automation

Silence is Expensive: How Publishers Can Optimize Bid Request Metadata to Capture the Exploding Demand for Programmatic Audio

A technical deep-dive for publishers on optimizing OpenRTB audio objects, leveraging metadata to unlock premium programmatic demand in the booming digital audio market.

Silence is Expensive: How Publishers Can Optimize Bid Request Metadata to Capture the Exploding Demand for Programmatic Audio

The Invisible Boom

We are currently witnessing a seismic shift in media consumption, one that is audible rather than visible. The digital audio market—spanning streaming music, podcasts, digital radio, and in-game audio—is exploding. Advertisers are following the ears, moving billions of dollars into programmatic audio channels to capture audiences in "screenless" moments: while driving, cooking, working out, or trying to fall asleep. However, for many publishers and supply-side platforms (SSPs), this boom is happening in a parallel universe where they are unable to fully participate. Despite having premium listeners and high-quality content, their inventory is selling for pennies on the dollar, or worse, going unfilled. The reason is rarely the quality of the audio content itself. The reason is metadata. In the display and video worlds, buyers have relied on visual verification and viewability tags to understand what they are buying. In the audio world, the environment is often "headless"—there is no screen to scrape, and frequently no visual component to verify. The Demand Side Platform (DSP) is flying blind, reliant entirely on the structured data passed in the bid request to make a purchasing decision. If your bid request does not explicitly tell the buyer the genre, the duration, the device context, and the technical protocol, the buyer's algorithm will default to "pass." In this article, we will dismantle the OpenRTB Audio object and reconstruct it to show exactly how publishers can optimize their signal quality to capture this exploding demand.

The Problem: Audio as "Video Without Pictures"

For years, the ad tech industry treated audio as a hacky subset of video. Legacy integrations often utilized VAST (Video Ad Serving Template) tags and simply stripped out the visual elements. While this worked for basic delivery, it created a data desert. When a DSP receives a bid request that looks like a video object but lacks width and height dimensions, it creates friction. Is it a mistake? Is it a broken video player? Or is it a podcast? To maximize yield, publishers must transition from these legacy hacks to fully robust, native implementations of the OpenRTB Audio object. This is not just a semantic difference; it is a financial one. Buyers invoke specific targeting parameters for audio—mood targeting, genre targeting, companion ad availability—that simply do not exist in video bid requests. If you are a publisher sitting on a treasure trove of podcast or streaming inventory, your first audit point is your bid stream. Are you selling audio, or are you selling "broken video"?

The Technical Core: Anatomy of the OpenRTB Audio Object

The OpenRTB specification (currently at version 2.6) provides a rich framework for describing audio opportunities. Yet, looking at the average bid stream, most publishers populate only the mandatory fields. This is the equivalent of trying to sell a luxury car by listing it as "Vehicle, 4 wheels." Let us look at the critical fields that separate premium inventory from the rest.

1. MIME Types and Protocols: The Gatekeepers

The most common reason for a bid rejection in audio is a protocol mismatch. The mimes array and protocols array in the bid request act as the primary filter.

  • mimes: You must list every format you can accept. While audio/mp4 and audio/mpeg are standard, omitting newer or higher-quality codecs can limit your demand pool.
  • protocols: This is where the industry history gets messy. DAAST (Digital Audio Ad Serving Template) was the original standard, but it has largely been subsumed into VAST 4.x. You need to signal support for the specific versions of VAST you can handle. If a buyer has a VAST 4.1 creative and you only signal VAST 2.0 support, they will not bid.

2. Duration and Linearity

Audio creatives are rigid. Unlike a display banner that can scale, a 30-second audio spot cannot be squeezed into a 15-second slot without destroying the user experience. Publishers must be explicit with minduration and maxduration.

  • Common Pitfall: Setting a maxduration of 30 seconds when you have a 60-second slot available. You are artificially excluding long-form storytelling ads (like host-read style creatives) that often carry higher CPMs.
  • Best Practice: Use arrays or ranges if your ad insertion point is flexible. If you have a podcast mid-roll that can take anything from 15s to 60s, signal that flexibility.

3. The Code Sample: Good vs. Bad

Let us look at what this looks like in JSON. The "Lazy" Implementation (Low Yield):

"imp": [
{
"id": "1",
"audio": {
"mimes": ["audio/mp4"],
"minduration": 15,
"maxduration": 30,
"protocols": [2]
}
}
]

The "Rich" Implementation (High Yield):

"imp": [
{
"id": "1",
"audio": {
"mimes": ["audio/mp4", "audio/mpeg", "audio/ogg"],
"minduration": 5,
"maxduration": 60,
"protocols": [2, 3, 7, 8], // Signaling VAST 2.0, 3.0, 4.0, 4.1
"startdelay": 0, // Pre-roll
"sequence": 1,
"battr": [1, 2], // Blocked creative attributes
"api": [7], // Open Measurement SDK support
"companionad": [
{
"w": 300,
"h": 250,
"id": "companion_1",
"asset": { "img": { "type": 1, "w": 300, "h": 250 } }
}
]
}
}
]

In the rich example, we have opened up the duration window, expanded protocol support to modern VAST versions, signaled that we support the Open Measurement SDK (crucial for verification), and offered a companion ad slot. This inventory is now visible to significantly more buyers.

Content Signals: The Contextual Goldmine

Because audio often lacks cookies (especially in app and smart speaker environments), contextual data is king. The content object within the bid request is where you prove the value of your audience.

The Power of cat (Categories)

In display, a buyer might target "News." In audio, they want "Daily News Briefing" or "True Crime" or "Lo-Fi Beats for Studying." You must utilize the full depth of the IAB Content Taxonomy. Sending a generic "Music" category (IAB-1) is insufficient. If you are streaming a playlist of 90s rock, send the specific sub-category.

  • Why it matters: Brand safety. A family-friendly brand might block "Podcasts" generally because they fear unstructured banter, but they will whitelist "Educational Podcasts" (IAB-5). If you don't send the granular category, you get blocked by the general filter.

Language and User Agent

Programmatic audio is borderless, but advertising is local. The wlang (Content Language) field is mandatory for any publisher with global reach. There is nothing more wasteful than serving a Spanish ad to a German listener. Furthermore, the User Agent (UA) string in audio is often unusual. It might be a Sonos speaker, an Alexa skill, or a specialized podcast app. Publishers should ensure their SSP is correctly parsing these UAs and passing them in the device object. If the UA looks like a bot to the DSP because it's non-standard, your traffic will be flagged as IVT (Invalid Traffic).

Podcasting and OpenRTB 2.6: The Structure of Sound

The release of OpenRTB 2.6 was a game-changer for audio, specifically for podcasting. It introduced the concept of "Structured Pods" to the bid request. In a typical podcast, you might have an ad break (a "pod") containing three slots. Previously, publishers would send three separate, unconnected bid requests. A DSP might win all three and play the same GEICO ad three times in a row. This is a terrible user experience and causes listeners to churn. With OpenRTB 2.6, you can describe the entire pod structure:

  • podid: A unique ID linking the requests.
  • slotinpod: The position (1 of 3, 2 of 3, etc.).

This allows DSPs to perform "competitive separation." They can ensure that if they buy Slot 1 for a car insurance company, they don't buy Slot 2 for another car insurance company. Paradoxically, giving buyers this control increases your yield. Buyers are willing to pay a premium for "exclusive" presence in a pod, knowing they won't be back-to-back with a rival.

Identity and Addressability: The Cookie-less Reality

Audio has been "cookie-less" long before Chrome decided to deprecate third-party cookies. Smart speakers do not have cookies. Mobile apps use resettable device IDs (MAIDs). To succeed in programmatic audio, publishers must pass whatever identity signals they have.

  • IFAs (Identifier for Advertising): On CTV and OTT audio devices, passing the IFA is critical for frequency capping.
  • PPID (Publisher Provided Identifier): If you have a logged-in user base (e.g., a music streaming subscription), passing a hashed PPID allows buyers to build frequency models across sessions without violating privacy.

There is also a growing reliance on "content signals as identity." If a user is listening to a niche podcast about "Vintage Watch Repair," that context is a strong proxy for demographic and income data, even without a user ID. This reinforces the need for deep, accurate metadata in the content object.

Verification: The Trust Gap

One of the historic barriers to programmatic audio adoption was the lack of measurement. "Did the ad play?" is harder to answer when the device is a Google Home sitting on a kitchen counter. The solution is the Open Measurement SDK (OMSDK) for Audio. Publishers must integrate SDKs that support OM. In your bid request, the api field should contain the integer 7, which signals support for OMID (Open Measurement Interface Definition). When a major agency buyer sets up a campaign, they often tick a box that says "Target only OMID-compliant inventory." If your metadata doesn't signal this support, you are filtered out before the auction even begins. You might have the technology installed, but if the metadata doesn't announce it, it effectively doesn't exist.

Audibility vs. Viewability

In display, we track "Viewability" (50% of pixels in view for 1 second). In audio, the metric is "Audibility."

  • The Metric: Was the audio player unmuted? Was the volume above a certain threshold? Was the tab active?
  • The Signal: Passing `playerstate` or `device.lmt` (Limit Ad Tracking) signals helps buyers understand the listening environment.

Strategic Implementation: A 3-Step Audit for Publishers

If you are a publisher looking to capture this revenue, here is your roadmap.

1. The Metadata Audit

Run a trace on your bid requests. Don't just look at the volume; look at the payload.

  • Are you populating content.genre, content.series, and content.title?
  • Are you signaling the correct VAST protocol versions?
  • Are you passing a valid device.ua and device.ifa where permitted?

At Red Volcano, we often see publishers who believe they are sending "premium" data, but their SSP integration is stripping out the granular content fields to save bandwidth. Ensure your pipes are wide enough for your data.

2. The Companion Ad Strategy

Audio is often consumed on devices that do have screens (phones, laptops). Yet, many publishers neglect the companionad object. Offering a companion banner slot alongside the audio spot increases the value of the impression. It gives the user a way to "click" on the audio ad. It turns a branding channel into a performance channel. Even if you don't have a banner to show, signaling the availability of the slot can boost your bid density.

3. Supply Path Optimization (SPO) Alignment

Buyers are consolidating their spend onto fewer, higher-quality SSPs. They are looking for the most direct path to the inventory. Ensure that your ads.txt (or app-ads.txt) is clean and that you are authorized correctly. For audio, because the supply chain can be fragmented (hosting provider -> RSS feed -> aggregator app -> listener), ensuring that the entity sending the bid request is authorized in the sellers.json chain is complex but vital. If the buyer sees a "reseller" label without a clear chain of custody, they will likely block the request.

The Future: Transcribed Metadata and AI

Looking ahead, the next frontier in audio optimization is AI-driven transcription. We are already seeing SSPs and data providers who transcribe audio content in real-time to generate keyword-level targeting segments. Instead of just targeting "Sports," a buyer could target the specific mention of "LeBron James" within a podcast episode. For publishers, this means the content.keywords field in the OpenRTB request will become the battleground. Publishers who can proactively transcribe their content and populate this field will win budget from contextual display campaigns that are looking to extend their reach into audio.

Conclusion

Programmatic audio is no longer an experimental budget; it is a core component of modern media planning. However, it remains a technically demanding channel where the details matter. As a publisher, your content is your product, but your metadata is your packaging. You wouldn't sell a premium product in a blank cardboard box. Similarly, you cannot sell premium audio inventory with a blank or generic bid request. By optimizing your OpenRTB implementation—focusing on protocols, duration flexibility, granular content categorization, and verification signals—you move your inventory from the "invisible" pile to the "must-buy" list. The demand is there. It is loud, it is growing, and it is looking for a signal. Make sure you are broadcasting loud and clear.