Carolina Devia Angarita is Head of Product Management at Adorama, a company that serves the photography and video verticals — both by selling equipment and by promoting its community of creators for knowledge sharing. She began her career as a product manager at Gilt.com, an online shopping company. From there, Carolina transitioned to Epicor Software and, later, became a senior product manager of global digital at Ralph Lauren. Before her current position at Adorama, she served in various leadership and product management positions at Barneys New York, Fareportal, Scentbird, and Macy’s.
In our conversation, Carolina talks about the importance of understanding who your customers are and what they’re thinking about when they come to your site or app. She discusses how a robust culture of experimentation helps teams embrace ambiguity. Carolina also shares her perspective that “data alone is useless” and that experimentation, along with following steps of the scientific method, will help to put the right data into context.
Prioritization is key to product management success. When it comes to the customer journey, there is no one-size-fits-all solution. It depends on your customers, how well you know them, what you sell, and what problem you’re solving. Nowadays, businesses compete with experience, especially within our sector where customers can find what we sell elsewhere. The only way to grab the customer’s attention is by portraying that perceived value and giving them the best experience they’ve had so that they not only buy but come back.
We start by understanding who our customers are and what they’re thinking about when they come to us. What do they see in our marketing channels? What do they read? What is their discovery experience when they come to the website, as well as what everything looks like from pre-purchase to post-purchase? Once we understand that in each of our many customer types, we define their pain points.
Of course, there is friction for the objectives that we want them to achieve on the website. How many pages do they have to go through? How many clicks do they have to make? We use heatmaps and value stream diagrams to see what customers are experiencing. Further, abandonment rates help us understand and dissect the journey and break it into small pieces. We can identify certain enhancements that we can make for each piece.
The third step is prioritization. We do this with a clear understanding of our mission. What is the main problem we want to solve? Once we align our enhancements with our mission, we go about identifying impact. Which enhancement is going to have the biggest reach? Which one is going to make customers the happiest so that the business value grows from there?
The last component is understanding of our capacity and in-house expertise. What can we do first, faster, and cheaper? These three variables sound like common sense, but they’re very important when it comes to prioritizing. Usually, this comes in the form of a score. Sometimes, it can be a yes or no, but it depends on the customer experience and how rigorous we want to get. Once we launch improvements, we install feedback loops to understand the customer experience, the enhancements we made, and how to evolve. It’s a never-ending process.
I worked for a large multi-category retailer in the United States. Like many other retailers, we experienced a high growth in traffic during the pandemic. However, we noticed there was a high drop-off, especially in discovery and from people visiting our homepage. They were not moving through the funnel to convert or to spend more time on the site. So, we started digging into qualitative and quantitative data — we used heatmap tools and looked into the abandonment and conversion rates.
As it turns out, most of the site visitors who dropped off were first going to the search bar. We found that searchability within our site wasn’t optimal and that there was an opportunity to refine and improve it. We did some work and optimized the algorithm on relevancy and high-search queries, and worked with engineering to develop better filters and facets for guided search.
We A/B tested our hypotheses from there. We showed 50 percent of our users a search bar that was more inviting. We showed the other 50 percent of users a personalized homepage that we called the “For You” page. We wanted to make sure customers knew it was for them. Since that phrase is used heavily in social media, younger generations know that it’s specifically curated for them. This homepage was a bit more intimate and simpler, and was very much about the customer. We showed them recommendations based on their purchases and activities on the site.
The conversion rate from the second group — the one that saw the personalized home page — was 30 percent higher than the first. In fact, the first group’s conversion rate was minimal — there was almost no impact from improving search. That was a very important learning. When we looked at the data alone, we thought we should focus on ease of search, but actually, users were searching because the homepage was not saying anything. Based on this, we focused on a whole page redesign. We saw a 25 percent increase in conversion from the homepage after that. It was a successful experiment that helped us understand where to focus our efforts.
In one of my roles, we had a large, successful subscription plan based around personalized health and skincare routines for our customers. This program brought in $60 million in incremental revenue in the first year. In the second year, we saw a decline — only 30 percent of new customers became members after the onboarding process.
We wanted to look at the root cause, and data was pointing to our onboarding process. It was lengthy and complex — we had forms and questions but not enough clarity around what would be gained at the end of this painful process. Our marketing channels shared great messages about the subscription, but once coming to the site, customers didn’t find it appealing for some reason.
We understood that the customers for this specific product were generally younger. Their main pain point was not knowing if our beauty solutions would work for them. So, we added a virtual try-on that was available on the user’s phone to the onboarding process. We also added progressive profiling to gather valuable information in small bites. We call this a snacking strategy — we ask them for things, then we delight them with something, and then we ask for something else.
Our new customer experience was lightweight, immersive, and had a personalized assistant who asked questions. This helped the customer feel like they were speaking with someone who could help them, instead of filling out boring forms. The initiative was very successful. We also gamified the experience. We were providing a lot of benefits in the subscription via loyalty points and status. So, instead of giving points at the beginning of the process, we switched to giving the point out gradually as the user provided more information and progressed through the process. The further they advanced through the onboarding process, the better it got for them.
Onboarding grew from 30 to 70 percent because of this strategy, and even better, the people who ended up converting were high-quality customers. We started with a larger set of customers and carefully spoon-fed them information about what we were all about. Our average order value grew from $70 to $98. Therefore, we had a healthier subscription with less churn and greater incremental revenue.
It depends on the problem we’re solving. If we just want to optimize the hierarchy or increase page taxonomy, I always recommend leaning on heatmaps or value stream diagrams, but with the assumption that we did our homework to understand who the customer is. If they’re just exploring versus buying or if they are a brand loyalist, there are specific behaviors we need to inspect.
With all of that aside, heatmaps help identify abandonment points, exit points, or dead clicks. Right off the bat, we can begin optimizing the page to fix those things. It’s important to make sure that calls to action are clear and that the site is categorized simply. A lot of these enhancements become official implementations after experimentation.
My best recommendation is to hypothesize based on who the customer is and have a clear understanding of what “success” looks like in terms of performance. With that, you can define smaller ways to optimize taxonomy because otherwise, this could be a very expensive and lengthy effort. AI is very helpful in removing the uncertainty of an idea or helping customers find things faster. It’s taking away a reliance on regular navigation, browsing categories, and subcategories. I’ve recently found that investing in AI and assistant recommendations makes a really good search strategy.
Looking through a product lens, revenue per visit tends to be important to me because it tells me what it means for the business when we get a visitor, regardless of who they are and what they do.
There are also engagement metrics, which depend on the product we’re launching or analyzing, the company’s goals, the phase of the company, etc. I will look at how much people are enjoying the product and how much time they spend on the site. What are they saying about it? Are they writing reviews? Are they engaging in any way? If the company is more mature and has already established product-market fit, I’m going to look at retention, repeat purchase, and customer time value.
Customer acquisition cost (CAC) is also very important because if customers come back, that lowers the CAC and increases the return value. Of course, metrics like profitability and revenue are very important to understand the overall health of the business. Revenue is the primary thing we talk about when we talk to our board or investors, so it’s always good to have those numbers at hand, but not use them as a sole metric of success.
Data alone is useless, but we lean on data to be able to hypothesize, so all of this is science. It’s like the steps of the scientific method, which is why I love product management. For data to make sense, you have to have an objective. Next, hypothesize about that specific objective and think about who the customer is who is interacting with your website or app. Look at both qualitative and quantitative data. It’s also important to put this data in the context of the overall digital store. Why? Because sometimes we, as product people, decide to downplay the potential of something to enhance something else.
Sometimes, strategically, there are reasons why certain pages or features don’t perform as well as others. There may also be times when we want them to have less visibility because we are highlighting something else. I like to look at the data and the story within the context of everything that’s happening to keep an eye on dependencies. Then, I work with a data analyst to clean the data and categorize it based on what matters for the story. If needed, they’ll define ways to create data models so that I can have a more clear and consistent hypothesis.
Ambiguity is part of many jobs, not just product. I try to help myself and my teams by normalizing it. If we don’t have ambiguity, we’re not taking enough risks — we’re playing too safe and that’s not sustainable. We’re also not learning or stretching in ways that we need to succeed. Experimentation is so important. A good culture of experimentation helps teams embrace ambiguity because they learn to hypothesize, analyze data, and uncover results when they put things in front of customers.
Ambiguity is the difference between us and robots. We’re able to work to figure things out. More junior product managers love to shine in this area. Creating a culture of experimentation is the primary tactic I use to handle ambiguity.
When it comes to customer segmentation, more traditional companies have delegated that job to marketing. I’ve noticed that the companies that work on customer analysis and segmentation together with product tend to have the best customer experience. I’ve had to break down a lot of walls, especially in large companies, to bring my input and share what I know from experience to the segmentation process. Now, with AI, things are changing and moving much faster toward a completely different, less rigid direction.
Behavioral actions, analysis, and what customers think, do, and say are what drive us to segment. It’s based on who customers are psychologically and what matters to them. Segmentation used to be based on demographics. How much did they spend? What’s their gender? That approach is starting to become a thing of the past. Now, we have a number of variables that enable us to cluster customers and have an idea of where to focus before we launch something.
LogRocket identifies friction points in the user experience so you can make informed decisions about product and design changes that must happen to hit your goals.
With LogRocket, you can understand the scope of the issues affecting your product and prioritize the changes that need to be made. LogRocket simplifies workflows by allowing Engineering, Product, UX, and Design teams to work from the same data as you, eliminating any confusion about what needs to be done.
Get your teams on the same page — try LogRocket today.
Want to get sent new PM Leadership Spotlights when they come out?
A product teardown is a systematic dissection of a product to analyze its features, design decisions, user experience, and overall strategy.
AI-generated personas not only save time but also provide a more accurate, scalable, and dynamic representation of your users.
A product sense interview measures a candidate’s ability to understand users deeply, frame problems clearly, and generate creative solutions.
A competency matrix is a structured framework organizations use to evaluate, align, and map skills across different roles.
Learn more →
