How We Grew an EdTech Startup’s Profit 25x in One Year with Analytics

Hello everyone! My name is Nick Valiotti, and I am the CEO of Valiotti Analytics. Here’s a story of how we implemented analytics for an EdTech startup, Refocus, which grew 25 times and reached the third round of investments while working with us.

We worked with them from the very start, building their data operations from scratch: digitizing marketing, sales, and product branches. I will describe how we did this and what we came to. First, we’ll get into challenges, then talk solutions, and finally I’ll brag (just a little) about the results.

Refocus is a prime example of quality analytics and its results

• We started out with nothing, from scratch. It’s always better to implement analytics at the beginning rather than when a company is already making $100M profit. This way, it’s cheaper, simpler, and allows for making the right decisions early on.
• We grew quickly over two years, and analytics grew along. This is an example of how analytics grows with the business and adapts to it, visibly impacting the company’s growth and profits.
• EdTech has its own specifics, but general principles are similar for everyone. Marketing, sales, product, cohort analysis — these are directions almost everyone wants to keep their eye on.

Initially, there was lots of data with little benefit, and it was unclear what to do with it

But Refocus’s data situation was quite common — most our clients come with something like this.

• Numerous data sources: ad accounts, CRM, LMS, payments, refunds, student contacts.
• Everything was stored in Google Sheets: each direction/department had its own sheets and metrics calculated independently.
• Tool problem: Refocus used Roistat to analyze advertising and promotion costs.

Business Impact:

• Many sources + Google Sheets = endless man-hours of manual labor. Refocus had to export data, load it into a sheet, match with another sheet, update… and then start all over again. A workflow like this takes a lot of time and still ends up being unreliable. Errors occur exactly when marketers or salespeople have to handle data manually. Oftentimes it’s for lack of focus, and sometimes it’s masculine feminine employee urge to fudge the numbers.

Not to mention they shouldn’t’ve wasted time on this at all. Salespeople should sell, and marketers should set up promotions, not sit over sheets.

• Distrust in data: It’s hard to ensure data accuracy when everything is held together by good faith. Roistat added to the challenge — the team doubted its accuracy. Hence, much data work brought little value — so how are you supposed to use data if you’re unsure of it?
• Table chaos: Each department had its own analytics, with data scattered across different folders and accounts. Bringing it all together to see the big picture was challenging.

The system depended on specific individuals who managed the analytics in their departments and knew where everything was. Not to be dramatic, but if one of these people left, no one would be able to navigate the ensuing chaos.

• Lack of understanding of how to use data. Analytics isn’t just about skimming the surface to see profit growth or plan execution. It provides detailed information about everything happening within the company, helps identify opportunities, and avoids risks. However, sometimes people simply don’t know how to achieve this: what should be done with all this data to extract insights?

So we faced quite a few challenges.

Task 1: Stop wasting time on sheets and start living

To achieve this, data must be collected automatically, at a set frequency, and in a specific format. For example, if you are advertising with Google Ads, you need to pull data daily: impressions, clicks, and their costs. There are many sources: ad platforms, CRMs, payment systems. The person handling this spends most of their work time on routine data extraction, which can lead to mistakes and burnout.

We automated all of this.

• What we proposed: Apache Airflow. This is an orchestrator — a software that manages data flows.
• Also suitable for: Almost everyone. Airflow is convenient, free, and available [almost] everywhere.
• Result: What Refocus employees would have done manually and with errors, Airflow did with almost no human intervention — quickly and without mistakes. It collected the data, standardized it, and stored it in the database.

A common fear is that the system will fail, and we won’t even know we’re getting incorrect data. Failures do happen, but the system monitors them and notifies a our employee. They will fix the problem before it shows up on the charts.

This is a significant advantage of custom analytics over off-the-shelf solutions: we build it ourselves and fully control it. You can always check if everything inside is working correctly.

Task 2: Move on from sheets to a warehouse to create a transparent, centralized data system

Imagine you need to contact Data Science course students from the Philippines who have requested a refund but haven’t received the money yet. The students’ names and contacts are stored in one table, and the refund requests in another.

When data is stored in Google Sheets, such a task turns into a quest of searching for links and access permissions, followed by googling the necessary formulas. When data is housed in an analytical warehouse, the problem is solved with a single query.

Someone might say, “It’s not my problem, let the subordinates figure it out.” However, this is yet another example of wasteful manual work that consumes your employees’ time and leads to errors in data.

A good system organizes processes to help employees, not create additional challenges.

Google Sheets aren’t meant for storing large volumes of data. When there are too many rows, they simply start lagging. Additionally, data isn’t well protected — it can be accidentally deleted, mistakenly edited, broken, and lost.

These problems are exactly what an analytical warehouse solves. Essentially, it’s the same tables, but organized into a centralized system where data from various sources is gathered. Everything is in one place and won’t get lost if the employee responsible for the spreadsheets leaves.

• Refocus suggested Google BigQuery for this purpose. It’s a cloud-based warehouse, meaning the data resides on the vendor’s servers, and they handle all technical support. It’s a great option for those without their own server.
• Who else would benefit: Anyone, but with certain geographical limitations. Check out their website to see if it’s available in your country.
• The result: instead of a heap of tables, a unified storage solution for the entire company was integrated into the analytics system. Nothing will be lost again, and there won’t be situations where someone accidentally deletes a column or doesn’t know who has access to a table.

Task 3: Get profit from data

To achieve this, we display them on dashboards.

The essence of a dashboard is to present data in a way that helps the client make decisions and grow their business.

Here’s how the sales overview dashboard looked. “Overview” means it provides a comprehensive view: profit, sales, average order value, lead sources. You can hover over each graph to see details or filter by dates. This tool is for managers to track the direction’s dynamics.

But this is general information. You can delve deeper into the adjacent dashboard and examine details thoroughly.

Here you can see the sales managers’ workload and its impact on conversion rates. Using this data, Refocus calculated the optimal number of leads per salesperson to ensure they remain productive without burning out.

They had multiple sales departments across different offices. The dashboard shows the achievements of these departments and their employees: you can see who’s at the top and whose metrics have declined. It also displays trends over time, so you can immediately understand whether a salesperson has consistently underperformed or if it’s a recent change.

Here’s how it looked for marketing. We tracked where leads were coming from and their costs. It immediately showed which channels were performing best and where it was worth investing money.

In a separate section, we checked if the data matched with “Roistat”. In the end, we completely stopped using it and transitioned fully to our own dashboards.

The dashboards allowed us to trace the customer journey from first contact to purchase and accurately calculate the cost per lead.

The student dashboard helped us monitor the number of active students, identify the most popular courses, and review refund statistics.

A real case: a sharp increase in refund requests helped us understand there were issues with a specific course. We quickly made adjustments, and the number of refund requests immediately decreased.

• Refocus suggested Tableau. It’s an industry standard and one of the best services for data visualization.
• Who else would benefit: Again, there are some geographical limitations. In case it’s not available for you, there’re decent alternatives.
• The result: Data that aids decision-making, budget allocation, product improvement, and business development.

What’s with scalability?

Refocus is a great case because the company grew rapidly, and analytics scaled with it.

We started with one source — pulling data from Amo into 2 dashboards. By the end of the project, we integrated 13 sources into the system, and the number of dashboards grew to 42. And it wasn’t just about creating dashboards — we also refined and improved them to make them more user-friendly and understandable.

We worked on this project for two years. Initially, we built the system and then continued to support and enhance it. On average, it takes 3–4 months from the start of the project to see the first results, depending on the scope and complexity.

Here’s what the timeline of the project looked like. We’ve marked only the major milestones — detailing every aspect would be too extensive for any single chart.

Comment below what else you’d like to know about analytics in general or the Refocus case in particular. I’d be happy to answer.