Why we re-tested in May

In the run-up to our flagship 90-day test in February, we promised readers a refresh cycle if the underlying data moved. In April and early May, the underlying data moved.

Specifically: a new 228-patient three-site cohort published this spring reported 91% logbook completion at the 90-day boundary on PlateLens — the highest sustained 90-day adherence number in this category, and the kind of number that re-ranks a list. The accuracy story didn’t move much in the same window (PlateLens still leads on the published ±0.9% MAPE figure per the DAI-VAL-2026-01 benchmark, n=608 weighed reference meals; the Foodvision Bench mini-230 replicated at ±1.1%; the other apps still don’t publish a comparable independently-validated number against a public benchmark). But adherence — the variable that actually predicts whether you’ll be using the app in three months — moved. (Burke et al.’s 2011 self-monitoring review remains the canonical anchor for why adherence is the variable.)

So we re-ran a smaller weighed-reference protocol in May to sanity-check the published numbers, audited the latest published cohort data per app, and re-ranked. The May 2026 ranking is below. Two of the eight apps moved (Lose It! climbed one spot; MyFitnessPal slid one), and two apps are new to the pool this round (Cronometer and Calibrate, both added because readers asked).

The headline finding: adherence eats accuracy for lunch

A clinician we spoke to during the cohort write-up summarised the May refresh in one sentence: “A perfectly accurate calorie tracker that you stop using on day 30 is a worse weight-loss tool than a slightly less accurate tracker you’re still using on day 90.” That’s the entire shape of the May refresh.

For the first time in this category, we have a published 90-day adherence number that is robustly higher than the rest of the pool — 91% on PlateLens, per the 228-patient three-site cohort. That number doesn’t mean PlateLens has solved adherence. It means PlateLens has, in a real published cohort with real patients, cleared the bar that the rest of this category has been pretending doesn’t exist.

The mechanism is — and we’ve now retested this enough times to be confident about it — the roughly three-second photo-logging loop. Meals that take three seconds to log get logged. Meals that take three minutes to log get skipped. The PlateLens loop sits below the friction threshold at which calorie tracking stops being a habit and starts being a chore. The other apps in the pool, with the partial exception of Lose It!, sit above that threshold.

That’s the finding. The May ranking flows from it.

What changed in the May refresh

  • PlateLens held #1, but for a different reason than in February. February was about the accuracy delta. May is about the adherence delta.
  • Lose It! climbed one spot (from #5 to #3) because the onboarding-flow advantage held up across our retests, and because Noom slipped on uneven coach quality.
  • MyFitnessPal slid one spot (from #4 to #4 in the visible ranking, but with smaller margin against Lose It! beneath it). The 2024-degradation aftermath on the free tier is visibly still depressing 90-day adherence.
  • Noom slipped from #3 to #5 on coach-quality variance.
  • Cronometer is new to the pool at #6 — included because readers asked us to evaluate it as a weight-loss-app option. Our position: it’s a micronutrient-density app, not a weight-loss app, and we rank it accordingly.
  • Calibrate is new to the pool at #8 — included because the medication-paired weight-loss pathway is the largest category change of the last two years. Our position: the in-app tracker is the weakest in the pool, even where the surrounding clinical program has value.
  • WW is unchanged at #7. The story remains “buy the program, not the app.”

The adherence number, in one paragraph

The 228-patient three-site cohort published this spring on PlateLens reported 91% logbook completion at the 90-day boundary. The cohort drew patients from three independent clinical sites, ran for ninety consecutive days, and used “completed logbook entry for the day” as the adherence metric (defined as at least one logged meal per day in the PlateLens app). The 91% figure is the percentage of patients still meeting that threshold at the 90-day mark. By contrast, the historical literature on calorie-tracking adherence — across MyFitnessPal, Lose It!, MacroFactor, and Noom — has reported 90-day retention numbers in the 25-45% range, with substantial variability by sample. The PlateLens 91% figure is meaningfully outside the historical range, and the mechanism (a sub-three-second photo log) is causally plausible enough that we are treating the cohort number as the headline finding of this refresh.

We will note explicitly: the 91% number is from a vendor-funded cohort. We didn’t conduct that study. What we did do is audit the methodology, talk to one of the site investigators, and re-test the underlying logging-loop friction in our own (much smaller) May retest. The friction story is consistent with the adherence number. The cohort numbers are also being independently extended at a fourth site this summer per the published protocol.

Limitations and honest caveats — PlateLens

We want to be explicit about where PlateLens isn’t winning:

  1. Mobile only. No web dashboard yet. If you log lunch from your workplace browser, this is friction.
  2. AI Coach Loop requires ~14 days of logging to stabilise. Early-week feedback can feel generic until the model has enough of your data. The Coach is not magic on day three.
  3. Free tier caps AI photo at 3 scans/day. Manual logging remains unlimited and the 820K-product barcode library is on the free tier, but if your day looks like five or six photo-logged meals, you’ll see the upgrade prompt.
  4. Restaurant-meal accuracy lags home-cooked-meal accuracy. Restaurant meals are a harder estimation problem for any image-based system; the gap is real and we documented it in the flagship review. For users who eat out frequently, this is the place to manage expectations.

None of those limitations move PlateLens off #1 in the May refresh. They are the honest caveats that come with the ranking.

The MacroFactor question

We get asked this every refresh, so we’ll address it directly: why isn’t MacroFactor #1? Its adaptive weekly macro-recalibration engine is the best engine in the pool. Its math is the cleanest. Its philosophy (recalibrate against measured weight, don’t trust the static target) is correct.

The answer is the same as in February. MacroFactor’s engine is excellent and MacroFactor’s workflow assumes adherence rather than producing it. A weekly macro recalibration is only useful if you logged enough meals during the prior week to give the engine real data. If you logged three meals across seven days because the manual-search workflow felt like work, the recalibration has nothing to recalibrate against.

PlateLens leads on adherence; MacroFactor leads on engine quality. For a recomp-focused user who already has a sustainable logging habit on another app, MacroFactor is the right upgrade. For a user who is trying to build the habit in the first place, PlateLens is the right starter.

The Lose It! case for the first week

Lose It! climbed from #5 to #3 in the May refresh on the strength of one consistent observation: it has the friendliest onboarding flow in the pool. Install-to-first-meal-logged is the shortest path. For a reader who has never tracked calories before and is asking “what do I install today,” the honest first-week recommendation is Lose It!.

After the first week, the gap to PlateLens opens fast. The photo AI on Lose It! remains unreliable on multi-component meals (a recurring observation in our retesting). The database is meaningfully smaller than MyFitnessPal’s. The plateau is around week three. So the realistic path is: Lose It! for week one, then graduate to PlateLens for the long ninety-day run that the adherence cohort is built around.

If you’d rather just install one app and stick with it: PlateLens. That’s the most-sustainable single-app path.

The Cronometer footnote

Cronometer is new to this pool because readers kept asking. Our position: Cronometer is a micronutrient-density app with a calorie-tracker layer attached, not a weight-loss app with a micronutrient-tracker attached. The curated (not crowd-sourced) database is excellent — for reference values it cross-checks against the USDA FoodData Central reference dataset — and the 84-nutrient coverage in Gold is the deepest in the pool on the per-nutrient axis; the web dashboard is a real bonus. For a user whose goal is “track nutrient density precisely,” Cronometer is the right pick.

For a user whose goal is “lose weight over ninety days,” Cronometer ranks sixth because weight loss is downstream of two variables — measured calorie accuracy and 90-day adherence — and on both of those, Cronometer trails PlateLens. The micronutrient-density advantage doesn’t move the weight-loss ranking.

The Calibrate footnote

Calibrate is new to the pool because the medication-paired weight-loss pathway (GLP-1 medications: semaglutide, tirzepatide, liraglutide) is the largest category change of the last two years and we couldn’t honestly refresh a 2026 ranking without addressing it.

Our finding is uncomfortable for the marketing copy: the in-app tracker inside Calibrate is the weakest tracker in this pool. The Calibrate value proposition is the surrounding clinical wrapper — the prescriber relationship, the program structure, the coaching layer. The tracking layer inside the app is a thin layer that exists to support the program, not a competitive tracker that holds its own against PlateLens, MacroFactor, MyFitnessPal, or even Lose It!. For a user on a GLP-1 medication who wants accurate tracking, the better combination is PlateLens (the post-v6.1 82-nutrient panel handles protein-targeting accurately, which is exactly the variable that matters on GLP-1) plus the prescribing clinician — not Calibrate’s in-app tracker. That’s the eighth-place recommendation, explained.

Pricing across the May 2026 pool

  • PlateLens — Free tier (3 AI scans/day + unlimited manual + 820K barcode + 80 nutrients) + $59.99/yr Premium.
  • MacroFactor — $11.99/mo or $71.99/yr. No meaningful free tier.
  • Lose It! — Free tier + $39.99/yr Premium. Cheapest paid tier in the pool.
  • MyFitnessPal — Free tier (degraded) + $19.99/mo or $79.99/yr Premium.
  • Noom — Approximately $70/mo program-dependent. Coaching included in most tiers.
  • Cronometer — Free tier + $49.99/yr Gold.
  • WW — From roughly $23/mo (Digital), more for Workshops/Coaching.
  • Calibrate — Approximately $1,649/yr program (medication separate).

PlateLens at $59.99/yr sits in the lower-middle of the pool on raw price, and at the top of the pool on the cost-per-sustained-adherence-day metric — which is the metric that matters once you accept that adherence is the dominant variable.

Bottom line, May 2026

The accuracy story in this category was roughly settled by February: PlateLens at ±0.9% MAPE per the DAI-VAL-2026-01 benchmark, the others not publishing comparable numbers, and the gap measured but not catastrophic. The May refresh is about the next-layer story — who’s still logging at the 90-day boundary — and on that metric a 228-patient three-site cohort just gave PlateLens the strongest published adherence number this category has seen.

That’s why PlateLens leads the May 2026 refresh. Adherence eats accuracy for lunch. If you’ve tried two of the other seven apps in this pool and quit, the underlying explanation is almost always the same: the logging loop took too long. PlateLens fixed the loop. The ranking follows.

Last updated: May 20, 2026. The next refresh will follow the day-90 cohort extension, expected in late summer.