Case study. Entry-point segmentation for TripleTen: learning to recognize our student
How a simple survey segmented potential students and helped fix the sales process
TL;DR. Project overview
TripleTen is an edtech startup that entered the US market essentially flying blind. The sales funnel was pulling in plenty of leads, but most of them weren’t converting: people were dropping off before finishing the trial, and the sales team was burning time on calls that went nowhere.
The tactical request was to optimize the funnel. But the real problem ran deeper: the company knew almost nothing about the people entering it. So I reframed the task – instead of fixing the funnel, first figure out who’s coming into it. To do that, I designed a short survey at the entry point of the trial period and built a segmentation system for potential students.
Key insight: all kinds of people were expressing interest in the program, but two factors turned out to be the strongest predictors of whether a student would graduate or drop out: their motivation to learn, and whether they had any prior coding experience.
Impact for the company
A simple data-collection tool emerged. It gave us key info on potential students and helped multiple teams at once — sales, marketing, product.
Impact for product
The segmentation inspired new product solutions and gave the product room to grow.
1 · Context and research objective
From funnel problems to the right research question
TripleTen is a small edtech startup that spun out of Yandex Practicum. Its core offering is three coding bootcamps.
The company was entering the US market with very little to go on. Program sales were growing, but the teams had no real sense of how to work with a USA audience, how to compete with established players, or how to build a business in one of the most competitive markets in the world.
The sales funnel ran through a free one-week trial: potential students could explore the platform and the learning content, and at the end of the trial, a sales rep would call them. But the funnel had serious problems all the way through:
There were plenty of sign-ups, but most people never showed up for the first webinar.
A large share of those who did register didn’t make it through the trial week.
Among those who did purchase a program, many dropped off after just two or three sprints.
The sales team was overloaded, spending too much time on calls with unqualified leads.
→ It was clear the funnel needed to be optimized.
Looking at the funnel more closely and talking to the sales team, I realized: we knew almost nothing about the people signing up for the trial. Sales reps were calling people knowing little more than their name, with no way of knowing in advance whether this was a target student or not.
→ I needed to find a way to collect information about potential students at the point of entry and segment them.
The final research objective:
🎯 Build a segmentation of potential students to optimize the sales funnel and to make the sales team more efficient.
2 · Research process
Research approach and methodology
Stage 1. #Desk research
Before settling on a methodology, I walked through the funnel myself, talked to the sales team, and mapped out the full picture, including where students were dropping off. I also gathered input from the sales reps – what information about a student would actually help them sell.
I could see the case for a round of interviews, but what we really needed was a way to collect this data continuously, not occasionally. So I designed and embedded a survey at the very start of the funnel.
Stage 2. #Survey
Based on my initial research, I put together a questionnaire. Drawing on what the sales team needed, I identified three types of data to collect:
Demographic characteristics — age, gender, US state of residence
Learning motivation — education level, coding experience, why they’re looking for additional education, willingness to study online for ten months
Purchase readiness — ability to pay, familiarity with TripleTen, desired start date
I knew that attention span at the very beginning of the funnel would be limited, so the selection of questions demanded careful prioritization.
For instance, in the demographics section, the most important data point was the student’s age. Data from previous cohorts showed that younger students adapted more easily to online learning and tended to have stronger motivation. Gender and state of residence; on the other hand, didn’t meaningfully affect outcomes.
The second major challenge in designing the questionnaire was the wording of the questions themselves. I needed to cut the risk of bias and figure out how to phrase things to get us honest answers.
Take readiness to study: we needed to know whether a potential student was willing to commit up to twenty hours per week. But asking it directly — “Practicum’s full 10-month program is conducted entirely online and requires around 20 hours per week. Do you have enough time to study at this pace?” — would likely prompt people to just say yes, giving us unreliable data.
An open-ended question worked better here: “How much time per week are you willing to dedicate to your studies?”
I kept one question from the demographics category; the rest focused on the other two: target and warmth.
target measured how closely someone matched the target audience profile, based on:
coding experience
motivation for professional retraining
willingness to dedicate regular weekly hours to studying
warmth measured purchase readiness, based on:
how soon the student wanted to start
viable payment options
overall familiarity with the program
I also developed a scoring system to evaluate survey responses. Each question had several answer options with an assigned point value – the closer an answer was to the target student profile, the higher the score. At the end of the survey, the points were automatically tallied across the two parameters, giving each lead a match score against the target profile.
Finally, working with the content and engineering teams, we embedded the survey at the very start of the trial and framed it as a way to get to know the incoming student.
Stage 3. #Segmentation
The survey ran on the platform as a test for a month, and after a successful pilot it became a permanent part of the product: the scoring system proved its worth, and all subsequent leads went through it.
A few months after launch, I had accumulated enough quantitative data. But the team was missing a qualitative lead segmentation. I ran a series of interviews with potential and recently enrolled students. Analyzing the responses, I found that two factors were the strongest predictors of success in the program: the type of motivation a student had, and whether they had any prior coding experience. Drawing on that insight, I built a segmentation compass along two axes — motivation type and coding experience — and identified four segments.
3 · Research results
What was developed
In the sales funnel
The funnel experience for students stayed the same, but now the team had key information about every student coming in. The sales team could focus on qualified leads, and deals started moving faster and more efficiently.
Because the answers were standard, a sales rep could glance at the table and immediately see how close this student was to the target profile, and how to approach the conversation. For example, whether to push on urgency or take a slower, warmer path. Sales reps could also cut the time spent on people who were simply burned out at their current job and hadn’t yet figured out what they actually wanted.
In marketing
We now had a data source showing who was interested in the program and what kind of people were entering the funnel By cross-referencing this data with marketing campaigns, the team could see which channels were bringing in more target students and optimize their effort based on real data, not guesswork.
In product
We now had a segmentation of potential students. It helped the product team build solutions beyond just the target segment and gave the product room to grow.
For instance, the Not warm, target segment inspired a career quiz. These were students who met the target criteria but hesitated about which program to choose. For example, some were torn between Data Science and Data Analytics programs. The quiz helped them find their direction and, along the way, nudged them toward a purchase.
4 · Reflection
Personal learnings
Research isn’t always about uncovering depth. Sometimes it’s about using research tools to tune up the processes. A short survey at the top of the funnel ended up optimizing the sales pipeline, giving marketing a tool to evaluate ad campaigns, and sparking new product ideas.
The data collected at the entry point isn’t valuable in itself. What matters is figuring out upfront how it will be used and what decisions it will enable. The impact was significant. The tool was simple.
But behind that simplicity lay real complexity. As a researcher, I needed to cut the risk of bias and of inaccurate answers, otherwise we’d end up with a segmentation built on shaky data. I had to figure out which questions people tend to answer honestly, and design a scoring system that would correctly capture both target and warmth.
And I had to keep reminding myself: attention at the point of entry is low, and getting someone to fill out a survey is hard. We’re essentially asking them to do something for us almost immediately. That meant thinking not just about the survey’s structure, but about tone of voice and the content framing.
I’m glad that with fairly simple tools, I managed to solve what turned out to be a genuinely complex problem.