Spot the Next Viral Song for a Record Label
More than 100,000 new songs are uploaded to platforms every day. The A&R (artist scouting) team at an independent label can only listen to about 300 a week and arrives late to emerging artists: by the time a song goes viral on TikTok, the major labels have already offered a contract. They want a radar that detects viral potential within the first 72 hours.
Where the money comes from
What to do now
- 1Weekly list of 300 prioritised songs for A&R.
- 2Measure how many of those detected end up going viral.
- 3Sign with staged contracts based on results.
For your committee
One page with the decision, the impact, the risks, the steps and the KPI to track.
The print dialog opens: choose "Save as PDF".
Watch out: more volume is not always more profit
Signing many one-song virals: promotion costs and advances are not recovered if the artist has no staying power. Combine viral potential with career signals.
KPI for the committee
Revenue per euro of advance and radar hit rate.
Detail for your data team
Proposed technique: Neural Networks (Deep Learning / CNN)
Why: The key information is in the audio: the hook, the production, the voice. Convolutional neural networks treat the spectrogram as an image and learn sound patterns that nobody can describe in a table. With 500,000 examples and pretrained audio models, it is the right approach.
Cost of getting it wrong: A false negative is a hit that the competition signs; a false positive is wasted A&R time (cheap). That is why recall is prioritised at the top of the list.
Data needed: Audio of 500,000 songs from the last 3 years, labelled as 'viral' (made it onto viral charts) or not.; Early signals from the first 72 h: plays, skips in the first 30 seconds, saves, uses in short videos.; Metadata: genre, duration, tempo, language, artist followers.