October 2019 Release Notes

Released on 10/01/2019

Persistent IDs

In the past, IDs have changed from build to build and have been used exclusively to identify documents. From this build forward IDs will persist for 99% of profiles. License customers will receive a mapping from person v7.3 IDs to persistent IDs alongside a changelog which is documented here.

Freshness

  • This quarter we have refreshed 42mm of our global work experience data.
  • Similarly, we have refreshed 12mm of our US work experience data
  • We've begun to see an uptick in 2019 start and end dates due to some new data union partners. Global license partners will see 2.6mm start dates in 2019 appended to already existing and new experience objects.

Coverage Increases

  • Our Facebook URL coverage in the API dataset increased to over 884mm from 838mm, with increased coverage in the following fields tied to Facebook in the API:

    • Name - 881mm
    • Location - 454mm
    • Experience - 280mm
    • Education - 261mm
  • We have also successfully merged more of the facebook dataset into our main dataset:

    • Facebook tied to LinkedIn -- from 3mm to 14.8mm
    • Facebook tied to Email-- from 15mm to 22mm
  • We have also increased our coverage of phone numbers tied to current job in our production dataset from 11.5mm to 16mm

  • Our coverage of current professional emails increased from 28.5mm to 37mm

Data Field Changes

There have been no changes in any API params. The datapull dataset has had the version_status field added. This is explained in the Persistent Ids

Minor Improvements

  • Removed a data source that was incorrectly identifying ~1% of profiles education history as work experience
  • Improved the quality of primary LinkedIn profile tagging to better reflect current, active, LinkedIn URLs
  • Improved merging logic across profiles to better handle changes in social profile username
  • Removed erroneous linkedin URLs that were being created by a misparse
  • Improved API matching logic for locations. We are now using phone area codes as a potential flag to improve match likelihood as well as some other minor fixes.