Why I left MyFitnessPal after eight years
An essay about a paywall, a sale, and the decision to stop renting tools that hold your data.
I logged into MyFitnessPal for the first time in May 2016, in a hotel room in Pittsburgh, on a Nexus 5X. The trigger was unremarkable. I’d gained weight in graduate school. I knew I should track. A roommate at the time was using MFP and had it open on the kitchen counter. I downloaded it.
The first few weeks were the usual onboarding magic — barcode scanning that felt indistinguishable from sorcery, the satisfying ritual of logging dinner before going to bed, the slow and surprising emergence of a chart that meant something. I lost the weight. I kept the app.
By 2018 the pattern was set. A meal eaten was a meal logged. The barcode scanner was the part of the app I used most: I would scan the items I bought at the store, on the way out of the store, for inventory tracking that doubled as eventual meal planning. I was logging maybe 25 foods a day. Across eight years that’s about 73,000 foods logged.
The 2018 breach should have been my exit. 150 million accounts. Email and password hashes. Standard fare for the era — Have I Been Pwned added it, the recommendation was to rotate the password, the press cycle moved on. I rotated. I kept using the app. I wrote it off as the cost of doing business with a free service backed by a real company. Under Armour was, at the time, that real company.
In 2020, Under Armour announced they were selling MyFitnessPal to Francisco Partners. The deal closed; the rebrand and the tonal shift took a couple of years to propagate; by 2023 the app was visibly different. More aggressive upsell. New “premium” wedges for features that had been free for a decade. A redesign that pushed the diary screen below a meal-recommendations carousel. Notifications I hadn’t asked for.
In April 2024, MyFitnessPal paywalled the barcode scanner.
I want to be specific about what that meant for me. The barcode scanner was not a feature. The barcode scanner was the use case. It was the entire reason I had built the habit in the way I had. Eight years of muscle memory of “scan, confirm, log” was suddenly a paywalled-after-three-scans-per-day flow. The “free tier” still existed, technically. In practice the free tier no longer did the thing I needed.
I didn’t pay. I considered it. I sat with the math: $80/yr, $400 over five years, $800 over the next decade. Not a budget problem. The friction wasn’t the dollars; it was the pattern. The acquisition had aligned the app’s incentives with squeezing more from existing users, and the squeeze was the new normal.
I decided to leave.
Leaving wasn’t graceful. The first thing I tried was filing a GDPR access request. (I’m not in the EU but their export portal works for everyone.) I got back a JSON archive of eight years of meals. The archive was lossy. About 12% of my meals had no calorie value attached — those were entries against foods that had been deleted or reorganised in the database in the intervening years and apparently not preserved in the per-user export. The macros were complete only for the most recent four years. The first four years had calories but not macros.
I filed a follow-up clarifying that I’d asked for all data, not the recent subset. Got back, twenty days later, a slightly more complete archive. Still lossy.
That was the moment I stopped trying to migrate the data faithfully and started thinking of the existing eight years as ablation data. The chart that mattered to me — long-term weight trend — I already had as a CSV from a separate spreadsheet. The food-level history was nostalgia I couldn’t faithfully recover anyway. I let it go.
The replacement stack came together over a weekend. OpenNutriTracker on the phone, set up with the OFF mirror running on the Pi I already had. A small Python script for a few specific use cases (annotating training meals before workouts) that I was always going to need anyway. A WebDAV target on the Pi for ONT’s exports. Borg backups to a $4/mo Hetzner.
The first month felt sterile. Eight years of friction and ritual, replaced with something that hadn’t yet acquired ritual of its own. The barcode scanner felt rougher because OFF didn’t have all the brands I scanned. I added them, on the few I cared about, by photographing the labels and submitting to OFF — which felt like a reversion to a kind of internet labour I’d forgotten existed. I have submitted 47 products to OFF since.
By month three the new stack felt normal. It is, materially, less feature-rich than MFP. I lost the restaurant-menu coverage. I lost the social-features I never used. I lost the way the app remembered “you usually log this for breakfast” after a few days.
What I have, in return: the data is on a disk in my closet. Nothing leaves the device except the OFF lookup itself. The Pi will keep working in 2030 unless I unplug it. I write what’s on this site partly because I want others who hit the same paywall to know that the exit is achievable and the cost-benefit is worth it more often than not.
The 2024 paywall was not the worst thing MyFitnessPal could have done to me. It was, in retrospect, a small adjustment to a financial model that had been building for years. It was the decision-point that finally weighed enough on the right side of the ledger.
Most decisions like this aren’t single dramatic moments. They’re a long accumulation of small ones — the ad personalisation toggle that moved screens, the meal recommendations carousel above the diary, the GDPR access response that was almost-but-not-quite-complete. Eventually the weight is enough.
I would say “you’ll know” but the entire point is that you might not, until the moment is past, and at that point the cost of leaving has compounded with the data you’ve handed over and the habits you’ve built around the tool. The earlier you take seriously what your tracker is doing with your data, the smaller the eventual cost of leaving. That’s the case for paying attention now.
I won’t pretend the FOSS replacement was as polished. It isn’t. I’d be lying. The barcode scanner is a percentage point or two slower. The app icon is uglier. The first-run experience requires looking up a setting in the developer menu. The micronutrient surface is shallower than what I’d had before.
I’d also be lying if I said the difference matters in any way that’s affected my actual goal. I am still tracking what I eat. The tracker has updated my weight chart correctly for 16 months. The data is mine.
That’s the whole thing.
Footnote, June 2025
A reader asked, fairly: “Have you tried Cronometer instead?” Yes. It’s the cleanest commercial choice from a privacy posture standpoint. I considered it. It’s still subscription, still cloud, still phones home. The reason I went FOSS instead of moving to Cronometer is that the next paywall is just a question of whether and when. I’d rather be off the rental clock entirely. See the Cronometer essay for the longer answer.
Footnote, April 2026
If you’re reading this and thinking “fine, but I don’t have a homelab and I won’t get one” — that’s a real position. The honest answer is that the rational closed-source choice in 2026 depends on what you value. See the would-I-pay-for-PlateLens piece for what that calculus looks like. None of the commercial apps are great. Some are less bad than MFP.