How Tracker Ranks Longlisted Candidates
When longlisted candidates are searched across the Tracker database, they are returned in a ranked order, but how is this ranking calculated?
Unlike many ranking engines that see frequency of search terms and a key indicator of the level of skill attained by an individual, we take a more human approach, attempting to replicate the same logical processes a human would go through if scanning a candidate’s resume. But this alone is not enough.
Our advantage over many autonomous sources such as job boards is that we have a special ingredient, namely the candidate’s history, both in terms of their resumes and relative success or failure on Jobs over many years – something you will never see in a resume. No candidate will have a section stating how many jobs they did not get selected for or failed at interview stage.
The basis of our ranking methodology therefore considers the following key facets:
|
Prevalence |
How often the word or phrases being searched appears in the resume and its context, such as “…extensive use of…” vs “…some exposure to…”. Prevalence is still a key aspect in identifying rank, especially when it comes to literally skills such as “WordPress” or “Sage Accounts” etc. |
|
Chronology |
Resumes may be littered with the search terms or derivatives of it, but when it happened is also of importance. Chronology looks at when these skills were used and whether they are part of career history or skills acquired. Someone with a French qualification from High School does not make them a candidate as a French teacher |
|
Duration |
Knowing how long a particular skill has been used or role has been worked is also a key factor in ranking. Each job in their history has a duration and within that prevalent skills so plays a fundamental part in ranking |
|
Weighted Terms |
Some search terms carry more weight than others but by providing more terms, so the results gain higher quality. By weighting some terms over others, factors such a Prevalence, Chronology and Duration get armed with better intelligence as to how each term should be prioritized |
|
Semantics |
Interpretation is important in order to get a more rounded view of the Candidate pool. Excluding Candidates simply because the terms were too restrictive or at worst, spelled wrong returns poor results. So, allowing for common variations of terms and including them in the ranking ensures coverage in consideration |
|
Historic Factors |
The all-important X-Factor that no resume search or ranking engine will do alone is in the Historic Factors. These sit outside the resume entirely and look at the Tracker database to apply weighting to all results, favouring successful candidates over unsuccessful ones |
Each of these key aspects of a Candidate’s profile, be that from their current resumes, the resume we got 5 years ago, or their history on the 10 shortlists they have been on, we can correlate the search to not only identify the candidates but pitch them against each other.
Once we have the terms of the search, we can begin applying scores against each of these factors, some being standalone scores, others an iterative score of one factor within another, for example the prevalence of a given skill over time (how recent and for how long).
So, although it appears to be complicated behind the scenes, it is still not as sophisticated as the human mind. It cannot detect a “gut feel” from the way a resume is written for example, it can only go on the hard facts presented and apply artificial interpretation to give a highly accurate ranking that alleviates much of the heavy lifting when reviewing hundreds of candidates.
From a technology perspective, Tracker uses a highly customised instance of the Lucene search engine platform. Using our parsing technology, we prime Tracker by extracting each element of information from the resumes provided and structure these in the Candidate records. These form the building blocks for the indexes we search when ranking long lists.
For example:
- Current Job Title
- Current Job Description and Dates
- Previous Job Description and Dates
- Job before that (and so on throughout their entire job history)
- Education
- Technical Skills
- Language Skills
- Soft Skills
- Achievements
Each Candidate has an extensive “profile” within the search engine allowing us to apply most of the main 5 factors against it including Prevalence, Chronology, Duration, Weighted Terms and Semantics; at any time, and very quickly. This profile is updated and refined each time a new resume is received.
The results can then be compared to what else we know about that Candidate from their history in Tracker, applying a light weighting to those we know to be more likely a successful candidate over another.
So in under a second, not only have we found the candidates, but we’ve also done a fair bit of work looking at each one individually against the search terms provided to give what we feel is one of the most technologically accurate suggestions as to how they rank.