About the Client & Product:

The client is responsible for an industry-leading e-discovery product. In mid-2020, there were huge shifts in the types of media available for the legal discovery process–primarily, there was a huge influx of audio and/or video-based media due to the COVID pandemic, remote work, and video calls.

MY Role

Lead Product Designer + User Research

Collaborators

Product Manager
(2x) FE Engineer
(1x) BE Engineer

Feature Validation

Prior to this feature’s development, discoverable media was uploaded individually, reviewed and transcribed by hand, and finally, marked up for discovery. The manual review step alone was a crawl, resulting in an average of 3 voicemails reviewed per hour.

SETTING THE STAGE:

This product uses processing engines during upload, which creates plain-text versions of files from native documents, among other useful things. This text-based processing also occurs when media files are uploaded, which inspired the product team with the idea that this feature could indeed be possible.

The team knew there was deep value to this feature, but wanted to validate that customer buy-in existed before beginning creative iteration. Before scheduling calls, an empathy map was created to reflect the primary concerns and observations from users interacting regularly with A/V content. (See below)

In addition, the market research conducted around understanding technical limitations and possibilities for this feature uncovered usable logic for auto-generating closed captions, searchable transcriptions, and more. These findings enriched the original scope of the feature, and were invaluable additions to a first version.

some questions the team posed:

• How might we display transcribed, text-based context while providing an anchor or bookmark to the native content?

• What parts of the A/V review process are consuming the most resource time?
  (Transcription, QA review of text transcription, tagging relevant timestamps, etc.)

• How are teams currently handing off produced A/V content to opposing counsel?

• How might we create timestamp anchor points around transcribed words?

Findings & Shipment

CLIENT FEEDBACK

• Need content to be sectioned, since our A/V content is quite lengthy. Typically, separating content by any verbal pause is our standard.

• Annotations that supports linking timestamp(s) will take this feature from nice, to a virtual necessity during A/V-heavy litigation.

• Potential to allow the user to be able to turn on/off speaker identification and tagging. e.g Speaker 1 renamed as “First name, Last name”.

• Allow the user to be save a sound bite from a particular keyframe.
  

FIRST VERSION and Shipment:

After considering customer feedback, we focused on the following features for a first version of A/V Transcription:

• Auto-transcription for both audio (voicemails, Slack calls, etc.) and video content (Zoom calls, security footage, etc.)

• Search

• Redact & bulk-redact via clicked, transcribed word

• Bulk actions for searched content, including adding transcript-based or media-based redactions, bleeps, blank frames, muting etc.

• Right click functionality for individually-selected transcript text, including redactions, bleeps, blank frames, muting, etc.
  

FOR FUTURE VERSIONS:

• Support for annotations, including video and text-based annotations

• Support for linking timestamp within comments and annotations, similar to YouTube functionality

• Support for renaming speakers

Impact & Results

The positive impact from this feature release was felt virtually instantaneously— legal teams were saving ample amounts of billable hours on A/V-heavy cases by offloading the burden of manual review and markup. In an average A/V centric case, there would be around 3000 voicemails, with teams able to manually review at a rate of 3/hr.

With the release of A/V Transcription, teams were able to review A/V content at an average rate of 25/hr, bringing an average 800% boost in task velocity.

Additionally, legal teams were seeing an average 10% of billable hours saved. Research and validation helped facilitate the overwhelming success of this feature. Not to mention, the impressive metrics empowered immediate planning around a v2.0 of this feature, which garnered praise and delight from our users!