Advertising and media have co-existed for over 5000 years. In fact, the first ad seen in print dates back to a papyrus fragment in 3000BC (albeit, an unfortunate one).
Despite this long history of advertising, new challenges continue to arise that must be solved if we want to continue progressing our culture and society.
The audio industry, as the fastest growing media industry in advertising in 2022, has surfaced new systemic challenges in audio advertising—both in optimizing media buying for brands and ensuring equal opportunity for creators.
Advertising is a major source of revenue for creatives and businesses in every media industry. Because ad tech solutions directly impact content monetization, it’s essential that they’re correctly designed to maximize the opportunity while safeguarding consumers and brands from harmful content.
As seen in digital and video advertising, marketers have two needs that must be fulfilled in regard to brand suitability:
- Assurance that their ads appear in safe and suitable environments and never alongside harmful content
- The ability to leverage suitability insights to target content that suits their audience’s psychographics and brand personality (i.e. some brands intentionally seek out podcasts with edgier humor or strong opinions on social issues aligned with their core audiences)
Despite the great advertising opportunity available in podcasts—there are over 400+ million listeners engaged globally—the same tools available for video and display advertising do not exist for classifying audio content.
Today, the methods by which many people approach advertising investment are both financially inefficient and ethically risky. For example, the most popular methods for investing in audio are manual review and keyword-driven solutions. These both present major challenges that we’ll discuss below.
To fully grasp the necessity for an audio-intelligent and automated solution for targeting podcasts, let’s break down how legacy ad tech solutions have created big problems for creators, publishers, and brands.
The Problem with Manual Podcast Review for Brand Safety & Suitability
When an agency team vets a podcast, that’s a human review. A real person spends real time listening to a few podcast episodes to assess its safety and alignment with a brand’s values. We call this a “manual review.”
A manual review has three major problems:
- It takes a ton of time to complete.
- It’s ineffective and risky for brands because even a safe podcast may have episodes or segments a brand would want to avoid.
- It’s subject to the bias of the assessor.
It’s impossible to manually review all the episodes in a large catalog to find a brand’s best ad placement. Effective targeting is about finding optimal matches in real-time. When faced with thousands of hours of content or more, you cannot expect a human to manually review a publisher’s catalog (let alone a few publishers). It just can’t be done… even on 2x playback speed.
Not every episode of a “safe” podcast will be safe and suitable for a brand. A host’s interests and conversations can vary widely per episode and change over time. By only reviewing a sample set of episodes, marketers can put their brand at risk by having ads appear alongside unsafe or unaligned content.
In a worst-case scenario, having your brand appear next to harmful content can cause serious damage to your business’ reputation—you could lose up to 67% of your consumers, according to a poll by DoubleVerify and Harris.
Due to the shortcomings of manual review, industry professionals have resorted to early ad tech solutions built on keyword targeting and anti-targeting alongside manually vetted show inclusion lists.
These solutions are similar to keyword blocklisting algorithms seen in the digital and content moderation industries and lack the precision needed when processing language into advertising data like GARM’s safety and suitability categories and IAB’s content taxonomies.
The Problem with Keyword Targeting in Brand Safety & Suitability Solutions
Keyword targeting and anti-targeting are driven by taking the transcript of a podcast and:
- Reviewing keywords to match for alignment with a brand’s value and message
- Reviewing keywords to flag content as risky for a brand
The issue with these keyword blocklists is that keywords alone do not contain enough information to support effective targeting. Keywords lack insight into the larger context of a statement.
When assessing brand safety and suitability, a single keyword can be used in multiple contexts across a wide range of risk levels. For example, let’s look at the word “dead” which is one of the most blocked keywords in recent years.
The word “dead” can be used to reference “a dead person,” “The Walking Dead,” “Deadpool,” “The Grateful Dead,” a youthful idiom to describe finding something hilarious, or otherwise. Four of these five use cases are brand-safe and sought-after topics. Now, think about what happens when scaling this filter across the 455,000+ active podcasts on Apple. That’s a ton of content incorrectly classified and removed from advertising inventory!
In fact, incorrect keyword blocking cost U.S. publishers $2.8 billion in 2019.
Simply put, keywords lack context and context matters—it makes all the difference in the accuracy of safety and suitability classifications and the performance of an ad placement.
One reason keywords lack context is that they don’t provide enough information behind the intention and sentiment of their use in conversation—how and why a word is used is just as important as what was said.
For example, the intention behind content with adult keywords may be educational (e.g. biology podcasts), while other mentions could be meant to entertain (e.g. comedy podcasts). Both podcasts could be highly suitable for different brands. It depends on the context.
For an in-depth review of real podcast examples and our model’s suitability and contextual analysis, see the results section in this breakdown on our AI/ML model by Mercan Topkara, Sounder’s Chief AI Officer.
Solutions built upon keyword targeting are unable to conduct deeper analyses that deliver actionable brand safety, suitability, and contextual insights in audio. The industry needs advertising solutions that can surface meaningful insights accurately and quickly from large catalogs and new quality content released daily.
Sounder’s Automated Brand Safety & Suitability Solution
Our brand safety and suitability solution for audio is the first of its kind. It’s the only automated solution on the market that leverages both IAB and GARM industry standards to support transparent and confident media buying in podcasts at scale.
As an ad tech solutions partner of the Global Alliance for Responsible Media (GARM), we recently shared one of the first white papers showing how to apply GARM’s suitability standards to audio media.
When processing an audio segment, our solution examines the context and understands, for example, that “shooting the ball” is a sports reference in an episode positively discussing the game. Most importantly, it also recognizes this moment as a strong fit for an athletic retailer to advertise.
How does it work?
Our solution is built on AI, ML, and NLP models that were trained specifically on podcast content to understand speech, a capability we call audio intelligence. This enables our technology to evaluate topics, sentiment, tone, timing, and intent when analyzing audio segments for context and targeting insights.
By design, our solution processes human speech in audio data for several layers of insights with lightning speed. Sounder’s Audio Data Cloud can generate episode ratings, full transcripts, topics, entities, summaries, categories, and more for one hour of audio content in less than two minutes.
For publishers, our solution can process their catalogs into brand safety and contextual insights at the episode, show, and catalog levels. This makes packaging and profiling their media for advertisers take minutes rather than hours—a huge efficiency boost to ad sales and operations teams.
For agencies, our solution can help marketers find and invest in the best podcast content for their clients. Say a bank wanted to advertise on content with high-quality and educational conversations around personal finance. Our solution will surface content like this throughout the database down to the segment level.
I’m happy to say that the lack of standardized safety, suitability, and contextual targeting data in audio will no longer be a factor in limiting the valuable opportunities hidden in podcast content.
With the industry already demonstrating rapid growth in revenue and listening, we anticipate the widespread adoption of audio-intelligent ad tech will further accelerate this rise.
Please feel free to visit our brand safety page to learn more or reach out to me directly at email@example.com.