The premise of predictive analytics is many centuries old. ‘Study the past if you would define the future,’ said Chinese philosopher Confucius two-and-a-half millennia ago.
And with the caveat that you can’t predict the unpredictable, Confucius wasn’t wrong.
Today we have access to more information than ever on which to base these predictions, together with the technology to crunch the numbers. The phrases big data, cloud computing and analytics tools have entered everyday vocabulary and trip off the tongues of HR practitioners as readily as they emerge from the mouths of IT professionals. As all the fuss around the GDPR has underlined, data is now integral to HR.
But while the use of dashboards and metrics has become commonplace, for the most part data analysis is still largely backward-looking in the sense that the primary focus is on measuring historic performance rather than planning for the future. How well did we do in hitting targets in period Y? Have we moved in the right direction when set against the earlier period X? And so forth.
Monitoring KPIs and the like is tremendously valuable for showing what organisations, teams and individuals are doing well at and identifying where there is room for improvement. But how much better would it be if data analysis could also provide a reasonably reliable picture of the future and be used to inform the people-related decisions of tomorrow?
While there are some instances of US firms using data to undertake predictive analysis on their workforce, the reality for most HR functions is that this capability is a long way away. According to the CIPD’s senior research advisor for human capital and governance Edward Houghton, there are a number of barriers preventing the function from moving into the predictive space, namely a lack of good-quality historical data with which to get started. This is because people data is often spread across multiple disconnected and unintegrated systems, meaning that different standards and definitions of data exist within the same organisation and even function.
The other major challenge is that many HR functions don’t have the skills and confidence at present to be able to undertake the deep analytics necessary. Recent CIPD research in association with Workday, People Analytics: driving business performance with people data, found that the UK lags behind other parts of the world when it comes to people analytics skills. Just 21% of UK HR professionals said they’re confident using advanced multivariate analysis, such as structural equation modelling, compared to 46% of professionals in south-east Asia.
“This lack of confidence shows in the analytics practice in organisations; even something as fairly basic as people data reporting is not common,” says Houghton. “HR professionals globally still have to get the basics right, before they can start to predict behaviour.”
But even though the majority are yet to make it to the starting line, there are some trailblazers.
Recruitment and retention
Take Microsoft. The software giant has used predictive analytics to try to forecast the likelihood of employees leaving. It bolstered its data science skills by hiring a marketing professional from the telecommunications sector, who was able to bring insights from their experience mapping customer behaviour and build predictive models for employee attrition. It’s a similar story at FTSE 250 gambling company the Rank Group, which has started looking at variables that may provide insight into whether employees are likely to leave or remain with the organisation. “We are very early on in this journey,” says Rank Group human resources director David Balls. “But we are looking at performance reviews, tenure, areas of the business, who their line manager is and so on.”
Taking performance reviews as a starting point, Rank is specifically focusing on employees in the ‘exceeds expectations’ category. Performance data is then married with other information, including a system called Talent Fit from a third-party group of occupational psychologists that matches individuals’ attributes against a blueprint created with the senior leadership team, together with feedback on their employment experience.
“For a top performer with a certain length of tenure in a certain area and working for a certain manager, we are trying to predict the likelihood that they might be turning over in a period of time,” says Balls. “And then we are reaching out to those individuals. If they are a great performer and we are going to lose them then you have the issue of hiring someone as good as them and completing the work that they are already doing, so it is incredibly important we know who these people are and do what we can to retain them.”
For Oliver Britnell, head of global workforce analytics at Experian, if predictive analytics is to have an impact it needs to connect to business issues and influence outcomes. “We have built a predictive capability that not only leverages our core people data, but also wider proprietary or external data and we have used this to drive attrition down by more than 4%, saving more than $12 million,” he says. “We are now at the stage of looking to commercialise this in the external market given the success we have had internally.”
Meanwhile Dominic Hammond, people analytics and insights leader at PwC, believes there is increasing scope to use predictive analytics with respect to quality of hire. In essence, seeking to bring down hiring risks by predicting the suitability of new hires through analysing source, role, profile and past experience.
This all sounds promising. But Microsoft did concede, when interviewed for the CIPD’s Valuing your Talent study, that it hadn’t perfected its approach. This study also found that these types of projects are very experimental, and often have high error rates attached to them.
“The reality is that predictive analytics is for many organisations hard to realise,” says Houghton. “Many simply don’t have the data sets – in terms of both size and quality – to be able to make accurate predictions, particularly regarding complex human behaviours. Using historical data for attrition, for example, has severe limitations; the reasons individuals leave are numerous and various contextual, environmental, social, and economic issues influence this decision.”
Houghton continues: “There are various historical trends that can provide useful insights, but developing predictions is difficult. Instead HR should map scenarios of various outcomes and use these to develop their strategies and operations. This is the level of prediction of possible outcomes that is possible for the majority of HR professionals.”
Reducing bias and increasing diversity
Using this data to recruit employees who will be more likely to stay loyal is just one way predictive analytics is poised to bolster the recruitment process. At Morgan Stanley the focus is on helping reduce biases and increase diversity within the process. “Data-driven recruiting, informed by predictive analytics and then re-tested in strengths-based interviewing, has been fundamental to us making sure that the people we select are the right people for the right roles; that is that they will enjoy their jobs, perform well and stay with the firm,” says Stephanie Ahrens, EMEA head of talent acquisition at Morgan Stanley.
It’s also helping stamp out any risk of candidates not showing their true selves. “Having algorithms analyse and score multiple data points makes it difficult for candidates to over-prepare or [be] fake, both in the blind application and the in-person interview. The organisation gets to see the ‘real’ candidate coming through and unconscious bias is significantly reduced,” Ahrens adds.
Charles Hipps, founder and CEO of Oleeo – the talent intelligence experts working with Morgan Stanley on its recruitment – says the key to stamping out recruitment bias is for HR and data officers to work together to devise a predictive system that meets organisational needs. He suggests that developing predictive tools designed to weed out historic biases may help businesses recruit the best candidates and strengthen themselves through diversity instead of falling prey to old habits.
“Research shows there is a need for this,” says Hipps. “Studies by the Social Mobility Commission have shown that numerous industries are failing to hire talented youngsters from less-advantaged backgrounds because they recruit from a small pool of elite universities and hire those who fit in with the culture; still favouring middle- and higher-income candidates who come from a handful of the country’s top universities.”
He points to further research from Royal Holloway University of London and the University of Birmingham, which found managers will often select candidates for client-facing jobs “who fit the ‘traditional’ image of a role, with many placing as much importance on an individual’s speech, accent, dress and behaviour as on their skills and qualifications”.
However, Roger Philby, founder of The Chemistry Group (an organisation specialising in predicting people performance), says: “to be clear and in contradiction to some of the algorithmic assessment players in the market” algorithms do not remove bias, they merely measure what they are told to measure. Removing bias, he elaborates, requires clearly-defined factors that predict job success and a measurement of these only.
Errors arise when an algorithm blindly reviews the data. For example, your high performers could all be male, leading an algorithm to take this as a ‘fact’ and scale the bias accordingly. This means bringing human insight into the definition of job success measures is critical.
“However,” says Philby, “do I believe that an algorithm can more accurately and reliably predict a person’s future job success than a human? Definitely, yes. If it knows what it is measuring.”
Upskilling and identifying potential
Outside the remits of recruitment and retention other opportunities exist in identifying employee potential. Philby argues that an individual’s potential is merely a propensity to change behaviour in the future. This, he explains, is determined by context; in that in one context an individual may have high potential but in another that same person may offer lower potential.
The key to making accurate predictions about future workplace performance lies in first nailing down a clear definition of what potential is. The Chemistry Group uses intellect, personality and motivation to do so.
“For one of the world’s largest consulting firms, as a business unit was being transformed, we used measurements of ‘potential’ and ‘capability’ to create ‘best job fit’ to roles,” says Philby. “It was critical for the management of the organisation’s talent to predict who not only had the capability today, but also the potential to do a job in the future.”
What else?
The opportunity for HR doesn’t end there. Learning effectiveness could also benefit from predictive analytics. For example, by asking what happens three months after employees take a training course organisations can predict future training outcomes and tweak course structures or investment in training. Then there’s also reward, in particular the perceived value of the respective components of a package.
The new backdrop of the GDPR makes using predictive capabilities more challenging, particularly in areas like recruitment, diversity or promotions, according to Experian’s Britnell. Nevertheless he feels that if built, governed and managed effectively, predictive models will enable HR and businesses more widely to use evidence-based insight to manage their people better.
And with the pace of change, including the shift to cross-functional, agile teams with less traditional organisational hierarchies and functional structures, the need to predict what’s around the corner is more pressing than ever, says global VP of people at AB InBev Ollie Roberts.
“HR sits on the biggest set of untapped data in the organisation,” says PwC’s Hammond. “In a world exploding with potential, the power of data-led insight is revolutionising business decisions. The way organisations embrace people analytics, including predictive capabilities, and the huge opportunities it captures will be of great importance. People analytics allows organisations to understand what’s working and what isn’t. It allows for the better matching of people to jobs and for more efficient and cost-effective recruitment and talent management.”
While predictive analytics clearly has its limitations and its take-up over the next few years is far from clear, there is little doubt that we are going to start seeing more of it in HR. After all, if the function doesn’t look to the future it’s going to end up in the past.