Skip to main content

Talroo's Search Algorithm

Simple systems often rank jobs primarily by keyword matches and bid price (CPC). That approach is straightforward, but it can overlook what matters to users and publishers: quality of job/user match and real-time inventory.

Talroo's proprietary algorithm uses a variety of signals to ensure the best jobs reach the best candidates.

Common Pitfalls that Talroo Addresses

Using a simple semantic matching and/or reverse-CPC rank can lead to several pitfalls:

  • High CPC Domination: High-bidding jobs that are not relevant to the jobseeker's background are surfaced every time in a reverse-CPC ranking regime
  • Synonyms and Abbreviations: Matching "doctor" queries to doctor jobs works, but what about "physician" queries? "MD"?
  • Ignoring User Data: Matching purely on the query ignores information about the user that may be critical

Talroo addresses these pitfalls through advanced machine learning strategies.

Inventory Changes

Real-time Status Changes

Jobs are frequently refreshed to ensure limited expired content is shown to users. User clicks on expired jobs is considered a rare event and are handled differently.

Expired Clicks

When jobs are taken down, cached data or xml feeds may still contain the job. Talroo handles clicks for expired jobs differently than jobs that are still live. See Expired

User data

PID (Personalization)

API based integrations can benefit from personalization in the results including history, conversion signals, answers to resumes and questions. See Personalization