TALENLYTICS™ SEARCH

Talenlytics ™ SEARCH is a predictive model that can predict and thus advise the hirer the talents who match the search criteria; and vice versa, to advise the talent on the jobs that match their criteria. A multivariate regression model will be built with explanatory variables (x_i) extracted from information of the talents and the jobs.

With the coefficients or weights input by the hirer, the multivariate regression model represented below will be used to general an Overall Match Score as:

Overall Match Score, y = β₀ + β₁x₁ + β₂x₂+ β₃x₃+ … + β_nx_n

The explanatory variables (x_i) include the following that are derived from the original variables:

Original variables	Original format	Derived variables (xi)	Derived format
Job category	Categorical	Job category complexity level	Boolean
Top skill	Categorical	Skills match percentage	Boolean
Second skill	Categorical	Skills match percentage	Boolean
Third skill	Categorical	Skills match percentage	Boolean
Country	Categorical	Location match factor	Boolean
Talenlytics™ Index	Numerical	Talenlytics™ Index	Numericals
Job rating	Numerical	Job rating	Numerical

Hirers will allocate initial weights (think of them as level of importance) or accept default initial weights, to the selection criteria, such as skills, location, knowledge score. The algorithm will search and match freelancers by computing an overall match score using the multivariate regression model, and return the matched results.

In addition, the algorithm will determine the immutable regions for all the criteria, to predict which selection criteria will result in new and better matches, in order to advise the hirer to adjust weights for the selection criteria. Immutable region for each selection criteria is the region of weight where adjustment in weight within this region will not alter the matched results. For example, if the immutable region for skills is from 0.6 to 0.9, adjusting the weight for skills within this region will not alter the matched results. Immutable regions will also indicate the directions for weight update. For example, increasing the weight for skills from 0.8 to beyond 0.9 will alter the matched results more effectively and return a new set of matched results, than decreasing the weight from 0.8 to less than 0.6.

Immutable regions are indicated as the highlight band on each criteria bar. The objective is to determine the immutable regions which can advise the hirer that changing the weights of the selection criteria will be able to return a new set of matched results that may best fit the job. Such an advisory function aims to overcome the unknown unknowns paradigm which encourage hirers to look beyond their initial set of matched results, and explore other possible results.

After confirming the shortlisted candidates which match the selection criteria, bid invitations can be sent out to these shortlisted freelancers. Our predictive search and matching algorithm is different from most traditional matching algorithm which usually match field-to-field (commonly known as Boolean retrieval technique), and will end up with many records which will satisfy the criteria. As the pool of freelancers grows, such field-to-field matching will not be effective and serves little purpose. Talenlytics™ Search will result in better matches and using immutable regions, the match records will be more precise and thus effective.