Skip to main content

Don’t Let Your Lack of Audio Discovery Speak Volumes

When preparing for discovery, many organizations focus on obvious data sources such as their e-mail, letters, spreadsheets, and reports, but they may be forgetting an essential part of discovery: audio files.

Even for organizations not subject to laws that require them to keep certain oral communications, such as the Dodd-Frank Wall Street Reform and Consumer Protection Act, audio recordings are becoming more prevalent in requests for production—particularly from regulators. For example, the Data Delivery Standards of the Security and Exchange Commission and the Commodity Futures Trading Commission specify details for how parties must produce both audio and video files. These agencies expect parties to provide files playable in Microsoft Windows Media Player format, along with their metadata.

All told, organizations may have to review thousands of hours of audio data in varying formats, and the review can be treacherous. Traditionally, parties have opted for manual listening or transcribing. With manual listening, reviewers must focus for long periods and take accurate notes about what they hear. If they miss something, then the entire recording may have to be listened to again. Transcription can be equally problematic. It is time-consuming and expensive, and its accuracy may suffer if done by people without legal experience. Background noise may muddle certain words, or the sound quality or connection could be poor. Once transcribed, certain nuances may be lost, such as tone and inflection. Even so, keyword searches are possible, but accents, dialect, and regionalisms can render searches somewhat less effective.

Given these hurdles, parties may have a legitimate argument that reviewing audio evidence poses an undue burden. However, before taking this approach, it would be wise to consider technology advances that can simplify the review process. One option is speech-to-text software, which can transcribe audio. However, it suffers from the same problems as transcription, and the software may have a difficult time juggling conversations between multiple speakers. Phonetic indexing, like that offered by Nexidia, offers a more reliable solution. Nexidia’s software breaks audio content into phonemes, which are the building blocks of speech, and uses them to create a searchable index of each file that reviewers can parse to find the most relevant parts of each recording. Because of this, the software can deliver more accurate results—phonetic indexing does not depend on accent or dictionary spellings to be searchable—and it can process in excess of 20,000 hours of audio per day.

In addition to leveraging technology, organizations with potentially discoverable audio data can follow several other best practices. First, they should include it, denoting its location and format, in their data map. They should also ensure that their litigation hold process incorporates audio data, as many users tend to delete voicemail on their company and mobile phones. Next, they should develop a plan for how to harvest audio files, along with their metadata. Finally, organizations should retain the help of experienced e-discovery experts, including linguists, to determine how best to deploy keyword searches to mine for relevant data.

Sheila Mackay is vice president at Conduent. She can be reached at

About the Author