of doing it is better. Sometimes the other way, or the first way, actually is better — but it’s also important to experiment, so that when you make a decision, it’s really the right decision for the work, because you know all of it.