How to Improve Exercise Selection Accuracy in Programs
1. Understand why exercise selection can miss the mark 0:00
The Programs feature uses multiple LLM prompts to generate exercise options.
A final prompt tries to choose the most relevant outputs for your intended exercise.
Sometimes the generated result does not match what you wanted, so you need to review and correct it manually.
2. Improve accuracy by improving your exercise library name and AKA database 1:16
The biggest factor in better exercise matching is having a well-built exercise library.
Add the names you actually use, not just the formal exercise name.
Include alternate labels and abbreviations under the "AKAs" so the system can match more freehand entries.
Example:
Instead of "2 Arm Back on Bench Press" add an AKA like "Back on Bench Press" if that is how you refer to the movement.
Add abbreviations in the AKAs such as "DA BOB Press" for double arm back on bench press if that is how you may sometimes refer to the exercise.
This helps for when you paste notes from outside the platform without needing to rewrite them.
A. Use built-in abbreviations and request new ones if needed 2:03
The platform already includes around 100 built-in abbreviations.
If a common abbreviation still does not match correctly, contact the team.
They can add it to the default library so it works for everyone.
3. Improve accuracy by building your exercise database to include more exercises.
Option 1: Build your library gradually if you are starting from scratch 2:13
If your exercise library is incomplete, you can still accept and add exercises as you build programs.
Over time, the system improves as you manually add exercises during workout creation.
Expect this to take a few weeks if you are building the library gradually.
Option 2: Import exercises automatically from YouTube or Vimeo 2:33
If you do not want to add exercises manually, use video-based syncing.
In Org Settings, go to YouTube Channel Sync.
Add either:
a YouTube channel URL, or
the channel ID
If needed, use YouTube’s channel sharing options to copy the channel ID and paste it in.
The system will search for videos that look like exercise demos rather than podcasts.
You can review the results and approve or reject them before they are added to your library.
Vimeo can also be connected.
Option 3: Request a manual import if your exercises live elsewhere 3:28
If your exercises are stored in a Google Sheet or another external source, contact support.
The team can help with a manual import into your library.
4. Contact support when matching still is not working well 3:45
If your exercises are already synced but matching is still inaccurate, reach out to support.
Sometimes the issue is a small coding tweak that the team needs to update.
Support can help keep the matching logic working well as the platform evolves.
5. Use the system to reduce manual programming work 3:59
The overall goal is to reduce manual entry for PTs and coaches.
Better exercise matching means less time spent on computer work.
That frees up more time for:
client communication
data analysis
coaching and relationship-building
The exercise selection accuracy is constantly improving via:
your feedback when it isn't performing as intended
LLM updated models (improves speed, but can sometimes regress in progress and cause new exercise selection issues - please let us know if you do notice these)