01 · The problem
A large audience. No buyers.
What do you do when you spend months building a course, launch it to hundreds of thousands of followers, and no one buys it?
That was the problem facing Mariana Antaya, an AI and machine-learning educator with 460,000 Instagram followers, after she launched her machine-learning portfolio course for beginners.
Mariana had built a huge audience, but she had very little information about who those people actually were or which of them were likely to buy.
Her existing email list had about 3,400 subscribers and had been largely inactive. That meant she could directly reach less than 1% of her total audience without depending on Instagram.
So when she originally launched the course, she had to make important decisions about the offer and positioning based mostly on assumptions: who the course was for, what problem they wanted solved, and what would convince them to buy.
The goal of our work wasn't simply to sell harder. It was to build a system that could turn her social audience into an owned, measurable audience—and then use that data to make the relaunch smarter.
02 · The approach
A four-stage system for turning attention into revenue.
We built the relaunch around four stages. Each stage had a clear job, and each one produced information that improved the next.
Move interested followers onto an email list while collecting first-party data.
Give new subscribers useful ML education before asking them to buy.
Invite the most interested people to a live masterclass and relaunch the offer.
Connect engagement and purchase data so the next launch starts smarter.
03 · Capture
Give followers a reason to leave the feed.
We used two free educational offers to bring interested followers onto Mariana's email list: a five-day machine-learning email course and a live ML masterclass.
Both educational offers did two jobs at once. They grew Mariana's owned audience and gave followers a useful preview of the material and teaching inside the paid course.
We promoted both free experiences through Instagram Stories. To maximize how many followers we reached and how many entered the funnel, we tested three different story strategies.
Illustrative story flows. Results are from the live campaign.
Adding a narrative before the CTA increased the link click rate by 2.7×. Replacing the link with a reply CTA then expanded reach from roughly 15,000 to roughly 40,000 views; ManyChat automatically sent the registration link to everyone who replied.
After someone subscribed, we asked a few questions about their ML experience, goals, frustrations, and blockers. Every signup became more than an email address. It became information we could use to understand the audience and improve the rest of the launch.
04 · Nurture
Earn attention before asking for the sale.
We did not move new subscribers straight from an Instagram Story into a sales sequence. The first five emails they received were 100% educational. There was no pitch.
Across five days, Mariana walked them through the core process of building a machine-learning model from end to end. The course gave subscribers a real win, introduced her teaching style, and gave them a reason to keep opening her emails.
Only after five days of education did we make the first ask: register for the free live masterclass. The email course did the nurturing. The masterclass would become the conversion event.
higher among people who took the free email course before registering for the masterclass
As you can see, nurturing the audience through the five-day email course had a meaningful impact on sales. People who took the course before registering for the masterclass purchased at 2.2× the rate of those who did not.
05 · Convert
The masterclass became the moment of sale.
By this point, followers had already moved through a series of progressively higher-commitment steps before we asked for the sale.
By the time someone reached the masterclass, they had already received useful content from Mariana, learned the fundamentals of the topic, and shown repeated interest in solving the problem the offer addressed.
Before the masterclass, nurture and reminder emails explained what registrants would learn, why it mattered, and why they should show up live. During the masterclass, Mariana taught first. Only after delivering the lesson did she introduce the paid bundle.
We repositioned the standalone course as the ML Builder Bundle. Instead of selling the course by itself, we packaged it with a model-evaluation scorecard, the Vibe Coding Roadmap, and an ML Starter Kit so the offer felt like a more complete path from learning to building. People who purchased during the live masterclass also received access to a live Q&A with Mariana.
The bundle was offered for $347 during the masterclass and $447 afterward. That gave people who attended live a clear reason to act while keeping the offer available after the event.
After the masterclass, we sent a sales sequence that recapped the lesson, addressed common objections, and continued promoting the bundle until the offer closed seven days later.
The launch took an offer that had previously generated no revenue and proved there was real demand inside Mariana's audience.
purchased during the live masterclass.
06 · Optimize
The launch produced a map of who buys and what to build next.
By the end of the launch, Mariana had grown her active email audience from roughly 3,400 to 8,800 subscribers and collected first-party data on 7,126 people.
We connected what people told us about themselves with their behavior throughout the funnel, from acquisition and email engagement to masterclass attendance and purchase. For the first time, we could compare buyers with non-buyers across the entire launch.
That meant we could answer questions the original launch could not:
- Who was actually in Mariana's audience?
- What did the people who purchased have in common?
- Which audience segments were most likely to buy?
- Which large segments might need a different product or offer?
People interested in finance and business applications converted most strongly.
People who were completely new to machine learning converted most strongly.
People looking for clear, guided structure converted most strongly.
16 of 44 buyers came from this intersection. The segment purchased at 1.21%—roughly 2.6× the audience average.
The data also showed where this offer underperformed. Portfolio-and-career learners were the audience's largest goal segment, but purchased at one of the lowest rates. That does not make them a bad audience. It suggests they may need a different product, price, or promise.
Mariana could now use the data to improve the next launch of this product, position new products for the people already most likely to buy, and create different offers for valuable segments the ML Builder Bundle did not serve as well. The next decision no longer has to begin with an assumption.
07 · Before and after
The change was bigger than the revenue.
460K followers, but little first-party audience data
→7,000+ first-party audience profiles
A cold email list of roughly 3,400 people
→8,800 active email subscribers
A course that had generated no revenue
→$15,668 gross revenue in seven days
Launch decisions based largely on assumptions
→Survey and behavioral data showing what worked
Dependent on Instagram to reach her audience
→An owned audience and reusable launch infrastructure
The launch did more than prove the course could sell.
Mariana finished with an audience she can reach directly, data showing who is most likely to buy, and a repeatable system she can make smarter with every launch.