What it does: Predicts short-term streaming performance (two weeks post-release) on Spotify for released tracks. Why use it: K5 helps optimize early-stage marketing campaigns and identify potential breakout tracks shortly after release. It provides actionable insights to:
  • Refine Marketing Strategies: Adjust campaign targeting, budget allocation, and creative based on predicted performance trends.
  • Capitalize on Momentum: Identify tracks exceeding expectations and double down on promotional efforts to amplify their reach.
  • Course Correct Underperforming Tracks: Detect underperforming tracks early and implement corrective actions to improve their trajectory.
  • Measure Campaign Effectiveness: Compare actual streaming performance against predictions to assess the impact of your marketing activities.
  • Inform Future Release Strategies: Learn from short-term performance patterns to improve future release planning and campaign execution.
When to use it: Use K5 immediately after a track’s release, with at least three days of streaming data available. How to use it in Songbird:
  1. Upload Audio: Upload the track’s audio file (MP3, WAV, or M4A). The filename must be the Spotify Track ID. (Include a clearly annotated screenshot of the upload area in Songbird).
  2. Input Metadata: Provide the track’s release date (YYYY-MM-DD), artist ID, and the last three days of streaming data from Spotify. Also, specify if the track is a focus track (Yes/No), any marketing alerts or initiatives, and your streaming goal. (Include a screenshot of the metadata input fields with labels).
  3. Run K5 Analysis: Click “Run Analysis” to generate predictions. (Include a screenshot highlighting the button).
  4. Interpret Results: Review the predictions, visualizations, and causal recommendations (if applicable) in Songbird. (Include a screenshot of the results page).
Key Inputs:
  • Audio File (MP3, WAV, or M4A): The released track’s audio.
  • Release Date: The track’s release date (YYYY-MM-DD).
  • Artist ID: The Spotify ID of the artist.
  • Spotify Streams (t-1, t-2, t-3): The streaming counts for the past three days.
  • ISRC: The International Standard Recording Code for the track.
  • Alerts (Optional): Any important marketing alerts related to the track.
  • Focus Track (Yes/No - Optional): Indicates if the track is a current marketing priority.
  • Streaming Goal (Optional): Target number of streams the label wants to achieve.
Key Outputs and Interpretation:
  • Predicted Daily Streams: K5 predicts daily streams for the next 7 days after the initial 3 input days. (Include a line chart visualization, like the one generated by Songbird, showing predicted vs. actual streams over time. Annotate the chart clearly.) Example: In the provided example output, the prediction field shows the predicted streams for each day (target_day), while actual_streams show the real streams from Spotify. Visualize this as a line chart inside the product.
  • Upper and Lower Confidence Bounds: The model provides confidence intervals (“upper” and “lower” bounds) for each prediction, indicating the range within which the actual stream count is likely to fall.
  • Velocity: Rate of change in streaming performance, indicating momentum. Higher velocity suggests faster growth. (If shown in the Songbird output, include an explanation and visual representation, perhaps a separate line on the chart showing velocity).
  • Recommendations (If applicable): For selected labels, K5 includes specific marketing interventions and their estimated impact (“lift”) on streaming performance. These are shown in Recommendations section.
For example, running a TikTok influencer campaign has an estimated lift of 1,425,020 streams, while running press/PR only has an estimated lift of 300,000 streams, meaning it is likely to cause a lower lift in streams. (Include screenshot) Limitations:
  • Short-Term Scope: K5’s predictive accuracy diminishes beyond the two-week timeframe.
  • Reliance on Initial Data: The model’s accuracy depends on the quality and reliability of the initial three days of streaming data.
  • External Factors: Unforeseen events, changes in listener behavior, and competitive releases can influence actual streaming performance, causing deviations from predictions.
Example Scenario: A label releases a new track and uses K5 to monitor its performance. After a few days, they observe that the track is underperforming compared to predictions. K5’s causal inference module recommends increasing investment in a specific marketing channel (e.g., TikTok influencer campaign). The label implements the recommendation and observes a significant positive impact on streaming numbers, demonstrating the value of K5’s insights for real-time campaign optimization. After the initial two weeks, they switch to the K7 model for long-term performance forecasting and strategic planning.