My name is Raghu, and along with Usha, I’ve built WindowEdu from the ground up. We’re an EdTech venture, and honestly, neither of us had any prior sales, marketing, or even extensive coding experience—just a solid background in mathematics. Our journey is is a testament to how small firms, armed with data analysis and AI, can revolutionize their sales process and achieve remarkable growth, even on a shoestring budget.
The Painful Reality of Traditional Marketing (and Its Analytics)
Like many startups, we initially explored traditional digital marketing. I invested around ₹2,03,400, and while we used analytics tools like Facebook Pixel and Google Analytics to track whatever activity we generated, the results from any associated efforts (like initial paid promotions) were disheartening. Our customer acquisition cost (CAC) was a staggering ₹2,034, while our course fee was just ₹999. After expenses, we were left with a mere ₹500 per student. It quickly became clear that even with sophisticated tracking, that much money in the vast digital landscape did virtually nothing for a small player like us in terms of effective customer acquisition.
We even explored affiliate marketing, hoping for a better return. But agencies offered us only about ₹300 per student for a year—far too little to sustain growth. We realized we needed a different approach, something unconventional and scalable.
The Genesis of Our Data-Driven Funnel: YouTube and AI
Our solution began with YouTube. We decided to build our own channel, creating valuable content for our target audience. A word of caution: don’t become too focused on YouTube if you’re a tiny team. Making a single video can take 2-3 hours. In our early days, we had 65 employees, which led to significant debt. Usha and I learned that lesson the hard way, so now it’s just the two of us building this. Even so, we manage to upload one or two videos every two days.
The beauty of YouTube, especially for a small channel like ours with 12,000 subscribers, is the insights it provides. We attract 4,000 to 5,000 new unique viewers every month. Our unique viewer count has been growing rapidly, by an astounding 15% in the last month alone, and 10% in the last 28 days! This wasn’t always the case; we didn’t have a content plan or post regularly before. But since we committed to consistent posting last month, we’ve seen this incredible acceleration. Our click-through rate (CTR) is a healthy 3-4%, which is far better than the global average for EdTech companies. This strong engagement, coupled with ad revenue (we get about ₹40 per 1,000 views, which is high for EdTech in Kerala, India, and globally), suggests our content quality is high and we’re reaching the right audience. Our AI confirms that we have content quality and are reaching the right audience.
This isn’t guesswork. We’ve built an AI agent using Claude that helps us analyze viewer patterns from YouTube analytics, telling us exactly where the dips and peaks in engagement occur. Claude is excellent for its reasoning capabilities in data analysis, even if its natural language output isn’t the best. For that, we use Gemini, which excels in crafting engaging titles, thumbnails (combined with Canva), and descriptions. So, we leverage both: Claude for deep analysis, Gemini for compelling output. Our content strategy is based on authenticity—we only discuss what’s genuinely in our videos.
Nurturing Leads and Automating Conversions
The next step in our funnel is to convert these engaged YouTube viewers into our Telegram group. This process is largely automated, with direct links in our video descriptions. In the Telegram group, we nurture them, building a strong community. Over time, they start asking about our paid courses and how to join.
Here’s where our custom-built system truly shines. For students to attend our practice exams, they need to download our app. The app then seamlessly pushes messages about our courses to them. We’ve developed our own analytics and webhooks, allowing us to map their behavior in real-time.
When a lead visits our landing page, our webhook( IN PHP- Hostinger specializes in PHP,No server setup needed – Just upload and it works)triggers, storing their names in a Google Sheet. If they visit multiple times, that’s also logged. Our Claude AI then analyzes this data daily, identifying patterns and highly interested leads. Even if they visit a payment page and leave, that’s triggered and stored. This data is critical, and we use paid Gemini(inbuilt) and Claude (manually uploads data in csv ) to analyze it directly within Google Sheets—it’s absolutely necessary for this level of insight.
Based on these behavioral insights, we’ve built an automation sequence. We send personalized messages to their WhatsApp and email. Our emails feature beautiful, modern templates created using Google App Script, and our WhatsApp messages include attractive, tiny landing pages. Google App Script is fantastic for professional email templates, although it does have a limitation of around 100 emails per day compared to the usual 500. For now, it’s enough, but we know we’ll need other services when we scale.
The Power of AI in Reducing CAC and Driving Growth
Currently, our conversion rate from YouTube to sales is around 0.5%. While the global average for EdTech is 3%, we only fully developed this funnel last month. My AI agent predicts we’ll reach that 3% target in just three months, which would be a massive breakthrough for WindowEdu, translating to 120 students per month! I’m more conservative, thinking it’ll be within 6 months to a year, but the potential is clear.
The truly revolutionary aspect is how this system impacts our customer acquisition cost. When we scale to 120 students per month, our CAC will dramatically drop to ₹30 per student, and even lower as our unique viewer numbers continue to increase. I don’t even count the YouTube content creation charges because the revenue we get from YouTube itself is substantial for us. This is a game-changer compared to the ₹2,034 from paid ads! We achieve this by integrating AI (Gemini, Claude, ChatGPT) at almost every step: from content creation to app development, tracking student habits, and even making personalized learning plans for them. Human interaction is still needed, but it’s minimal, highly efficient, and not time-consuming at all.
One challenge is the lack of specific conversion rate data for Kerala, India, so we’re forging our own path here. But our strategy is simple and automated: upload YouTube videos, link to our Telegram channel, encourage app downloads for exams, and then let the automated follow-up system take over.
The initial cost for setting up this entire system was incredibly low, with a monthly recurring expense of only ₹3,500. Everything, from the webhooks to the AI analysis, was built by Usha and me. We had no prior experience in sales, marketing, or complex coding, just a strong foundation in mathematics. If we could build our own hyper-efficient, data-driven sales funnel, then anyone interested in building such a system can do it too. And I’d be happy to help them.