Audio Content Recognition is a technology that can identify a piece of audio or video content by analyzing small parts of the audio contained within the content. ACR can identify which show is being played on a TV by its soundtrack.
Overview
Audio Content Recognition is a technology that is used to identify a piece of content, using the soundtrack of that content, with no human intervention. The technique has been used in consumer applications like Shazam to allow a user to identify the title of a song when the software can analyze the audio of the song.
Use in Television Data Collection
Recent generations of Smart TVs have embedded chips and software that allow them to identify the program that is being watched on the television by analyzing the audio of the program. ACR chips can detect any content that has an associated audio track, including both full-length television content and shorter TV commercials. The data is then pseudonymized, aggregated, and pushed to the television manufacturers. The resulting data sets are used to measure the audiences that were watching content, leveraged to help advertisers plan their television advertising budget, and measure the effectiveness of an advertising spot.
Comparison to Nielsen
Historically TV measurement has been panel based. The leader in the space has been Nielsen. ACR offers three primary advantages to Nielsen's measurement products:
- Scale: ACR data can be collected across millions of devices, compared to a Nielsen panel, which might have only tens of thousands of households.
- Precision: ACR data can be collected second-by-second whereas Nielsen data is typically aggregated in 15-minute blocks
- Transparency: ACR data can be made available directly to the consumers (analytics companies, advertisers, television networks), allowing them to build their solutions and derive their insights. Nielsen typically does more packaging of their data, making those entities more reliant on Nielsen's analysts.
While ACR provides several advantages, it does have some shortcomings when compared to the Nielsen offering:
- Unrepresentative: Nielsen goes to great lengths to make sure its panel-based approach is representative of the populations being measured. ACR technology only lives within Smart TVs, which are not proportionally distributed to the general population. This can be overcome by downsampling ACR data to make it proportionally representative.
- More Complex: Because the data is available at greater granularity, users of the information often have to create a set of heuristics to make sure the data gives them meaningful insights. Nielsen has presumably spent decades working on this problem.
- Privacy: While the data collected via ACR is pseudonymous, detractors point out that because the data is more granular, there existing privacy risks associated with ACR data collection and usage.
Additional Resources
Wikipedia: Audio Content Recognition
DMN: 4 Things You Should Know About ACR