Fort's pitch is accuracy: the press repeats the claim that it records strength training "more precisely than any competing product" and recognises "over 50 different exercises". Yet the How it works page lists no exercises and gives no rep or set accuracy figures, and the5krunner's review states plainly that "Fort has published no accuracy data" and has not said which exercises it covers. Fort already publishes validation numbers for sleep (trained on 5,000+ nights, validated on 1,500+) and says its models use 20,000 reps, so the method exists. Only the strength figures, the core of the product, are missing.
First action
Export the current exercise taxonomy and the per-exercise set-detection results from the 20,000-rep dataset into one sheet.
How to measure it
How it works → preorder conversion
- Now
- –
- Target
- +20%
- By
- 60 days
Steps
Export the list
Every exercise the model recognises, with its muscle group and detection status.
Founder & CTO 2 days
Compute accuracy
Set detection and rep-range accuracy by exercise group against video-counted reps, held out from training.
Founder & CTO 1 week
Write the methods note
Lifters, reps, conditions, and known weak spots, in plain language.
Founder & CPO 2 days
Design the sections
Exercise list by muscle using the existing diagrams; one accuracy table.
Founder & CPO 3 days
Publish and link
On How it works, a new FAQ entry and the press strip.
Founder & CEO 1 day
Tell reviewers
Send the table to the outlets that asked for accuracy data.
Founder & CEO 1 day
Update each release
Refresh numbers and the exercise list with each firmware or model update.
Founding Engineer, Software each release
Why it works
Add two sections to /pages/how-it-works and link them from the FAQ. (1) A searchable exercise list: every movement Fort detects, grouped by muscle (the page already has front and back muscle diagrams), each marked detected automatically, detected with a confirm tap, or coming later. (2) An accuracy table in the same style as the sleep claim: set detection and rep-range accuracy by exercise group against video-counted reps, with the number of lifters and reps tested and the conditions, plus a plain note on where it's weaker (for example single-arm movements). Add a short FAQ entry, "How accurate is Fort?", and a dated methods note. Use the same table in press kits and replies to reviewers.
The buyer is paying $309 now for a device that arrives in Q2 2027. A claim they can check is what makes a long preorder feel safe, and an open exercise list answers "will it track my program?" before it turns into an email or a cancellation. It also gives reviewers and AI search assistants something specific to quote. Today they can only repeat "no accuracy data".
Ready to use
CopyAccuracy table layout
Exercise group | Exercises | Set detection | Rep range within ±1 | Reps tested | Lifters Upper push (bench, overhead press, push-up…) | n | __% | __% | n | n Upper pull (row, pull-up, lat pulldown…) | n | __% | __% | n | n Lower (squat, deadlift, lunge…) | n | __% | __% | n | n Single-arm / single-leg | n | __% | __% | n | n Method: video-counted reps, held out from training, collected [month year]. Weak spots: …
Risks and how to handle them
Numbers are lower than the marketing implies
Publish them by exercise group with a clear roadmap; honest numbers beat no numbers with this audience.
Competitors copy the format
Being first to publish is the point; refresh it every release.
Evidence
- Your websiteHow it works: Fort "automatically tracks strength training, using motion sensors to detect sets and estimate rep ranges"; models "draw on thousands of hours of collected data, including 20,000 reps"; the sleep model was trained on over 5,000 nights and validated on over 1,500. The page lists no supported exercises and gives no rep or set accuracy figures; the call to action is "Pre-Order Now ($309)".Source
- Pressthe5krunner (February 19, 2026): Fort claims to recognise "more than 50 exercises" but has not said which; "Fort has published no accuracy data" and has not compared its bar-speed estimates with a barbell-mounted reference; rep counting raises "no obvious reason to be sceptical", while proximity to failure is the hard part; it compares Fort with Whoop, Garmin, Apple Watch, Polar Loop and Bevel, and says the central claim will be settled "by independent testing after it ships".Source
- PressNotebookcheck: by "the manufacturer's marketing", Fort is designed to record strength training "more precisely than any competing product" and recognises "over 50 different exercises"; Fort "will initially only deliver within the USA".Source
- Your websiteFAQ: "The Fort app requires an iPhone with iOS 18 or later. Android will be supported by April 2027." "Batch 2 is available for preorder, with estimated shipping in Q2 2027" (Batch 1 sold out). "Fort starts at $309"; the preorder includes the device, a band, a charging dock and the first year of premium software, then "premium software costs $80 per year". 7-day battery; 17.4 g. No FAQ entries on accuracy, the exercise list, returns, warranty or shipping countries.Source