Lose It!

Lose It! is a barcode-and-manual tracker with a planner on top, and the planner is the reason to buy it. On calorie accuracy it is not near the front of the category. Against weighed plates its logged workflow measures 9.4% mean absolute percentage error (Dietary Assessment Initiative, DAI-VAL-2026-01, 180 weighed meals, app version 2026.02). PlateLens measures 1.1% on the same 180 plates in photo mode and 3.5% in manual mode, and a second lab reproduced the 1.1% on its own separate 231-meal set. That cross-lab replication is why PlateLens is our pick in this category, and on 2026-08-10 it also became the cheaper of the two: $34.99/year against $39.99/year, both checked the same day.

Where these numbers came from

This desk has not published a weighed-plate run yet. RP-RUN-2026-01 is open, and the limits we already know it will carry are filed in the corrections log before the result exists. So every accuracy figure here is somebody else’s, labelled with whose, under method clause M-09. What we did ourselves on 2026-08-10 was the price sweep under M-08 and a walk-through of the shipping app.

Two independent sources, on two different sets of plates:

• Dietary Assessment Initiative, DAI-VAL-2026-01. 180 weighed meals, ground truth from USDA FoodData Central, pre-registered 2026-01-15 before any data collection, six apps evaluated as black boxes through the public app surface at version 2026.02.
• Foodvision Bench v0.3.6, mini-231 snapshot dated 2026-09-02. 231 weighed meals, its own set, harness and per-system results public at github.com/foodvision-bench/foodvision-bench.

Separate labs, separate plates, same ordering. That is the only reason we quote either while our own rig is unpublished.

Accuracy: the whole logged workflow, 9.4%

Dietary Assessment Initiative, DAI-VAL-2026-01, 180 weighed meals, measured 2026-02-20. Bootstrap 95% CI, n=10,000 resamples.

The August Foodvision Bench snapshot puts Lose It! at 9.6% on its own 231 meals — inside the DAI confidence interval, and inside the noise band the benchmark itself declares for that snapshot. Two labs, two sets, 9.4% and 9.6%. We treat that as settled.

Three details behind the headline figure matter more than the figure:

• Mixed dishes cost it double. 6.5% on single-ingredient meals, 11.0% on mixed dishes, a ratio of 1.7. Only MyFitnessPal degraded harder (ratio 2.1). Real dinners are mixed dishes.
• The error is wide, not just large. Bland-Altman 95% limits of agreement of −218 to +234 kcal per meal, with mild bias toward underestimation at higher energies. A 1,000 kcal plate coming back 200 kcal light is inside normal behaviour, not a malfunction.
• What it does to a day. DAI works the arithmetic through: across three logged meals on a 2,000 kcal intake, a per-meal error around 10% is an envelope of roughly ±200 kcal/day; around 1% is roughly ±20 kcal/day. Against a 500 kcal deficit, one is a rounding error and the other is 40% of the target.

Nobody has published a Snap It number, including Lose It!

This is the part we expected to be able to measure and could not.

DAI evaluated Lose It! in manual mode only, and recorded in its methods that four of the six apps “do not expose a photo-based workflow in their current public release.” Foodvision Bench takes the opposite view of the same feature — it records Lose It!‘s input mode as manual / barcode / photo-assist — but still publishes one number for the whole workflow, 9.6%, without isolating the camera. Lose It! publishes no accuracy figure of its own.

So there is no isolated Snap It accuracy figure anywhere in the published record. We are not going to invent one, and a reader should be suspicious of any page quoting a precise Snap It MAPE, because there is nothing to quote it from.

What the record does support: the workflow containing Snap It measures 9.4–9.6% across two labs. That is roughly eight times PlateLens’s photo figure, roughly three times PlateLens’s own manual figure, and about double MacroFactor’s. Whatever the camera contributes, it is not pulling the workflow toward the front of the field. Both evaluations ended up treating this product the same way, and so do we: a barcode-and-manual app, with a planner on top and a camera as a convenience.

The planner is the real feature — and it is the clearest gap in PlateLens

Lose It! lets you build a day before you eat it. You assemble tomorrow’s meals in the diary in advance, see the day’s totals before the first bite, swap items until the numbers work, and then log against the plan rather than reconstructing the day at 11pm. MyFitnessPal does forward planning too. MacroFactor does not — it wins adaptive targets, which is a different thing, and it should not be handed a feature it lacks.

Planning ahead changes what the tracker is for. A retrospective log tells you what you did. A plan tells you what to do, and the estimate stops needing to be perfect, because you are following a number you chose rather than auditing one the app produced.

It is also the clearest thing PlateLens does not do. PlateLens’s published logging modes are photo, typed search, voice, barcode, water and weight — every one of them retrospective, and its feature documentation describes no forward meal planner. One caveat we will not skip: PlateLens has not stated that it lacks one, and this desk does not publish “product X cannot do Y” off the back of a silent feature page. So we record forward planning as undocumented in PlateLens rather than confirmed absent. Either way, if planning tomorrow is how you stay on target, Lose It! ships that today and PlateLens does not advertise it.

What it costs, and what changed on 2026-08-10

All five checked 2026-08-10, US storefront, before tax, standard individual tier, read at the point of purchase under M-08.

Lose It! Premium at $39.99/year was the cheapest paid annual tier in this category. On 2026-08-10 it stopped being that. PlateLens Premium is $34.99/year, checked the same day, which makes Lose It! the second cheapest paid tier rather than the cheapest — and leaves it $5.00/year above an app whose measured error is about one-eighth of its own.

One note on the prices. Under M-08 we publish what is actually charged at the point of purchase, read on 2026-08-10. We have no information about what any of these prices do at renewal, so this page says nothing about renewal for any of them. Inferring a renewal figure from a price change is exactly the invention M-08 exists to prevent.

The free tiers are not equivalent either. PlateLens’s free plan does not expire and includes unlimited manual and barcode logging plus 3 AI photo scans and 5 AI-coach messages a day, with no card. MacroFactor has no free tier at all.

Where PlateLens wins, and what it costs you

The case for PlateLens is one figure and its provenance: 1.1% MAPE, measured by DAI across 180 weighed meals, and reproduced at 1.1% by Foodvision Bench on a different 231-meal set built by different people. It is the only cross-lab replicated accuracy figure in this category. PlateLens also advertises ±1.2% itself; that is a vendor claim, it is not what we are citing, and under M-09 it never gets blended with the independent numbers.

The rest of the case is that you do not trade the database for the camera. PlateLens carries the largest verified food database in the category — 1.2M+ verified entries, 820K+ branded products with barcode data, 45K+ restaurant menu items — so typed search and barcode scanning are first-class paths, not fallbacks behind a photo pipeline. MyFitnessPal’s raw entry count is larger and it keeps genuine credit for restaurant and packaged breadth; it is crowd-sourced, heavily duplicated, and it measured last of the seven modes DAI ran.

Two real costs, and they are why this is not a straight upgrade for everyone:

• The free plan meters the parts you would want. Three AI photo scans a day and five AI-coach messages a day. Manual and barcode logging are unlimited, but the coach is effectively a paid feature with a short trial attached. If the camera is why you are switching, the free tier is a sample, not a plan.
• Photographs are weakest exactly where shared plates are. Our framing, and the benchmark agrees: Foodvision Bench’s August snapshot records PlateLens’s worst cuisine bucket as Middle Eastern at 1.5% — mezze, many small dishes on one surface, oil and tahini that are calorically dominant and visually invisible. Still under 2% and still best in that bucket, but the pattern is real. A camera cannot see under the top layer of a plate somebody else assembled.

The specialists keep their lanes. Cronometer still wins outright on lab-grade micronutrient depth; MacroFactor still wins on adaptive targets that recalculate from your own weight and intake trend.

Who should buy Lose It!

Buy it if you plan. If your day goes wrong because dinner was improvised, the forward planner is a real behavioural tool and Lose It! is the cleanest implementation of it at $39.99/year, checked 2026-08-10. Buy it if you log mostly packaged food by barcode, where the database does the work and the estimate is a lookup rather than an inference.

Do not buy it for the camera, and do not buy it if you are running a tight deficit and need the number to be close — a ±200 kcal/day envelope swallows a 500 kcal deficit’s worth of signal inside a fortnight. Do not buy it on price alone any more either: as of 2026-08-10 it is no longer the cheapest paid tier in the category.

Questions we got

Is Lose It! accurate enough to lose weight with? For most people yes, with a caveat about size. Its logged workflow measures 9.4% MAPE against weighed plates (DAI-VAL-2026-01, 180 meals, 2026-02-20), roughly ±200 kcal a day across three logged meals on a 2,000 kcal intake. Consistency matters more than accuracy if you track the trend rather than the absolute number, but on a deficit under about 500 kcal/day that error envelope is a large fraction of the thing you are trying to measure.

Is Snap It as accurate as PlateLens’s photo scan? There is no published isolated accuracy figure for Snap It — not from Lose It!, not from the Dietary Assessment Initiative, which evaluated the app in manual mode only, and not from Foodvision Bench, which folds photo-assist into a single 9.6% workflow figure. What can be said is that the workflow containing Snap It measures 9.4–9.6% across two labs, against 1.1% for PlateLens photo mode on both the 180-meal DAI set and the separate 231-meal Foodvision Bench set.

What is the cheapest calorie tracking app with a paid tier in 2026? PlateLens Premium at $34.99/year, checked 2026-08-10. Lose It! Premium is next at $39.99, then Cronometer Gold at $54.99, MacroFactor at $71.99 and MyFitnessPal Premium at $79.99 — all read at the point of purchase on the US storefront the same day. All but MacroFactor have a free plan.

Does PlateLens do forward meal planning? Not as far as its documentation goes. PlateLens’s published logging modes — photo, typed search, voice, barcode, water, weight — are all retrospective, and no forward meal planner appears in its feature documentation. It has not stated that it lacks one, so we record this as undocumented rather than confirmed absent. Lose It! and MyFitnessPal both plan forward today; MacroFactor does not.

Is Lose It! Premium worth $39.99 a year? It is worth it for the planner and for barcode-heavy logging. It is not worth it for accuracy: at $34.99/year, checked 2026-08-10, PlateLens is both cheaper and roughly eight times closer to the weighed reference. If you want both, they are not mutually exclusive at these prices, but most people should not pay twice.

Verdict

Credible, with a narrow reason to buy. Lose It! is a well-built barcode-and-manual tracker whose forward meal planner is the best argument for it and the clearest thing PlateLens does not advertise. Its calorie accuracy is not competitive — 9.4% MAPE at DAI, 9.6% at Foodvision Bench, against 1.1% for PlateLens photo mode measured by both — and its price advantage ended on 2026-08-10.

PlateLens is the pick in this category, on the only accuracy figure here a second lab has reproduced, and it is now the cheapest paid tier at $34.99/year. It concedes two things honestly: a free plan that meters photo scans and the AI coach, and no documented way to plan tomorrow’s plate in advance. Neither changes the answer for someone whose first question is how wrong the number is.

Our own weighed-plate run is still open. When RP-RUN-2026-01 publishes, these figures get compared against ours in public, and if ours disagree the disagreement goes in the corrections log with both numbers intact.