Need for Quality Skills Data

We will explore one function of HR - Recruitment and examine the root cause of its poor efficiency.

HR covers different functions including Recruitment, Learning & Development, Workforce Development and others. In the functions where skills are involved data quality of skills matter. And when quality of data is poor effectiveness of the function is poor. We will analyze the recruitment function and see its dependency on skills data. Finding jobs or people is frustratingly costly, time consuming, and takes a lot of effort. Job seekers are frustrated that they are unable to discover the right openings, that they are swamped with job alerts and calls that are unrelated to them and that the hiring process is tediously long. Frustration is even higher at the hiring manager’s end. Their concerns are similar. Inability to find the right people at the right time, receiving resumes that are irrelevant, and having to spend a long precious time in the recruitment process. Ultimately, there are two fundamental functions in recruitment, (1) discoverability and (2) matching.

The e-commerce industry is also in a similar function of enabling discoverability and matching; discoverability and matching of right products at one end and right customers at the other end. E-commerce is doing a fabulous job despite there being millions of products.

Why is it that we can’t do in recruitment what is done well in e-commerce?

In the last five or six years there has been a lot of investment into start-ups trying to solve this problem using AI. However, the problem persists and no scaleable and viable solution has been found. The reason these have failed is that it is not an AI problem. After all, even AI can perform well only when there is quality data. And herein lies the key to solving the problem: quality data on skills.

One reason is that the key information regarding products is well captured. The data on products is well parameterized. Different products require different kinds of parameters of data. What dataset is required for, say, TVs is not what is needed for pillow covers. Both segments require data but different parameters of data. That difference apart, discoverability and matching works well because the data is captured well for all products. So comparing TVs is easy, data driven, and insightful. Imagine data of people and jobs captured in the same manner: parameterised, quantitative, structured, neat, and less textual. Wouldn’t the recruitment space see a boost in discoverability and matching and bring efficiency? It will. However, here lies a real problem.

There are two components of data when it comes to people and jobs. One is general (or should we call it demographic) information such as email, location, job titles, degrees, past companies and such. Then there is this skills information like functional/technical skills, soft skills, knowledge, domain experience, and others. The former, from a data perspective, is easy to capture and analyse. The general information is clean, has patterns (emails for example), consists of discrete elements, and is quantitative (years of experience for instance). However, the latter i.e. the skills information suffer from lack of these. This information is fuzzy, subjective, unclear, lacks a pattern and consists of non-discrete elements. Because of this, expression of these is tough and difficult to analyze.

If we have to solve the problem of discoverability and matching in the recruitment space ,we need to solve the skills data problem.