Programmatic Audio Advertising

The Programmatic Audio Landscape

Programmatic audio spans three main contexts: streaming music platforms (Spotify, Pandora, Amazon Music), podcast advertising networks, and digital radio (iHeart, digital AM/FM streams). Each has different inventory characteristics, targeting capabilities, and measurement approaches.

Audio is a companion medium — listeners multitask. They cannot click. The response to audio advertising is brand recall, direct URL visits (if they remember), and search. Attribution is harder than display or video, which affects how audio fits into performance-focused campaign structures.

Targeting in Audio

Audio targeting uses a combination of: declared user data (age, gender from account registration), behavioural data (listening history, playlist preferences), contextual signals (what they are listening to now), location data, and device signals. First-party data from streaming platforms is generally high quality because users log in with real credentials.

Podcasts are targetable by show, category, audience demographics, and contextual themes. Host-read ads (dynamically inserted but performed by the host) command premium CPMs because they blend into content more naturally than pre-recorded spots.

Measurement Challenges

Podcast measurement has improved significantly with prefix-based attribution (tracking download URLs through a measurement prefix) and pixel-based attribution for streaming. But the fundamental challenge — audio cannot be clicked — means last-click attribution systematically undercredits audio's contribution.

Brand lift studies, search lift measurement, and direct response URLs (vanity URLs and promo codes mentioned in audio ads) provide supplementary measurement. Attribution modelling that accounts for audio touchpoints provides a more complete picture.

Creative in Audio Is a Different Discipline

Audio is the one programmatic channel where the creative cannot be skimmed, skipped by an ad blocker, or seen out of the corner of an eye. It is also the one where the listener is usually doing something else — driving, cooking, walking — and cannot click.

Three consequences shape production. The first five seconds carry the brand, because there is no logo doing that work in parallel. A call to action has to survive not being clickable, which in practice means a memorable phrase or a spoken URL rather than a promo code nobody can write down. And host-read and announcer-read spots behave differently enough to be treated as separate formats — a host read borrows the host's credibility and cannot be reused across shows; an announcer read is portable and does not.

Dynamic ad insertion lets you vary the spot by listener and by moment, which is genuinely useful and quietly expensive: every variant is a production cost and a QA surface. Decide how many variants you will actually analyse before you commission them.

Podcast and Streaming Are Not One Channel

They are bought through similar pipes and they measure fundamentally differently.

Streaming audio — music services and radio simulcast — behaves like the rest of programmatic. The ad is served at the moment of listening, so an impression means something close to what it means in display, and targeting can use the signals the platform holds about the listener.

Podcasts historically counted downloads, which is not the same thing as a listen: a file can be downloaded automatically and never played. The IAB measurement guidelines exist precisely to constrain what may be counted, and whether a supplier is certified against them is the single most useful question to ask before buying.

The practical rule: do not compare a podcast CPM against a streaming CPM without first establishing that the denominator means the same thing. Much of the apparent price difference between the two is a measurement difference wearing a pricing costume.

Sources

What each claim on this page rests on. Entries are typed so you can see which are primary.