Using advance Tech for predictive analytics in employee retention

Using advance Tech for predictive analytics in employee retention

This technique can help managers reduce attrition costs.

The future of human resources is changing. Like the rest of the business world, chief human resource officers (CHROs) and their teams are beginning to find that they need to focus on building a robust analytics capability to best prepare for the data-driven world.

“CHROs have said that they feel [pressured] as the only ones not bringing data to the table. The business is expecting HR to have similar numbers to marketing, though maybe not finance or operations,” observed Andrew Marritt, CEO of OrganizationView, a people analytics practice based in St Moritz, Switzerland. According to Marritt, the data-centric modern HR leader needs to know not only what has happened, but what is likely to happen.

A key HR concern for businesses is employee retention. There are significant financial and intangible costs associated with losing loyal and high-performing employees. Investments need to be made to find, hire, and train their replacements. There could also be a negative impact on the stakeholders they worked with regularly such as suppliers, colleagues, and customers. Some companies are starting to look to predictive analytics to increase their ability to mitigate the risk of employee turnover and increase retention.

Investment in building a people analytics capability need not be big at first, and businesses can benefit greatly from it. “Our research shows that the financial costs associated with attrition can range anywhere between 13% and 23% of annual compensation depending on the function/level of the employees under the scope of the study. In our experience, a focused attrition analytics predictive model can help lower this risk by 5% to 8% annually,” said Neeraj Tandon, director for workforce analytics and planning, Asia-Pacific, at Willis Towers Watson, in Gurgaon, India.

WHAT’S NEW

Traditional HR analytics are descriptive in nature and examine employee data across different dimensions such as department and demographics to identify past patterns within metrics like turnover and retention. Conclusions are then used to formulate talent policies. Descriptive analytics, however, cannot predict future outcomes at an individual employee level.

Predictive analytics does this by going a step further and using the evidence from descriptive analytics as inputs for advanced techniques like statistical modelling and machine learning. These methodologies provide forward-looking measures such as flight risk, which quantifies the likelihood of an employee’s leaving the organisation within a certain period of time.

Predictive analytics also identifies hidden connections between key factors contributing to employee turnover. The main predictor variables normally studied include pay, promotion, performance reviews, time spent at work, commute distance, and relationship with a manager. (See the chart, “Factors Contributing to Voluntary Turnover”, for a breakdown of key reasons for attrition at a sample organisation.) Organisations also use external data such as labour market indicators and the current economic scenario as causative variables while formulating hypotheses and building models for retention. HR teams and managers use the findings from the modelling to better design timely interventions to help retain employees.

Factors contributing to voluntary turnover

An ADP Research Institute white paper examined the factors leading to voluntary turnover at a sample company. The graphic below breaks down the reasons cited. By collecting and analysing the factors that contribute to turnover, companies can institute policies and procedures to address concerns.

In this example, management may want to focus its retention efforts on industry veterans who have not been with the company for very long or look at implementing more lenient telecommuting rules to ease attrition.

Source: ADP Research Institute white paper, Revelations From Workforce Turnover Study.

Deloitte estimates that about 8% of global businesses leverage predictive analytics for talent management, and the ones that do tend to be larger. According to Brian Kropp, group vice president at Gartner, organisations that develop this capability tend to be in sectors that are intellectual property dependent such as financial services, healthcare, and fast-moving consumer goods. Globally, businesses in all major economies are working towards acquiring this competence.

COST VERSUS BENEFITS

Organisations looking to develop competence in predictive analytics have several options. Consulting organisations offer expertise towards building this capability. For businesses looking to set up internal capabilities for smaller capital outlay, many choose to employ or train in-house data scientists who may turn to inexpensive software such as IBM SPSS or free open-source software known as R for their initial modelling.

External vendors that set up human capital management systems with predictive analytics capabilities are also available at different price points. However, experts warn that internal teams should make sure that the human capital management systems offered integrate with data systems within the organisation. The systems should not overpromise and underdeliver in terms of features and tools, and vendors should provide the guidance to use them insightfully.

DATA-BASED CHALLENGES

According to Bersin by Deloitte, an HR research organisation, setting up clean and accurate data streams is, and will remain, a challenge for people analytics. As the research indicates, most big organisations have five to seven systems of record for their human resources data. This means that information often used in predictive modelling is inaccurate or unavailable, a serious stumbling block.

“As statisticians, we do deploy multiple data treatments to improve the quality of data. However, often data on some important variable are incomplete, and as a result we ignore these variables. Some of these variables could be important to predict the outputs. Hence, it’s important that organisations continuously focus on data quality improvement,” Tandon said.

Companies should run specific data quality programs to make the data fit for modelling. These programs would be of greater effectiveness if they were directed at key variables that predict output variables such as attrition rather than across the entire dataset, he added.

BUILDING A GOOD MODEL

Besides clean, accurate data streams, a few further steps can be taken to ensure that predictive retention models are a robust tool for decision-making. For one, studying the workforce in clusters of employees with similar characteristics and reasons for leaving the organisation is essential for building models that lead to targeted and effective retention strategies, according to Tandon.

Model building also goes through multiple iterations to ensure it fits the data optimally, which includes choosing or eliminating causative variables scientifically, and testing the model on an existing dataset to gauge how accurately it predicts actual outcomes. With the acknowledgement that numbers do not tell the entire story, intuition is also factored into models. “There is a good reason people are intuitive; they have got experience,” explained Marritt on how this contributes to the model’s effectiveness.

However, a degree of inaccuracy is associated with predictive modelling, and this is where HR and managers play an important role. “Data should just be another voice at the table. Decisions have to be made by humans,” said Marritt, on how these tools can influence employees’ working lives. It is always better to roll up the data and use them at an aggregated level such as teams, rather than at an individual level, because the implications of making an incorrect decision are considerable, he added.

Last but not least, as with any new initiative, organisations must recognise that adequate coaching and oversight mechanisms should be in place to help users leverage the technique correctly and thoughtfully. According to Tandon, managers are being trained on the key objectives of developing attrition models and coached on how to use the information to prevent high-performing employees from leaving, without creating a bias against the identified individuals.

Central governing teams (often comprising business and HR team members) monitor and track interventions taken by line managers to reduce attrition risk for employees identified as a high flight risk. This also helps organisations bring some level of consistency in interventions to control attrition, Tandon added.

TARGETED APPROACHES

Once these checks and balances are in place, a data-driven approach that includes predictive analytics is seen to bring greater transparency and balance to decision-making. “There have been instances where decisions were made by those who were the most vocal. This will be harder in a world where data is needed to support decisions,” observed Marritt.

The key causative variables that emerge during modelling will also help organisations craft more effective retention strategies. If commute distance emerges as a major driver, for example, greater efforts can be directed towards options such as remote working. If a limited training budget is available, it can be used to provide inputs for those employee segments that have a high flight risk. While HR and managers have always designed these interventions, a forward-looking, rigorous technique enables them to direct time and money towards these efforts with greater precision and with greater confidence in the outcome.

Furthermore, finding unexpected patterns in the data can help design retention strategies that make strong business sense. Marritt’s team at OrganizationView, for instance, found that high work pressure was a key cause for attrition at a certain financial services organisation. However, it was more so for low to midlevel performers while top performers actually thrived under high pressure and were more likely to leave in its absence. Since high-performer attrition had a greater financial impact, the organisation focused on this rather than overall attrition.

THE NEAR HORIZON

Companies are experiencing a massive change in the data they have about customers, and the same change is coming to what they know about employees, according to Kropp. Organisations that figure this out and get there faster will retain a higher-quality workforce. It will be the single most successful differentiating factor on that front, and a must-have for businesses that cross a thousand employees, he added.

Over the last three years, Gartner has also seen a significant increase in the number of organisations that collect employee data in unconventional ways, such as social media activity, speed of keystrokes, mood recognition, email text and frequency, and wearable microphones. Organisations are attempting to understand employee behaviour and experience through these experiments, and some of them will be input into models, which will increasingly graduate from predicting flight risk and quality of hire, which are relatively easy to measure, to hard-to-define variables such as employee engagement and performance, Kropp said.

On the maturity front, while only a small percentage of organisations surveyed by Deloitte currently have the capability for people analytics, in a more recent survey 69% of businesses say they are integrating data to build a people analytics database. The analytics function will also grow into a multidisciplinary team that will solve business-critical problems to drive business results.

Source : FM UK

Share This