Source & Credits: insights.untapt.com
Fifteen years ago the recruitment market was split into two; permanent recruitment was a “job-led market” and contract was a “candidate-led market”. Today, it is generally a candidate-driven market. The reasons are all known and nothing new: low unemployment, high demand for new skills, the uncertain economic future of Brexit, etc… It means that candidates, particularly the skilled ones, are in control.
There are many that believe the solution is to simply attract more candidates. But as of December 2018, unemployment in the US and UK were 3.9% & 3.7%, with 6.6 million & 861 thousand current job openings respectively. There are practically more jobs than people looking for work. In the financial markets, when demand for an asset outstrips supply, prices go up and the asset becomes a valuable commodity. If there is no expectation for an increase in supply then investors will keep hold of these assets.
Good candidates are a precious commodity today. So it is no surprise that recruitment leaders are looking to new technology to improve the candidate experience, and so hold onto their valuable commodity and build strong relations with them.
Competitive markets drive sophistication as each player seeks an edge. These days, the edge in almost any market is through better use of data. Over the last 2 years, we have witnessed the rise of “Talent Analytics” and “People Science” both as job titles and as a new technological approach. These represent the application of Data Science to the world of people and talent.
What is Data Science?
Data science is a multi-disciplinary field that uses scientific methods, processes, algorithms and systems to extract knowledge and insights from structured and unstructured data.
With the growth of the Digital Economy the world has seen fundamental shifts to using the masses of data created every day to make more informed decisions. An abundance of statistics claim that 2.5 million terabytes of data are generated every single day from 6 billion devices. Data science is the speciality that enables companies to predict outcomes and make decisions based on data and not instinct.
Recruitment agencies are a massive contributor to those 2.5 million terabytes with an ever growing abundance of data: candidate profiles, company information, job specs, interview decisions, hiring decisions and salary information. Making sense of all this data enables an agency to improve the service they give to clients and more importantly, as argued above, the candidate experience.
The challenge with analysing recruitment/talent data:
Recruitment data has proven difficult to analyse. It changes quickly. It suffers from legacy systems and poor record keeping. But this is true of many other data sets. The main reason recruitment data has been resistant to the machines is because these forms of data — comprising mostly resumes and profiles -are hugely complicated and unstructured in differing formats and languages.
Resumes are complex. The standard approach for categorising and filtering candidates and jobs relied on the use of keywords and assigning tags, which is an imprecise and work intensive process. However, new possibilities began to emerge in 2012 with the development of deep learning models. In many areas, these models have driven a revolution. Phones now recognize faces, Google Translate is really translating, and self driving cars are just around the corner. In 2012 for example, Google had no products with the deep learning in them. Today they have over 300 products which reference this approach.
What is Deep Learning?
A deep learning network is composed of several layers of algorithms called artificial neurons, which are inspired by biological brain cells. As illustrated via this layer-by-layer visualisation of the best-known deep learning model (a machine-vision architecture named AlexNet), each successive layer of artificial neurons within a deep learning model enables the model to encode the data input into it in increasingly complex and increasingly abstract ways.
In the same way that deep learning models can represent visual data hierarchically, it is also well-suited to modelling the structure of natural language, whether it be spoken or written language.
As examples:
- Voice recognition and transcription
- Automated conversations with customer-support call centres
- Predicting the sentiment of financial reports
- Suggesting replies to electronic messages, e.g., email
- Translating between languages
- Application of Data Science for improving candidate experience
With the ability to breakdown the natural language of candidate resumes , a data scientist can analyse up to 16 million factors of a candidate’s experience. We are entering a new era where, much like with quantitative trading in financial markets, recruitment companies can now analyse much more data and automate processes.
There are many solutions and insights that data science can deliver to the recruitment industry. At untapt, the interest has been on using data science to improve the candidate experience as well as to provide insight for hiring managers–accelerating the entire process and reducing the workload.
Amongst other approaches, we have been using data science with clients to:
- Drive relevant communication by providing insight on when and what to communicate with prospective candidates
- Develop automated recruitment platforms that facilitate meaningful connections
- Predict future leaders and identify transferable skills
- Analyze workforce to identify skill gaps at a given firm
- Map out talent pools and compare candidates in vector space (see illustration below)
- Identify areas of training for career progression
- Provide analytics on hiring trends and companies
- How to approach Data Science
Applying data science in any industry requires a thoughtful approach. Although there are off-the-shelf solutions, the best results come with appropriate planning. The first step is to decide what changes are needed to the current process to give a better candidate experience. Is it a new way to service them by predicting outcomes or automating operations to offer faster, better-targeted services? Maybe the candidates are happy with the process, but the recruiters themselves are being crushed by the workload.
Teams need to gather relevant data for the project before starting the design phase. An understanding of what data is available will help the design of a bespoke data modelling approach. Typically, we iterate on this model several times, incorporating feedback from the client as well as key outcome metrics. Once a model is ready, it can be delivered into real-time production systems in multiple different ways, ranging from building an entire custom platform and user interface to providing an API (application programming interface) to which your existing system can connect.
Changing the candidate experience
Competition for the best candidates is so high that there needs to be a change in approach for attracting and rediscovering talent. 60 percent of job seekers report a negative candidate experience with the employers they engage. Recruitment doesn’t have to be about making more calls and searching for more candidates. Especially in this age of record low unemployment and knowledge based economies, it isn’t a “numbers game”; it is a relationship game. Sourcing and speaking to more candidates doesn’t guarantee success. The candidate feeling like a valued asset and having beneficial engagement is critical for delivering results. Recruitment agencies that efficiently leverage talent data can differentiate themselves from their peers by maximizing the value of each encounter with a candidate or a client, and data science tools such as deep learning help do that on a mass scale and at high speed.
“Without data, you’re just another person with an opinion.” W. Edwards Deming
Untapt is a Data Science and Technology Company who focus exclusively in the field of Talent and Recruitment. www.untapt.com