Understanding Skills Scarcity in An External Market — Part 1
What is Skills Scarcity? How to measure it? Where to use it?
Skills are the new currency in the changing world of work. Studies [1] show that the half-life of skills is five years, meaning that what we learned today will become obsolete or be forgotten in just five years. According to a recent survey by Mercer Global Talent Trends [2], more than half of organizations which responded to the survey, are targeting at upskilling and reskilling their critical talent pools to drive workforce transformation. On the other hand, based on a survey of 10K people across UK, Germany, China, India and US, PWC reported that 74% of employees are ready to learn new skills or retrain to remain employable in the future [3]. Skills are definitely becoming increasingly essential to the success of both individuals and enterprises.
Given this context, two essential questions for enterprise HR organizations are: 1) which future skills to grow in their employees so as to support their career development? and 2) which skills to retain so as to keep the company at a competitive advantage?
At IBM, we have developed a skills planning platform called Skills Value Framework (SVF) to seek answers for those two questions. As shown in the figure below, the SVF has two dimensions where the x-axis indicates the internal business demand of a skill and the y-axis indicates the scarcity level of a skill in an external market. In IBM’s expertise taxonomy, a skill is termed as a Job Role Specialty (JRS) in the format of JOB ROLE:SPECIALTY. In the example below, two JRS are given, namely Systems Admin (the job role) with Windows (specialty), and Software Developer (the job role) with AI (specialty).
Each Skills Value Framework dimension has 3 values, so together they form a 9-box. In particular, a Business Demand value has three values in its range: Design, Maintain, and Grow, which all depend on the level of internal business needs for specific skills. Skills Scarcity measures the level of skills supply as compared to the demand in an external market, and its range values are Low, Medium, and High. A High scarcity value indicates that the specific JRS or skill is very scarce in a particular country, and it would be difficult to acquire and/or retain such skill. In contrast, a Low scarcity value would indicate that the supply of a particular skill is readily available and easy to find in the given country. As an example, for the “System Admin: Windows” JRS as shown in the figure above, it is considered to have a Declining business demand yet a Medium skills scarcity in the United States. At IBM, we use this skills framework to guide the shifting of our workforce towards higher valued skills.
We have measured skills scarcity for all Job Role Specialties by country, as a workforce in one location can lack people with a specific skill whereas another country could have many workers with that same skill.
For the rest of this blog, I will focus on the Skills Scarcity aspect of this framework.
A. How Do We Measure Skills Scarcity?
The figure below shows the overall methodology of determining skills scarcity in a nutshell. At a very high level, the measurement has tapped into both internal data and external data for deriving the scarcity values. The final integrated data is further reviewed and approved by business units before it is released for downstream applications.
A.1. Use of Internal Data
We have used IBM internal data to build an AI model to predict the Skills Scarcity based on various factors that capture the supply and demand status of a skill in an external market as seen through IBM’s internal lens. To achieve this goal, we identified 30+ factors along the following six dimensions: external hiring, job applications, offer acceptance, compensation, voluntary attrition and skills lifetime. These factors are based on the input from various IBM SMEs (Subject Matter Experts) including labor economists, compensation market pricing experts, career & skills consultants, and business units’ representatives.
To build our labeled dataset for model training, we recruited SMEs in Talent Acquisition to annotate the Skills Scarcity data for a collection of Country-JRS pairs. All of these pairs had a continuous and sufficient number of hiring activities over the past two years. Talent acquisition specialists are the people with first-hand experience in hiring for specific countries and regions, and are thus our best resource for providing the view on Skills Scarcity from an external market perspective.
Using this annotated data, we built machine learning models to capture the underlying data characteristics and behavior patterns of the training data associated with scarcity of different levels. Many machine learning algorithms are used to train the models including XGBoost, SVM (Support Vector Machine) and Decision Trees. Rigorous model testings are conducted to ensure good model performance by measuring various performance metrics (Precision, Recall, F-1 score, extreme error rate, etc.), performing 5-fold cross validations, as well as examining model performance variation over different population segments (e.g. by business units, by geo/country, etc.)
Finally, the models are deployed and applied to predict the scarcity level for a given Country-JRS pair, along with a confidence value.
A.2. Use of External Data
Our external partner provides a Hiring Difficulty Score (HDS) that indicates how difficult it is to fill a job position. This score is measured by comparing the hiring difficulty in a designated metropolitan area with the average conditions across the country for similar positions. The HDS ranges in value from 1 to 99, where 1 represents the easiest job to fill and 99 represents the most difficult job to fill.
The external vendor has used many data points to measure the HDS, including the skills demand in the market reflected through job postings, talent supply, compensation, unemployment stats published by governments and organizations.
A.3. Integration of Internal and External Data
Once we have obtained the scarcity data from our internal model and the HDS from external vendor, we fuse them together to achieve the final data by taking into account both the confidence on the quality of each data source, as well as our preference on balancing the internal and external views.
A.4. Business Review of the Skills Scarcity Data
The last step of our Skills Scarcity production process is to have the business units review the scarcity data so as to ensure that the predicted scarcity values of particular skills in specific countries meet their strategic skill priorities.
Based on the review outcome, some data overrides could happen during this step. For instance, we may have predicted the scarcity value of JRS “Software Developer: Cloud” in China to be Medium, yet given that this is a strategic area for our business, the proposed actual value is High.
Such a review process helps improve a unit’s confidence and trust in the quality of the scarcity data, which ultimately leads to its adoption by additional business units to assist decision-making in many use cases. In doing this, we keep human in the loop with AI models.
B. What Are the Use Cases of Skills Scarcity?
Skills Scarcity can be used to support many use cases. Two examples related to employee skill development and compensation investment are briefly described below.
B.1. Employee Skill Development
Leveraging our Skills Value Framework, as introduced earlier, which combines both Skills Scarcity and our internal demand view on roles/skills, we can draw many insights from the data. For instance:
· If a role is growing in demand and has high scarcity in the market, this means that we expect to need more people performing in this role and that it is very challenging to find skilled resources from external marketplace. As a result, we may want to consider proactively training existing employees for these roles as opposed to searching for external talent that may be very challenging to find and costly to recruit
· If a role is declining in demand and has low scarcity in the market, then it means that there are a lot of people with the skills to fill the role for which no more people are needed. So essentially, we have too many people with this skill. In this case, we want to proactively look at the adjacent skills of existing employees in these roles to determine if and how we have these employees reskill to roles that are higher in demand by our units
B.2. Compensation Investment
Skills Scarcity can also be used to optimize the compensation investment to support a skills-based pay strategy. For instance, business units can leverage Skills Scarcity data to prioritize the retention of employees with High scarcity skills. Moreover, during a company’s salary planning cycle, the scarcity data can be used as one of the factors to determine the salary increase amount at individual employee level.
Conclusion
In this post, I introduced the concept of Skills Scarcity, described a data-driven approach to measure its value, and illustrated two exemplary use cases. As skills are becoming the new currency of today’s world, I would strongly encourage every organization to come up with a holistic view and plan on how to effectively upskill/reskill its employees and align their skill & career development with the organization’s strategic direction.
References
[1] P. Estes, “The Half-Life of Skills”, available at https://hrdailyadvisor.blr.com/2020/03/25/the-half-life-of-skills/, March 25, 2020
[2] B. Heger, “2021 Global Talent Trends Report: The Future of Work”, available at https://www.brianheger.com/2021-global-talent-trends-report-the-future-of-work-mercer/, February 25, 2021
[3] PWC, “Workforce of the Future — The Competing Forces Shaping 2030”, available at https://www.pwc.com/gx/en/services/people-organisation/publications/workforce-of-the-future.html
In Part 2 of this blog, I will talk about the AI trust we built around Skills Scarcity, as well as taking a deeper look into the impact brought by COVID-19.
