---
title: "Cold-Starting Podcast Ads with Multi-Task Learning"
url: https://daily.dev/posts/cold-starting-podcast-ads-with-multi-task-learning-0wzksgjib
source_url: https://research.atspotify.com/2026/3/cold-starting-podcast-ads-with-multi-task-learning/
type: article
source: "Spotify Research"
published: 2026-03-17T13:00:33.480Z
updated: 2026-03-17T13:00:59.549Z
tags: ["machine-learning", "spotify"]
reading_time: 8
upvotes: 0
comments: 0
language: en
---

> ## Documentation Index
> Fetch the complete documentation index at: https://daily.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Cold-Starting Podcast Ads with Multi-Task Learning

**[Spotify Research](https://daily.dev/sources/spotify_research)** · 8 min read · 0 upvotes · 0 comments

## Summary

Spotify Research presents a unified multi-task learning (MTL) framework that jointly optimizes podcast ads and promotions within a single model, replacing previously siloed systems. The model shares a common encoder for user, content, and context representations, with task-specific prediction heads for streams, clicks, likes, and follows. Two key techniques—directional loss masking and source-balanced sampling—prevent negative transfer between channels. Offline ablations show the 5-task MTL model achieves +4.5% Promotions AP and +50.2% Ads AP over the promotions-only baseline. Online A/B tests across 180+ markets showed ~18% higher impression-to-stream rates, ~20% lower cost-per-stream, and ~18% more total streams. Cold-start gains were even larger for emerging creators, with up to ~60% improvement in impression-to-stream rate for low-popularity podcasts. The unified approach also reduces engineering complexity and accelerates new product launches.

## Full article

daily.dev links to this article rather than hosting it. Read it at the original source: <https://research.atspotify.com/2026/3/cold-starting-podcast-ads-with-multi-task-learning/>

## Similar posts on daily.dev

- [Profile-aware LLM-as-a-Judge for Podcasts: A Better Middle Ground Between Offline Metrics and A/B Tests](https://daily.dev/posts/profile-aware-llm-as-a-judge-for-podcasts-a-better-middle-ground-between-offline-metrics-and-a-b-te-8karfizw2) · Spotify Research · 0 upvotes · 0 comments
- [Generalized user representations for large-scale recommendations](https://daily.dev/posts/generalized-user-representations-for-large-scale-recommendations-m9jaa9lvm) · Spotify Research · 0 upvotes · 0 comments
- [Teaching Large Language Models to Speak Spotify: How Semantic IDs Enable Personalization](https://daily.dev/posts/teaching-large-language-models-to-speak-spotify-how-semantic-ids-enable-personalization-i4u6pcpnq) · Spotify Research · 1 upvotes · 0 comments
- [Why We’ve Been Optimizing the Wrong Thing in LLMs for Years](https://daily.dev/posts/why-we-ve-been-optimizing-the-wrong-thing-in-llms-for-years-sgoikytic) · Towards Data Science · 2 upvotes · 0 comments
- [Prompt-to-Slate: Diffusion Models for Prompt-Conditioned Slate Generation](https://daily.dev/posts/prompt-to-slate-diffusion-models-for-prompt-conditioned-slate-generation-zhdu1vt8h) · Spotify Research · 0 upvotes · 0 comments

---

Tags: [#machine-learning](https://daily.dev/tags/machine-learning), [#spotify](https://daily.dev/tags/spotify)

[View this post on daily.dev](https://daily.dev/posts/cold-starting-podcast-ads-with-multi-task-learning-0wzksgjib)