Olivier Le Moal - Fotolia

HR analytics tools promise edge in war for talent

HR data analytics is said to enhance hiring decisions, improve employee retention, and boost company performance. Fact or fiction?

“Our people are our greatest asset.” This business platitude, if true, makes it all the more surprising that many businesses seem to know so little about who exactly their workers are.

“We still speak with organisations who have many small outposts in different parts of the world, and corporate headquarters has no idea how many people are working for them,” says Gartner principal research analyst Helen Poitevin. “Examples bubble up showing a 15% difference between how many people they thought they had on board and the real number they discovered when they implemented a global HR system.”

Keeping track of these numbers has been the stock in trade for suppliers of human resource management software for years, she says. At the very least, HR software can help companies understand how many people they will need to hire, and their financial exposure as employees retire and draw pensions in the coming years.

But in recent years some suppliers have begun to make bigger claims about their HR software, suggesting that the right kind of data analytics can help boost company performance and guide the organisation on who to hire and how to manage them.

“More big vendors are now interested in forecasting who might be leaving the organisation and why,” Poitevin says. “They also want to predict who will be a good fit in terms of recruiting. You get these predictions because technology is ready.”

In September 2015, researchers from Harvard, Yale and Toronto universities (Danielle Li, Lisa B Kahn and Mitchell Hoffman) showed that people selected by predictive analytics for call centre jobs based on a performance test stayed 15% longer than those selected by people who could override test results. “Our results suggest that firms can improve worker quality by limiting managerial discretion. This is because, when faced with similar applicant pools, managers who exercise more discretion – as measured by their likelihood of overruling job test recommendations – systematically end up with worse hires.”

It is a telling result, but one limited to the type of role that businesses are recruiting for. Poitevin says that analytics is used a lot less in recruiting for senior roles, partly because the existing selection process is more open-ended, with multiple interviews and role playing, and partly because analytics saves the most time and effort when applied at scale. Businesses simply do not hire a CFO all that often.

Unstructured data

Tom Pohlmann, who head values and strategy with consultancy firm Mu Sigma, says that analytics is being used to elevate the strategic standing of the HR department, but adds that it involves a huge trawl of unstructured data, often stored in disparate systems.

“That’s the challenge with HR,” says Pohlmann. “There is some data that is codified and very static. We also get into a lot of unstructured data – in PDFs, resumés, performance reviews and so on – that is not coded; it is open-ended. Some larger retailers are also using internal social networks such as Yammer to gauge employee sentiment.”

Pohlmann adds that such data can be used to predict who is likely to leave an organisation, which can help in devising strategies to reduce employee churn.

For more about HR analytics

Oracle human capital management strategy director Andy Campbell says Oracle Advanced Analytics, which powers Fusion HCM Predictive Analytics tools, can help mine unstructured data in HR.

“If you’re looking to find out why people are leaving the company, you can do a review of why others have left, and analyse unstructured data of their last few appraisals. Was it the opportunity to work on an interesting project? You can also look at satisfaction surveys. Together they can give you an indication of the state of mind of those leaving – and let you do something about it.”

False positives

But Poitevin says that users should be cautious when using these techniques, particularly with suggestions as to who is likely to leave. The problem, she says, is that the process throws up a lot of false positives: people exhibit all the indicators of leavers, but don’t end up leaving. Given this problem, the tools must be applied with extreme caution as it’s difficult for managers to “unknow” the analytics results, and make adjustments to their behaviour.

“People do it so they can say they are using predictive analytics, but as far as use on the ground goes, it is limited by behavioural and ethical considerations,” Poitevin says.


Paget Miles, worldwide leader for HR analytics at IBM, says that use of HR analytics has been limited by access to data scientists who can formulate queries on behalf of HR professionals. He says IBM’s Watson cognitive computing system gets over this limitation by allowing natural language querying of company data, which has been combined with knowledge of HR terminology.

It means that companies are beginning to see unexpected connections in data, Miles says. “One large global business we worked with swept up a range of company data to figure out return on investment from its learning programmes. They found out within minutes that lots of people were leaving after doing a particular course; they also found these individuals were leaving to get paid more at competing firms.”

Smaller suppliers are also creeping into the market to try to help solve some of HR’s more difficult puzzles. For example, TalentPool is using clustering algorithms to help SMEs access graduates who best meet their needs (see case study below) while Saberr is reaching into the minds of prospective candidates and current employees to help figure out which recruits would fit in with existing teams (see second case study below).

Analytics in HR could be arriving at the right time. According to professional services firm Deloitte, the capability gap for virtually every talent issue grew in its 2015 survey. Both business leaders and HR respondents themselves say that HR is either failing or barely making its grades. “At a time when talent is indisputably a CEO-level issue, this should be setting off alarms in every HR organisation,” states the report.

The growing arsenal of analytics tools and predictive techniques could help HR answer that call.

Case study: Talentpool

Not all businesses rely on suppliers to provide analytics for HR problems. Graduate search firm TalentPool has built its own database and is applying analytics techniques to help small and medium-sized enterprises (SMEs) find the right people for their positions.

TalentPool co-founder Andrew Lavelle says that at a time of great competition for the best graduates, it can be hard for SMEs to find graduate recruits. The company has developed clustering technology based on open source tool k-modes to bring together candidates who meet employers’ needs based on statistical similarities with successful candidates.

It is so hands-off, it’s a breath of fresh air
Dan Fruhman, Deliveroo

As the database grows, Lavelle says machine learning will be brought to bear to address the graduate selection problem. “From a technology point of view, the clever bit is yet to come. Once you gather a large data set of 100,000 real candidates with real-world interaction, recommendation engines as used in online dating and by Netflix have a very large potential.”

Companies including Tata, Hiscox and Allianz use the TalentPool matching service. Online food delivery firm Deliveroo has also benefited.

Dan Fruhman, north-west US manager for Deliveroo, says the company had received $200m funding and was struggling to recruit the right graduates to meet its expansion plans to cover 55 cities worldwide.

“With TalentPool, there is no sifting through rubbish, as you do with other online services or recruitment firms. We get a pool of people who are mainly interested in joining hot startups and see the potential to progress. Compared with recruitment firms, it is so hands-off it’s a breath of fresh air.”

Case study: Saberr 

Recruitment analytics firm Saberr was founded with inspiration from Noam Wasserman, professor of business administration at Harvard Business School. After studying 6,000 startups, Wasserman had found that 65% failed due to irreconcilable frictions between the founding team, rather than a faulty product or lack of funding.

Saberr uses output from psychometric tests, together with measures of personal values, to try to predict whether prospective candidates will fit well with existing team members. It is a cloud service which uses an algorithm based on years of research in psychology, together with the candidate’s and team’s test results to arrive at a risk score indicating overlapping values or tolerance for other’s values.

People who scored higher were more successful
Alastair Barber, Capco

Capco, a specialist financial services consultancy, has found Saberr useful in recruiting. Managing principal Alastair Barber says the business was asking candidates to fill in the Saberr online survey and comparing the results with scores from those seen as the “culture carriers” at Capco. They then looked at whether this comparison could be used to predict the outcome of the recruitment process.

“The data pointed to the idea that people who scored higher were more successful. If they resonate with the culture carriers, they got through the recruitment process,” Barber says.

Although Capco has currently halted its licencing of the Saberr tool, Barber says it could prove valuable in taking a stage out of the recruitment process, freeing up valuable time in the business. It could also be used to select the best teams to work on client projects from the internal pool of talent, he says.

Next Steps

Grain firm grows with HR analtyics tools

How possible changes to fiduciary rule could touch HR tech

Read more on Business process management (BPM)