Programmatic audio in your own DSP seat, starting with a supply path audit and an inclusion list, judged on a geo holdout rather than last click.
About this service
Audio has no viewport. Every quality check your team relies on elsewhere, viewability, scroll depth, time in view, does not exist here, and the people who profit from invalid traffic understand that better than most buyers do. So the first work on a programmatic audio account is not a bidding strategy. It is a supply path audit: sellers.json, ads.txt and app-ads.txt reconciled across every route to the same publisher, and every route that cannot be traced to a named seller blocked before a dollar moves.
Where the money is placed:
Buying runs in DV360, The Trade Desk or Amazon DSP, usually whichever seat already holds the client's display and retargeting, because a separate audio seat creates a frequency problem nobody can subsequently solve. Supply reaches those seats through Triton Digital's a2x, AdsWizz, Spotify Ads Exchange, SiriusXM Media, and Megaphone's targeted marketplace for dynamically inserted podcast inventory. Every account runs on an inclusion list. Open exchange audio without one is not a strategy, it is a subscription to somebody else's arbitrage.
Formats and their traps:
VAST 4 audio with a companion banner that is regularly served to a screen switched off, in a car, or in a pocket, which is why companion clicks are a delivery check and never a result. Ad pods, where position materially changes the product and is often undisclosed unless the deal terms demand it. Fifteen and thirty second variants requested per pod position rather than assuming one length travels everywhere. And a standing check that a line item labelled audio is not video inventory with a static frame, which we still find in the smaller supply sources.
Frequency, said plainly:
Much of this inventory arrives with no cookie at all: a smart speaker, a car dashboard, a podcast application. Frequency is enforced publisher-side against IP and user agent, and cross-seat frequency capping in audio is a fiction. We cap where capping actually functions, we buy fewer paths so the caps mean something, and we tell you what share of the buy is genuinely uncapped instead of showing one number that implies control we do not have.
How performance is judged:
Not on last click, and not on view-through windows measured in weeks, which in audio manufacture conversions out of coincidence. The buy is judged against a proxy agreed and signed before launch, branded search volume, direct sessions, code redemption, or store visits where a location partner is already contracted, read against a geo holdout. Ghost bid incrementality is not available across audio supply the way it is in display, so a matched-market design is the honest instrument and we will not claim more precision than it carries.
Not included:
Creative production and script variants, which are separate work. Buying through a reseller's seat, because we will not run money where we cannot see log-level data or where a fee sits inside an undisclosed margin. And audio placed inside an automated goal-seeking product that decides its own supply, which is the opposite of everything above.
Who this is wrong for:
Brands spending under roughly eight thousand dollars a month in audio, where buying direct from the platforms beats DSP fees and the tech tax by a distance, and we will say so rather than take the retainer. Teams who want listen-through rate on a dashboard: it will read above ninety percent indefinitely and tell you nothing. And anyone who cannot give us access to their own analytics, because every honest measurement in this channel lives on the advertiser's side, never on the platform's.