Freemium Done Right: Lessons From a Multi-Billion-Dollar App — Chris Hulls, Life360

Free location sharing is a ~10-times-a-day utility that keeps the app top of mind; paid tier sells safety/peace of mind that people actually use. Core map, location history and place alerts stay free (some limits on free location history). Chosen because carriers (AT&T, Verizon) and later Apple Find My offered competing location products, so the only way to compete was a better product for free. Founders found people said they would pay for emergency safety alone but didn't ('out of sight, out of mind').

Set a somewhat arbitrary rule to ignore monetization until 10 million active users, later than most companies, because location was expected to become a commodity (carriers, later Find My). Era also had no in-app purchase or push notifications.

Internal 'free user Bill of Rights' treated as a constitution only changeable at exec level. Rights include: core map, location history and place alerts must be free (the paywall can move a bit); no deceptive tactics or dark patterns; be transparent and notify users before changes; ads must not get in the way of the job to be done. Motivation: each small cut looks harmless in a holdout test, but 20+ cuts over years erode love and word of mouth in ways data can't show.

Size tests by absolute N, not percent of base: at ~100M users, 1% is ~1M people. For new or costly ideas, run a painted-door test (button that fails gracefully) on ~100 (even 10) users and accept ~70% certainty, then run the real test. Low-risk copy changes (headers, email headlines) can go to everyone to pick up 0.1% effects. Percent steps (1, 5, 10, 50%) are for rollouts, not tests.

About a year before the episode, after ~16 years of subscription-only revenue, added home-screen banner ads for free users using their own ad unit (ad exchange demand, some AdMob testing). Team ran a holdout group to check the ads' effect. Subscribing removes ads.

Direct deal with Uber: when a user lands at an airport the app shows 'Hey David, welcome to SFO. Do you need an Uber?' with one tap. Does not interrupt the workflow. A second similar partner deal was about to be announced.

Previously sold de-identified raw GPS feeds to third parties for location-based ad targeting (filtering sensitive locations). Stopped after repeated bad press and regulator/FTC scrutiny. Now sells aggregated data by census block through Placer.ai (e.g. a retailer sees '72% of visitors came from this census block', no raw GPS lines). Rule: be transparent and the CEO uses any data product himself.

Modeled on Credit Karma's card matching: use known driving behavior to tell top drivers (e.g. top 5%) to share data with insurers and save money, and match hard-to-insure drivers with specialists. An earlier attempt into insurance with Allstate only produced a sliver of product. Could eventually replace subscriptions with free everything.

No ads served to users under a certain age; large ad categories excluded; a team reviews ads; a company-wide Slack channel for reporting bad ads.

Define what can't break (location accuracy, battery drain, life-safety features) versus what can. Killing a user's battery is a one-way door: users remember even after the fix. Suspected regressions are 'guilty till proven innocent' and always investigated. Guest admits the company swung too far into program management process around this.