Data in hiring: The recruitment revolution

Data is being used at each stage of the recruitment process, but there is still some way to go before recruiters use analytics to predict performance

Right people, right roles, right time. You would be hard pressed to find an HR professional who would argue against those three aims. And yet recruitment has traditionally been wildly unscientific, based on gut feel and unreliable methods. Unstructured interviewing, used by 92% of organisations according to CIPD research, has a predictive validity rate of 0.31.

Getting recruitment right makes a major difference to organisational success. Boston Consulting Group found that delivering on recruitment leads to a 3.5-times difference in revenue growth between the most and least capable companies. It is the most critical HR process in determining future organisational success, followed by onboarding and retention, and more impactful than both talent management and leadership development.

According to Deloitte’s 2018 Human Capital Trends Report, 85% of HR leaders rate people analytics as important, but only 42% rate their teams as ready to capitalise on this. There remains work to do turning hype into reality. So how exactly are HR leaders adding science to their recruitment processes? And is it working?

“People analytics has really taken off this year and recruitment is a space with lots of activity,” says consultant and former IBM global director of people analytics David Green. “It’s a revolution, with data being used at each stage of the recruitment process.”

According to Tomas Chamorro-Premuzic, professor of business psychology at University College London and Colombia University: “Recruitment lends itself to analytics because you are trying to make a clear-cut decision: do we hire this person or not?”

“Knowledge is power,” believes Adi Koblenz, head of talent acquisition at Ovo Energy. “[Using data] helps us remove a level of ambiguity over recruitment,” she says. “It’s such a competitive market. It’s not just about bringing in talent, but the right talent.” She has two focuses: assessing quality of process and quality of hire.

On quality of process, the team is analysing which managers are better at predicting talent and hiring well. Top hirers could be used to develop others. Data has enabled Ovo to be more strategic about advertising, proving that posting ads for developers on specialist blogs results in a higher quality of hire than general job boards.

Assessing the drivers behind the likelihood of getting hired, including assessing source quality, is something HR analytics practitioner Kevin Moore has also had success with. In a previous role at US media provider Gannett, publisher of USA Today, he worked on a predictive project on hiring drivers, including candidate source. “It turned up a surprise finding showing one source had an extremely low likelihood to be hired,” says Moore. “This source was costing us about $20,000 a year and was supposed to yield high returns due to its extensive posting network.” The contract was subsequently discontinued.

Boosting diversity is another driver – although over-relying on data has its own risks, as Amazon recently discovered when it found its artificial intelligence (AI) recruiting tool was biased against women. Insurance firm Aviva is using data to increase diversity, collecting anonymised data at application, combined with demographic data. “It’s about understanding our candidate market to see how we can drive a different candidate population,” explains UK resourcing lead Rachel Glaze.

At PwC a deep understanding of the talent market is helping drive long-term thinking, informing whether to buy, borrow or build talent, says head of workforce strategy Helen Hopkin. Hiring in the tech space where there are fewer female candidates led PwC to launch an apprenticeship scheme to build its pipeline.

“It’s using data and looking across the recruitment and employee lifecycle,” adds Hopkin. “What tools can help inform decision-making and where is the talent we need?” PwC uses LinkedIn, using a natural language processing tool to spot profiles with the right skills and experience (based on the firm’s ‘The PwC Professional’ desirable attributes).

A link to wider strategy, rather than recruitment playing with analytics in glorious isolation, is critical. “We need to understand the business strategy and tie what we measure to outcomes,” says Green. “We’ve traditionally looked at HR in silos, and people analytics created another vertical. But companies doing it well see it as a horizontal, the touchpoints along the employee journey.” He predicts that organisations will start using network analysis in onboarding, for example.

At technology giant Philips, head of talent intelligence Toby Culshaw leads a team that sits across the business, providing insight into the risk and feasibility of strategy from a people perspective. This could be opening a site, acquisitions or functional expansion.

“The key thing is making decisions at an earlier stage,” says Culshaw. “If we are planning new locations, are they future-proof?” The team did a piece of work on expanding an R&D function in Cambridge, Massachusetts, looking at site feasibility, alternative sites, competitors, the inflow and outflow of talent, and so on. “The output was that we shifted the whole North American head office there. We were part of that chain.”

Heather Whiteman is VP, global head of people strategy, analytics, digital learning and HR operations at GE Digital. She also views people analytics as a horizontal. “We have over 100 million data points to help us understand our talent, including which capabilities we have and what we need,” she says. That ties into targeted L&D opportunities, and knowledge of skills gaps at a granular level to inform hiring, “not just a software engineer, but one with a specific skill”.

Having this wealth of data allows the company to predict what matters in roles. In sales, for example, recruiters know what leads to higher revenues. “We know those skills create more value, so we should go after those people,” says Whiteman.

The team has also built an algorithm that matches employees with other – potentially more suitable – roles. “It can say – based on your skills, experience and interests – how well you fit with your role and other roles,” Whiteman explains. This helps internal moves
and promotions.

When it comes to predicting success, Koblenz is running a “huge” project looking at performance in relation to the aptitude tests and assessments Ovo uses, to help predict performance and test prediction inputs. “In recruitment we often make assumptions about how to measure success: data helps you see if that’s right,” she says. Ovo previously assumed contact centre hires with call centre experience would do better. “But we found people from a retail background were stronger talent. That’s why data is such an important part of what we do.”

Diarmuid Harvey, senior business psychologist at The Chemistry Group, believes CV sifting will soon be replaced by text analytics. He adds that predictive tools have the biggest impact in larger organisations, with more data and variables to tease out. “If an employer is having issues with retention, there is an enormous amount of data gathered through the ATS,” he says. “These can be used to predict how long employees might stay, or how likely they are to leave.”

As Ovo has found, call centres and similar contexts are comparatively easy environments for which to create predictive models. “It’s harder to predict performance for more senior and influential jobs,” says Chamorro-Premuzic. “To do analytics well and benefit from AI, we need to have that level of performance data on leaders. The reality is we don’t.”

One multinational firm’s recruitment leader, who wishes to remain anonymous, agrees. “Show me the analytics that could predict someone could be a CEO and I would start to buy in, but in managerial roles, I don’t think we are there yet,” he says. “For predicting success, I haven’t seen anything that beats a competency-based interview plus psychometric assessment.”

Many of the recruitment leaders HR magazine spoke to stressed the work they are doing is not predictive. “We use intelligence, but I wouldn’t say we do predictive analytics,” says Culshaw. “If we start using those terms, pretending we are doing it when we are not, we will have no credibility with the business.”

In the future though, Chamorro-Premuzic believes there is a goldmine of personal data online that could be used to predict performance. “In an age where people have given away so much personal data, it could be organised to provide predictions based on interests, values, personality and abilities.

“Imagine if there were apps on Twitter, Facebook, Netflix, Spotify – even your Uber rating could be a measure of emotional intelligence. Imagine there was something to bring them together – a person’s digital reputation – and consumers were incentivised to share that data with recruiters. That would be an efficient way to match people to jobs.”

He acknowledges there are ethical and legal hurdles, but asserts it will happen. For now, GE Digital’s Whiteman encourages hiring professionals not yet using data to give it a go.

“You don’t need a big team or sophisticated tools,” she says. “It’s easier than people think. Don’t be scared, just try.”

This piece featured in our What's on the cards for hiring? ebook in partnership with LinkedIn. Read the full supplement, including extra box-outs, here