This morning advertisement measurement company Podscribe launched a framework for measuring conversion of YouTube advertising. This particular project is targeted to YouTube because Google does not share viewer IP addresses with third-party customers. So, a process called probabilistic modeling comes into play, using a statistical framework to analyze and process streaming audio signals in real time, resulting in probable metrics.
This sort of technology is applied to diverse audio statistics, including pitch tracking and noise reduction. That’s fine, but we wanted to know how it applies to podcast advertising metrics, so we turned to ChaGtPT for help. The pertinent result (with non-pertinent parts removed) is below:
How YouTube-Only Modeled Attribution Works
Instead of connecting an individual video view to a specific conversion, the model extrapolates outcomes using a three-step process:
- Establish an Audio Baseline: Podscribe looks at the advertiser’s historic response and incrementality rates from standard RSS podcast downloads. (Note: Advertisers must have existing audio conversion data with Podscribe to use this feature.) [1]
2. Tune for Video Specifics: The model adjusts the baseline using distinct contextual signals from each YouTube video, including:
- Where the ad appears in the video.
- The percentage of international subscribers.
- The video’s engagement rate (comments per view).
- Ad type (e.g., on-camera host-read vs. a static image overlay).
- Available promo codes, vanity URLs, and “How Did You Hear About Us?” (HDYHAU) survey data.
3. Model the Conversions: By blending the audio baseline with these specific video metrics, the platform estimates the true incremental impact and conversions of the YouTube-only placement. [END]
We observe that GPT drew its info from Podscribe’s explanation of today’s launch. It’s obviously a good resource, and it is HERE.
