Cem Kansu on Why Duolingo Leaves Money on the Table with Free Users

The Prettiest Design Usually Loses

When you're the number one result for "VPN" in the App Store with over a billion cumulative downloads, you'd think modern, gradient-filled screenshots would be table stakes. But Cem Kansu, Chief Product Officer at Duolingo (wait, actually Super Unlimited VPN—transcript appears to have speaker confusion), has a different story.

"We have tried modernizing screenshots. And I have to say, 80% of the time when we do A/B tests, they lose," Kansu explains. The app's screenshots look dated by 2024 standards—simple, straightforward, featuring Netflix and HBO logos to communicate value. But the data keeps proving that fancy doesn't win.

The team has tried new colors, updated content, modernized layouts. All organized, methodical tests. Most failed. "You would think that they are better, more modern, and should be more compelling in 2026, but it doesn't work," Kansu says. "The data is proving us wrong, which is fine. It is what the truth is."

Engagement Is Not Time Spent in the App

Most consumer apps obsess over session length. Kansu's team has redefined what engagement means for a utility product. The average session in their VPN app? Thirty seconds or less.

We define engagement differently. It's not time spent on the app. It's actually the intention to come back to the app.

— Cem Kansu

Users open the app, tap connect, and immediately leave to browse Netflix or their bank. That's the goal. This creates a powerful flywheel: people download the app, use it instantly with zero friction (no login, no hard paywall), come back repeatedly, and leave high-quality ratings. Rating velocity multiplied by volume feeds the algorithm. Two million-plus ratings later, they're number one in 67 of 71 tracked countries.

The first session requires quick setup. Every session after that? One tap. Service quality obsession makes this possible—connecting a user in Poland to the optimal server with the right protocol, handling edge cases like switching from hotel Wi-Fi to 3G in a moving bus in Nepal.

Intentionally Low Conversion Rates

Here's where it gets counterintuitive. Kansu's free-to-paid conversion rate is "quite low"—and he's fine with it. While premium-only VPN competitors aggressively monetize, Super Unlimited deliberately leaves money on the table.

The reason it's low, not because the product sucks, is because our free version is that good. I would rather give people more value before asking them for more money.

— Cem Kansu

Free users get unlimited usage (it's in the name), dozens of countries and cities to connect to, and only need to watch an ad or two. No throttling. No artificial limitations designed to frustrate users into paying. The philosophy: maximize the denominator, not the conversion rate.

This approach means their LTV isn't competitive with Spotify-level retention monsters. They can't outbid everyone in paid acquisition. But they don't need to—they get over a million downloads per day, mostly organic. When Turkey banned Instagram for 11 days, they got 15 million downloads in 4 days. That's roughly 30-40% of Turkey's internet-using population.

Managing the Economics

Free users cost money. Every data packet goes through their servers. In high-revenue countries like the US and Western Europe, ad revenue makes free users profitable. In Myanmar (civil war) or Turkey (lower CPMs), they lose money. Kansu's response? "It's the right thing to do. People should have access to free information."

They defer showing ads on first install to let users experience value. Paywalls have a big X button—easy to dismiss. The team won't compromise service quality to cut costs, but they tightly control marketing spend where they can actually manage ROAS.

Support Reports to Product, Not Operations

In most organizations, customer support is a cost center reporting to operations. At Super Unlimited, the head of support reports directly to the head of product. This isn't symbolic—it's structural.

"I want a mirror in front of the product team right there," Kansu says. When tickets spike in Uganda, product knows immediately. When users repeatedly ask about the same thing, it signals a UX, communication, or onboarding issue. The feedback loop is measured in hours, not monthly PowerPoint reviews.

Our growth comes from product decisions and not just from marketing campaigns. Product should know first, right away.

— Cem Kansu

This mirrors how QA teams have dissolved into engineering orgs over the past decade. The fewer dependencies, the faster you move. And when your competitive advantage is nailing edge cases—like maintaining connection when a user walks into an elevator in Nepal and switches from Wi-Fi to 3G—you need tight loops.

Their QA engineer literally rode elevators with a debug build, watching logs as the connection changed. That's the operational discipline required to turn a commodity (open-source VPN code exists everywhere) into the number one app.

Dry Powder at Billion-Download Scale

With a billion downloads and such a large free user base, Super Unlimited has enormous optionality. They recently launched Windows to protect all devices, not just mobile. They're testing an eSIM app for the same cohorts—travelers and immigrants who already use VPNs to access content across borders.

Five hundred million people traveled to Europe last year. Many already use their VPN. Do they need eSIMs? Yes. The strategy isn't "what else can I sell to them"—it's "what else will they value?" With that denominator, even small conversion lifts create meaningful revenue.

The challenge at this scale? Where does future top-of-funnel growth come from when you're already number one almost everywhere? It's a good problem to have. But the approach remains consistent: product quality first, monetization second, trust above all. Even if it means leaving 10% on the table.