Free Tier Strategy: When Does Freemium Pay for Itself?

Why a free tier is neither a cost nor an engine

By the time a free tier reaches an agenda, finance is calling it a cost and growth is calling it an engine. Both are holding the wrong number. The cost is usually an allocation that charges the free tier for the paying tier's usage, and the conversion figure is usually paying users over all users.

Neither number settles the question, which is narrower: is the share of free users who start paying large enough to cover the ones who never do? A free tier that fails that test can still be the cheapest way you buy a customer. A free tier that passes it can still take more than a year to pay back.

The one division: the break-even conversion rate

The method reduces the decision to one division and two readings around it. Every input has a usual way of going wrong, and the wrong version is the one already in the deck.

In the worked case used in the Free Tier Strategy Pack, a fictional company has a free tier of 3.8 million monthly active users who cost $2.60 each to serve and earn $0.40, so the free base costs $8.36m a year. Its deck quotes 5.9 percent conversion. The flow, 61,000 free users who started paying last year, is 1.6 percent. One conversion is worth $292, so the break-even rate is 0.75 percent and the cushion is 0.85 points.

The free pays for itself twice over, and a year's free base still takes 1.3 years to pay back, which is the finding finance was actually describing. At the cost per free user finance had been carrying, the break-even rate would have been 1.95 percent and the tier would have been cut.

Count the flow, not the ratio

The best-known free tiers report a stock of payers, not a flow of conversions. Only three of the nine free tiers in the Free Tier Strategy Pack carry a conversion figure of any kind, and none of the three is a flow.

Spotify's founders bet in Stockholm in 2006 that they could give the product away, with ads and some friction, and that a slice of listeners would pay to remove the friction. By late 2024 Spotify had 675 million monthly users and 263 million subscribers. That proportion is the famous 40% conversion rate, and it is a ratio of two stocks: no filing says what share of the free base started paying in any year.

The free tier also took a long time to pay. Spotify's first full year of operating profit came in 2024, with royalties consuming about 70% of revenue and gross margin reaching only 32.2% even in its record fourth quarter. The free tier bankrolled a decade and a half of losses before the pocket it was built for paid.

The New York Times built a metered free tier on March 28, 2011: 20 free articles a month, with readers arriving from search, blogs or social links not counted at all. By December 31, 2025 the Times had about 12.78 million total subscribers and operating profit of $431.6 million. It counts subscribers and never the free readers, so its figures show that the pocket pays without showing at what rate the free converts.

Netflix had a free tier it never designed: more than 100 million households sharing a password, by its own estimate in 2022. Netflix tested paid sharing in Chile, Costa Rica and Peru in 2022, then launched it in the United States and more than 100 markets on May 23, 2023, at $7.99 a month per extra member in the US. Paid net adds were 5.9 million in the second quarter of 2023, and net adds reached a record 13.1 million in the fourth. Netflix does not attribute those additions to former borrowers, so even this conversion is reported as net additions rather than as a flow.

Find the pocket that pays: the other side

The convert is only one of three pockets that can pay for a free tier. In the second, the free product goes to one side of a market so that another side pays, and the conversion rate of the free users is the wrong test.

Google bought Android for a figure estimated at about $50 million and then chose not to sell it. Rather than license it to handset makers for a fee, Google made Android free and open-source and in November 2007 organized 34 companies into the Open Handset Alliance. Its aim was to keep its services at the default center of mobile and protect search, not to be in the operating-system business. No handset maker converts; search pays.

Robinhood launched its trading app in 2015 with free trades as the default. The free trade was the loss leader, and the cash and shares it parked on the platform were the product. In 2024 Robinhood's net interest revenue was roughly $1.1 billion, larger than its transaction-based line, on total net revenues of $2.95 billion.

The other pocket has costs of its own. In December 2020 the SEC found Robinhood customers had received inferior prices that deprived them of $34.1 million in aggregate even after accounting for zero commissions, and Robinhood paid a $65 million civil penalty.

Alibaba used free listings to take a market. When eBay bought EachNet in 2003, which then held 85% of China's consumer-to-consumer market, Alibaba launched Taobao with free listings for sellers against eBay's fees, and with Alipay escrow to supply the trust the market lacked. By 2006 eBay had shut its standalone China operations. The case shows free listings taking a market; it does not show which pocket paid for them.

The layer on top: free users who buy, and free users who do not

The third pocket is a layer sold on top of the free product. It can pay for everything or for nothing, and the difference is whether the free users are the ones buying the layer.

Epic planned to sell Fortnite Battle Royale inside its paid game and, two weeks before launch, released it free instead, worried that charging up front would strangle growth. It went live on September 26, 2017 with a promise: "we will not sell items that give a competitive advantage." The money came from V-Bucks, the currency players buy to spend inside the game. Fortnite is estimated to have earned $3.5 billion in 2023, around 80% of Epic's total, though Epic does not publish consolidated financials.

The layer can be repriced once the base depends on it. In March 2026 Epic reduced V-Bucks purchasing power for the first time: the $8.99 pack that once bought 1,000 V-Bucks now bought 800.

The Epic Games Store is the other end. It launched in December 2018 with a 12% cut for developers against Steam's 30%, and gave away premium games week after week. Between 2019 and 2024 the Epic Games Store grew from 108 million to 295 million registered PC users, while third-party revenue, what players spend on games Epic does not own, moved from roughly $251 million to $255 million. In 2024 alone it gave away 89 games with a retail value of about $2.2 billion.

In November 2023 the store's general manager said in court that it was still not profitable and that growth remained the goal. The free base grew; the layer it was meant to feed did not.

Before you close it, price the free

A free tier that does not pay for itself is not automatically one to close. The lever is often the cost of the free rather than the free, and charging from the first day has a record of its own.

HubSpot capped its free tools at a maximum of five edit seats from March 5, 2024, with unlimited free view-only users, a cap that did not exist under its earlier model. HubSpot's stated aim was to let customers start for free and "pay as they grow." That is a move on the cost and shape of the free tier, not a decision to end it.

Quibi tried the opposite and charged from the start. Quibi launched on April 6, 2020, selling a subscription to short-form video against TikTok and YouTube, which charged nothing, and spent over $1 billion commissioning more than 8,500 episodes. It had projected 7.4 million paying subscribers in its first year. Quibi announced its shutdown on October 21, 2020, and its library went to Roku for "significantly less than $100 million."

Quibi had no free tier to judge. Its case is the other column of the comparison: what a customer costs through the paid channel when a competitor's product is free. The method puts that cost beside the cost of a customer bought through the free.

What these cases show

Each case is drawn from the Stratrix case library and checked against the filings and reporting listed under Sources. The last column says what each one does not establish.

Running the decision

What this method cannot tell you

Free tiers that worked are written about, and the ones that quietly cost money are not. The cases above show how each pocket works. They cannot support a success rate for free tiers, and this page does not claim one.

The arithmetic also stops short in three places. It does not price what a free base is worth beyond conversion, which in some cases is the whole return. It carries a conversion to the average tenure of a paying user, although converts from free may stay a different length of time. And it takes the share who would have paid anyway as a stated assumption, because no company has measured it.

Questions