Single workout analysis
A practical guide to getting the most out of AI-assisted post-session review — what the AI Assistant analyses, what it needs from you, and how to interpret what it tells you.
This feature is designed to support the coach's eye, not replace it. The AI Assistant works from data and patterns; the coach works from knowledge of the whole person.*
What the analysis produces
Single workout analysis examines one completed session and returns a structured report covering nine areas:
| Section | What it contains |
|---|---|
| Summary | 2–3 sentence overview of how the session went |
| Key takeaway | One sentence the athlete should carry forward from this workout |
| Planned vs actual | Adherence to target distance, duration, and intensity zones |
| Performance insights | 1–5 specific observations about execution quality — pacing, HR behaviour, cardiac drift, interval consistency |
| Effort calibration | How the athlete's perceived effort (their rating) compared to what the data shows |
| Interval quality | For structured sessions: consistency across repeats, fatigue across the set, recovery quality between intervals |
| Weather impact | How temperature and humidity affected performance, if data is available |
| Athlete feedback insights | Observations extracted from the athlete's post-session comments — flags pain, equipment issues, energy level, motivation |
| Recommendations | 1–6 specific, actionable next steps for upcoming sessions |
| Achievements | Personal bests, breakthroughs, or positive patterns worth acknowledging |
Interval quality and weather impact are included only when relevant data is available. Achievements are always present — if no personal bests exist, the AI Assistant acknowledges positive aspects such as consistency or plan adherence.
What the system fetches automatically
Before generating the analysis, the system retrieves the full workout record from Good Coach App. You do not need to copy or paste any data.
Activity data (from GPS device, Strava, or Garmin):
- Distance, active and total duration
- Heart rate (average and max) and power metrics
- Lap-by-lap breakdown — up to 20 laps with pace, HR, and power per lap
- Elevation, cadence, calories
- Cardiac drift and aerobic efficiency (AE) pace, where available
Planned workout structure:
- Target duration or distance, workout goal and description
- Structured interval details if the session was a structured workout
- Sport type and activity category
Athlete feedback:
- Effort rating (1–10 scale) and quality rating (1–5 scale) logged after the session
- Post-session text comments
- Weather: feels-like temperature and humidity
- Coach memory — your accumulated methodology and preferences, which shapes how the AI Assistant frames observations and recommendations
- Athlete memory — the persistent profile of this athlete, built from previous analyses
Your inputs
Workout selection (required): The specific session to analyse.
Coach notes (optional): Free-text guidance that steers the analysis. These are treated as high-priority instructions and override accumulated memory when they conflict. Short notes ("looks fine") do not have meaningful impact — specific ones do.
Include athlete context (optional, off by default): Adds training load (acute and chronic load, load ratio), health state, fatigue level, recent volume trends, and current training phase to the analysis prompt. When on, the AI Assistant is required to reference this context in at least 2–3 of its performance insights, making the analysis situationally aware rather than just session-level.
Response type (optional, full by default): Full returns all nine sections. Short returns only the summary and key takeaway — useful for quick reviews of routine sessions.
Athlete context — when to turn it on
The athlete context flag is off by default because most routine sessions do not need it. When it is off, the AI Assistant analyses the workout on its own merits: how well it was executed, what the data shows, what to take forward. This is faster and cheaper.
Turn it on when the session should be understood in relation to the athlete's current state:
- The athlete is in a high-load week or shows an elevated load ratio
- The athlete reported high fatigue, pain, or illness
- The session follows a hard block, a race, or a long break
- You want the AI Assistant to assess whether the session result is expected given recent training stress
When context is included, the AI Assistant does not just describe the session — it explains it. A slower tempo run reads differently when the system knows the athlete just completed three hard days in a row.
Coach notes — what makes them useful
Coach notes are your primary tool for directing the analysis toward what matters most. The AI Assistant will follow them as mandatory focus areas.
| Less useful | More useful |
|---|---|
| "Check the data" | "The athlete said their legs felt dead — does the HR data back that up?" |
| "Pacing was off" | "Compare pacing in this threshold session to the one from two weeks ago" |
| "Good session" | "Flag anything in the interval splits that suggests early fatigue" |
| "He was tired" | "Athlete had a high-stress week at work — context this appropriately" |
Short or vague notes add little. Specific observations, questions, or flags add a lot.
Memory and continuity
See AI Assistant memory for a full explanation of how both memory types work across all generators.
After every analysis, athlete memory is automatically updated with observations worth retaining: pacing tendencies, effort calibration patterns, how the athlete responds to different workout types, notable athlete comments. The most recent 3–5 sessions are kept as a log; older entries are folded into the broader athlete profile.
This is how the AI Assistant builds a picture of the athlete over time. An athlete with six months of analyses will have a rich profile — the AI Assistant already knows their tendency to go out too hard in tempo runs, their typical cardiac drift range, their injury history — without you restating any of it.
Coach memory is not updated from single workout analysis. It is built from week generation and cycles generation.
What the analysis cannot do
It cannot watch the athlete. If something happened during the session that is not in the data or the athlete's comments, the AI Assistant has no way to know. Use coach notes to surface anything important that the data does not capture.
It cannot compare to sessions it has not seen. The AI Assistant works from what is in memory and what is in the current workout record. If you want it to compare this session to one from three months ago, say so in notes — memory may contain that context, but not always.
It cannot account for context you have not provided. If the athlete's sleep was disrupted, they are fighting a cold, or they ran on an unusually hilly route, tell it. These factors matter and the system cannot infer them from activity data alone.
Structured interval analysis requires structured data. If the session was a structured workout but the activity file contains only a single continuous effort (no laps), interval quality analysis will not be meaningful. This is a data capture issue, not an AI Assistant limitation.