There are two aspect to talent analysis. We look at the current talent supply as well as predict the forecast for next few years. Predictive analytic techniques are used for both the types of analysis.
Talent Supply Forecast: We have devised predictive models/ equations for every combination of Domain and Location available on our platform. This roughly translates to over 20,000 independent models that help us predict the future values of installed talent based on a set of wide ranging variables optimized for each model. Every model is driven by a set of primary and secondary variables that range from historic installed talent numbers to number of companies and profiles in any given location. These models utilize techniques ranging from linear regression or logistic regression to complex decision trees.
Installed Talent Supply: As mentioned above we have over 20,000 models that help in predicting the installed talent. In addition to those 20k models we also have a set of models that we use to estimate the skill and function level talent distribution. The overall mechanism involved in this is complex but in simple terms we can describe it in the following way. The domain level talent is at high level and easier to estimate through research. Based on the installed talent data by domain we do cluster analysis to identify the locations that share similar traits. We use a method of vector quantization similar to k-means clustering algorithm method and identify a prototype city. In addition to the statistical implications we also implement external research based findings to identify a central location for any given cluster. Usually the central location has optimum data that has been cross verified using our external data mining techniques as well as verified government reports. Using this and a host of multiple other parameters we then try and fill any gaps for the locations in the same cluster.
Fresh Talent Pool is estimated by studying graduating student populations across multiple majors, who will join the workforce next year, for a specific Domain and City of interest. Fresh Talent will overlap across Domains, as we can never truly estimate the function or skills non-working professionals hold. Fresh Talent in the ‘Software Product Development’ domain in Washington DC for example, is an estimation based on the number of students who are enrolled in Washington DC universities, studying specific subjects (i.e Computer Science) which could eventually place them in a software product development based job role. There is no guarantee however, that students in those areas of study will move into Software, so we also consider a subset of students across other domains, which in this example, could also extend into Hardware Product Development or IT Services.