Super Unlimited

A free product good enough to compound downloads and ratings replaced ad spend. The free VPN gets over a million downloads a day, mostly organic, has passed a billion total, and ranks number one for "VPN" in 67 of 71 tracked iOS countries (number two in the rest).

There is no login and no hard paywall; free users watch an ad or two and can connect to dozens of countries, and the app is called "Unlimited" on purpose. Tanuj Chatterjee rejects throttling or removing free features: "Our top of the funnel is our superpower. Don't mess with it."

Sub Club▶ 14:59

For a utility, measure intent to return, not time in app. Users tap connect and leave for Netflix or the browser, so average engagement is about 30 seconds or less per session. Success is whether they come back.

Removing login and heavy paywalls keeps the second visit to one tap. "It's not time spent on the app. It's actually the intention to come back to the app."

Sub Club▶ 5:06

Low free-to-paid conversion is an accepted trade for a huge denominator. Conversion is low because the free version is good, yet the business is profitable and growing. With a base this large, a small lift in conversion is worth a lot.

Smaller premium rivals monetize more aggressively; Tanuj calls his approach a denominator game and credits it partly for the organic share of users. No retention, LTV or conversion figures are given. Scale is also "a blessing and a curse": there are fewer people left to add at the top of the funnel.

Sub Club▶ 13:34

Anyone can build 85% of a VPN; the last 15% wins. Open source code lets anyone ship a basic VPN in about two weeks, so the team competed on edge cases: best server and protocol per country, airlines blocking users as bots, handoffs between Wi-Fi, 3G and 4G.

The QA lead rode an elevator several times with a debug build to read logs when the connection dropped. Service quality was the main thing keeping them from number one, and fixing it was the biggest growth driver.

Sub Club▶ 4:06

Their own and rivals' reviews set the fix list, starting from complaints. From the start the team read its own reviews and fixed what people complained about, then read competitors' reviews to find problems it could solve better. Removing friction came first.

The target was complaints, not missing features. Over four or five years of this, with little PR or branding, the app reached number one.

Sub Club▶ 3:31

Splitting ratings by country exposed fixable regional quality problems. The team investigated markets where ratings fell well below the US and found quality issues it could fix, which helped the climb in rankings.

Sub Club▶ 41:35

Support reports to the head of product so pain surfaces immediately. A spike in service quality tickets from a country like Uganda reaches product at once instead of in a monthly KPI review.

It covers UX, onboarding, education and regional service issues alike. "I want a mirror in front of the product team right there."

Sub Club▶ 25:21

Big algorithm changes ship in small chunks behind feature flags. The server and protocol selection algorithm first launched for five countries behind feature flags rather than waiting months to build it for every country.

Scope can be cut; privacy, trust and service quality cannot. "We will do it in smaller chunks rather than build a big massive Taj Mahal and launch it."

Sub Club▶ 43:45

Marketing is the flexible cost; service quality is never compromised. Every data packet passes through the company's own tunneling infrastructure, now tens of thousands of servers in dozens of countries, with a bare metal data plane for cheaper bandwidth and the control plane in standard cloud.

Cloud capacity absorbs traffic spikes because bare metal takes a week to deliver; at their scale, cloud bandwidth alone would make them "lose their shirts." Serving low-revenue countries sometimes loses money, so marketing spend is what rises or falls.

Sub Club▶ 18:27

Don't optimize infrastructure costs before product market fit. The company grew for years on off-the-shelf cloud and only later built its own bare metal data plane.

Tanuj applies the same logic to AI inference costs, which tend to fall over time. "Don't optimize too early. Find a product market fit, you know, grow the funnel."

Sub Club▶ 37:06

Leaving 10% of revenue on the table protects the trust driving growth. No ad shows on the first launch, and paywalls have a big visible X to close them.

"There's a lot of trust that you destroy with overdoing monetization."

Sub Club▶ 20:18

Ads are a small, deliberately unoptimized revenue stream. With sessions of 30 seconds or less, the app runs interstitials, full-screen ads and native banners without tuning traffic for them. The US and Western Europe are profitable.

In some countries match rates and CPMs are very low, and Tanuj calls that "totally okay."

Sub Club

Grow paid revenue by adding value and products, not tightening the paywall. Paid adds no ads, top tier support, more locations, faster 10G servers and, since late last year, multi-platform use. Windows, Mac, an eSIM app and a planned second phone number app extend the bundle.

Retention is lower than at subscription leaders like Spotify, so one plan covering a family's devices helps. The eSIM targets travelers and immigrants hitting access problems abroad and is being cleaned up before promotion.

Sub Club▶ 13:34

Cross-sell only to users with intent; a broad eSIM ad failed. An early eSIM ad shown to the whole base failed because people without travel plans had no reason to want it.

Not every phone supports eSIM, so targeting has to be deliberate.

Sub Club▶ 20:18

Ask for ratings only after the user got value, and rarely. The app prompts after delivering value and avoids repeat or badly timed asks, building 2 million plus high-quality ratings on its App Store page.

No rating gymnastics; download volume and returning users do the work. "We basically ask for ratings after we've delivered value to the user."

Sub Club▶ 5:49

Modernized screenshots lost about 80% of A/B tests to familiar ones. About 80% of redesigned screenshot tests lost and the team often reverted to the originals; the other 20% won only slightly, with no meaningful difference overall.

As the number one app, they avoid big simultaneous changes and test colors and content methodically. Tanuj: "the data is proving us wrong, which is fine."

Sub Club▶ 7:16

Serve ban-driven spikes fully; the base settles at a higher normal. When Turkey banned Instagram for 11 days, the app got 15 million downloads in 4 days. After a spike, the user base settles at a new, higher level.

Spikes also come from poorer countries like Myanmar, and the company serves them even where it loses money, keeping cloud capacity ready to absorb surges.

Sub Club▶ 40:28

SEO stays slow and secondary; in-app marketing drives Windows adoption. Ahead of the late-last-year Windows launch, the team started programmatic content through Daydream, but SEO conversion is weak and web traffic is not among the top five channels.

Gemini takes the top of results pages, and premium-only rivals with years of desktop SEO still win search. The main lever for getting mobile users onto Windows is product marketing inside the mobile app.

Sub Club▶ 10:00

Paid search runs only on chosen keywords and markets, capped by ROAS. Apple and Google search ads target selected keywords and markets, with ROAS watched closely and a spending ceiling, while managing cannibalization of organic installs.

On Android the app is number one only in the US, so extra marketing push goes where it is not yet first.

Sub Club▶ 24:00