What it does: Predicts first-week streaming performance on Spotify for unreleased tracks.

Why use it: K2 allows you to make data-driven decisions before a track’s release, enabling you to:

  • Prioritize and Optimize Releases: Focus your marketing and promotional activities on the tracks with the highest potential for success. Optimize release schedules and campaign strategies based on predicted performance.
  • Identify Strong Performing Tracks: Discover breakout tracks early, allowing you to allocate resources effectively and capitalize on their momentum.
  • Set Realistic Goals: Establish achievable streaming targets and measure campaign effectiveness against objective predictions.
  • Improve A&R Decisions: Use predictions to guide A&R strategies, identify promising artists and songwriters, and prioritize the development of high-potential tracks.
  • Refine Go-To-Marketing Strategies: Tailor campaigns to specific predicted performance tiers, ensuring that marketing messages and budgets align with realistic expectations.

When to use it: Use K2 after the track is finalized but before its release.

How to use it in Songbird:

  1. Upload Audio: Upload the finalized audio file (MP3, WAV, or M4A) to Songbird. The filename must be the Spotify Track ID. (Include a screenshot of the upload interface in Songbird, with clear annotations showing where to upload the file).

  2. Input Metadata: Enter the track’s planned release date (YYYY-MM-DD) and the artist’s Chartmetic ID. (Include a screenshot of the metadata input fields, clearly labeled).

  3. (Optional) Focus Track Designation: If this is a priority track for your marketing campaign, designate it as a “Focus Track.” This adjusts the model’s predictions to account for increased promotional efforts. You can also enter a confidence percentage for your focus track selection (e.g., 80%). (Include a screenshot showing the Focus Track input.) If you are uncertain about this selection, use the focus track probability prediction from the model itself.

  4. Run K2 Analysis: Click “Run Analysis” to generate predictions. (Include a screenshot highlighting the “Run Analysis” button.)

  5. Interpret Results: Review the prediction results displayed in Songbird. (Include a screenshot of the results page).

Key Inputs:

  • Audio File (MP3, WAV, M4A): The finalized version of the track.
  • Release Date: The planned release date of the track (YYYY-MM-DD).
  • Artist ID: The Chartmetric ID of the artist.
  • (Optional) Focus Track: A binary indicator (0 or 1) specifying if the track is a priority.
  • (Optional) Confidence in Focus Track (Percentage): Level of confidence in Focus Track choice.
  • (Optional) Streaming Goal: An explicit number of plays the label wants to achieve.

Key Outputs and Interpretation:

K2 provides a probability distribution across different streaming tiers, indicating the likelihood of the track falling within each tier during its first week of release.

(Include a screenshot example of K2 output in Songbird. Annotate the screenshot to clearly label the key outputs.)

  • Probability Distribution: The model outputs the probability (percentage) of the track achieving various streaming milestones during its first week.

The tiers are:

  • less 25k
  • 25k-50k
  • 50k-100k
  • 100k-300k
  • 300k-500k
  • 500k-1M
  • 1M-3M
  • 3M+

A higher percentage in a given tier indicates a greater likelihood of the track reaching that streaming level.

  • Focus Track Probability: The model outputs the likelihood of the track being a focus track. A focus track is typically a track the artist actively promotes on socials or other digital channels. Use this prediction to help inform whether to designate a track as “Focus.”

  • (If Goal is Provided) Probability of Reaching Goal: If you provided a target number of plays the label wants to achieve during the track’s first week of release.

Limitations:

  • Probabilistic Predictions: Predictions are probabilities, not guarantees. Actual results can vary due to numerous external factors.
  • Short-Term Focus: K2 is optimized for short-term, first-week predictions. Accuracy decreases for longer-term projections.
  • Dynamic Market Conditions: Model accuracy can be affected by unpredictable market trends, viral events, or changes in listener behavior.

Example Scenario:

A label is deciding between two tracks for an upcoming release. K2 predicts that Track A has a 50% probability of reaching the 100k-300k tier, while Track B has only a 20% chance of reaching that tier.

The artist and label decides to prioritize Track A for a more extensive marketing campaign, leveraging the insights from K2 to allocate resources efficiently and maximize the track’s potential for success.

They also designate Track A as a “Focus Track” to reflect their increased promotional efforts and rerun the K2 model. With the new input of setting the focus_track flag, K2 prediction shifts towards higher streaming tiers with higher confidence.

This detailed structure, including visuals and a clear explanation of inputs, outputs, limitations, and a concrete example scenario, serves as a template for other model cards. Remember to adapt the specifics to each model’s unique characteristics and functionality.

Predictive models overviewK5 post release short term