🖥️How Recommendations are Generated

How Recommendations are Generated

Our recommendation engine uses a blend of data analytics and AI-driven insights to provide personalized advice based on your performance data.

How It Works

1. Data Collection: Once your video has been processed, our system gathers a variety of performance metrics—from speed and cadence in running to jump height and reactive strength index in jumping, and bar speed and power in weightlifting.

2. Benchmark Comparison: These metrics are compared against optimal ranges and historical data from your previous sessions. The system identifies deviations that may indicate areas for improvement.

3. Rule-Based & AI Analysis: Our engine uses predefined rules along with machine learning models to analyze the data. For example:

   • Running: If your ground contact time is longer than ideal or cadence is low, the system will flag these metrics.

   • Jumping: Discrepancies between jump height and flight time or an imbalanced reactive strength index trigger specific feedback.

   • Weightlifting: Lower than expected bar speed or high catch times may prompt recommendations on technique adjustments.

4. Generating the Feedback: Using generative AI, the system then crafts clear, actionable recommendations. These suggestions are displayed alongside your report, providing context and steps for improvement.

5. Presentation: Recommendations are shown in a concise, expandable format. Each recommendation includes a short explanation, supporting data, and often a benchmark or target value to aim for.

What to Expect in the Recommendations Section

• Tailored Advice: Each recommendation is specific to your current performance. For instance, a runner might receive a suggestion to “Increase your cadence by 5 steps per minute to improve efficiency.”

• Visual and Textual Feedback: The feedback appears both as text and, where applicable, within graphical annotations in your reports.

• Continuous Refinement: As you log more sessions, our system continuously learns from your data. This means recommendations evolve over time to remain relevant to your progress and training goals.

Examples:

Example Shoe Recommendations

1. Jordan Retro 6 G White/Khaki

   • Metrics: Moderate cushion (8-12 mm), 6.5 mm heel-to-toe drop.

   • Recommendation: "Great for stability with your 11.0 km/h pace, but consider a lower drop to reduce contact time."

   • Supporting Data: Contact time > 6s suggests inefficiency with current drop.

2. Air Jordan 7 Retro SE Vachetta

   • Metrics: High cushion (420.5 mm avg), low drop (0-4 mm).

   • Recommendation: "Ideal for reducing landing impact and improving flight time consistency."

   • Supporting Data: Flight time variability detected in data.

Enhanced Examples

Example 1: Improving Efficiency

• Data: Contact time averages 9s, stride length 420.5 mm.

• Recommendation: "Switch to a zero-drop shoe (e.g., Jordan Nu Retro 1) to reduce contact time to 3-6s and boost stride length to 430 mm."

• Target: Efficiency increase of 5-10% at 11.0 km/h.

Example 2: Balancing Comfort and Speed

• Data: Cushion avg 6.5 mm, speed 11.0 km/h.

• Recommendation: "Opt for moderate cushioning (8-12 mm) like the Jordan Retro 6 G to maintain comfort without sacrificing pace."

• Target: Consistent flight time across strides.