09 Aug Is the recruitment industry set for a big data revolution?
Sports teams have realised the potential of taking a data-driven approach to recruitment, but with staffing taking up a large chunk of firms’ costs, is corporate recruiting ripe for the same transformation?
LinkedIn’s algorithms use recruiters’ established preferences to suggest candidates for jobs. Photograph: David Loh/Reuters
By John Burn-Murdoch – theguardian.com
Midway through his documentary The Outside View, Rob Symes asks Kevin Roberts, CEO of Saatchi & Saatchi, how the advertising giant manages recruitment. “Our employee turnover rate is 30%”, comes the proud response.
But should such a slash and burn approach to staffing be aspired to? Depending on the industry you look at, staff costs account for between roughly 30% and 80% of an organisations outgoings.
With hiring, firing and training taking up significant portions of this cost bracket, there are evidently considerable savings to be made by playing the long game when it comes to recruitment.
The data-driven approach to recruitment is taking off in sport, with Michael Lewis’ bestselling book Moneyball either the catalyst or the commentary, depending on your viewpoint. Anybody familiar with the use of analytics in football will have noticed the rise this summer of the “using data to find transfer targets” theme.
But in the broader business world, adoption of these methods has to date been limited to a handful of examples, most of which have come from the usual suspects – Silicon Valley giants such as Facebook and LinkedIn.
The HR team needs a statistician
Symes, a recruitment consultant for Campbell Black and the man at the heart of the aforementioned film, believes the primary obstacle for businesses to overcome is a skills gap in human resources.
“HR people – traditionally – are brilliant for legal, they’re brilliant for motivating people, but they’re not numbers orientated. Only 10% of HR professionals at FTSE 100 firms have a degree in a subject involving statistics – maths, economics, sciences”, says Symes.
This lack of analytical capabilities can make the hiring process overly subjective, and another lesson modern businesses may be able to take from sport is the concept of the aggregation of marginal gains, most recently popularised by cycling mastermind Sir Dave Brailsford.
Aggregation of marginal gains
It may not seem like rocket science, but a natural tendency – even among scientists – for allowing prejudices to hold sway over empirical evidence to the contrary, makes this concept easier to blithely reference than to adopt.
“Michael Lewis did make Moneyball a bit hyperbolic – data doesn’t solve every problem. But what it does is it gives you an advantage over any of your competitors who are working with untested theories and anecdotal evidence. Even if it’s a 2%, 3% advantage, it’s worth having”, says Symes.
It’s all very well pointing to data as a silver bullet, but it’s not for want of trying that many businesses have stuck with the traditional approach to recruitment. There is scarce little data for the forward-thinking recruiter to use, and where it does exist there are few, if any established methods for extracting insights that can be fed back into a firm’s hiring strategy.
“Once people are aware that there is a data gap in recruiting, the next step is collecting data. Existing systems have various data points – what time a member of staff arrives and leaves, how they are progressing towards previously established and measurable goals – but these are very rarely joined up.
“This is where big data comes in. You need software, algorithms, and – at the real cutting edge – machine learning – to spit out significant patterns that can then be used to make decisions”, says Symes.
At this point, of course, we have the human/machine interface, but this is where the need for numerate HR teams is at its clearest. In order to be able to fully understand what the data is showing and to fully trust it, while at the same time being able to put any insights into the context of real world recruitment, this dual skillset – HR and a working knowledge of statistics – should be an absolute minimum.
Directly monitoring the workforce
“The three functions of HR are to make companies healthier, wealthier and happier. Healthy and happy employees will work better, and a more productive workforce will make a company wealthier.
“A data-driven HR department would use data to test different ways of achieving these three aims. To use an extreme example, a company may offer its staff the opportunity to use devices that monitor blood sugar, mental alertness and so on. It may sound a bit science fiction right now, but when you think about it as an employer, the health and wellbeing of your workforce during the working day is going to impact upon your bottom line”, says Symes.
Monitoring and optimising employees’ health and wellbeing could yield quick gains, but these are examples of retrospective techniques. If the base workforce isn’t up to scratch, improving productivity can only achieve so much.
The real challenge is being able to use data to create an image of the ideal employee for a given role, and then to carry out predictive analytics to reveal who from a pool of candidates should be appointed in order to maximise the chance of them developing into a similarly valuable asset.
“Any aspect of performance that you think is important should be measured. If we can find the key data points that indicate leadership qualities, then we can recruit leaders.
“The traditional recruitment model is you recruit an employee, and they either sink or swim. What I’m interested in is using data to define the elite portion of your workforce, and then training your recruiters, or your whole recruitment system, to seek out those same characteristics”, says Symes.
The problem with psychometric tests
An obvious solution springs to mind – psychometric tests. But top recruitment consultants are already well are of a number of weaknesses to the data these methods yield.
“I interviewed 25 CEOs in Silicon Valley and the UK technology sector, and asked them how important were psychometrics to them in terms of making a recruitment decision. 10 was important, 0 was unimportant, and they came back with an average of 1.5. The reason? You can game psychometrics”, says Symes.
This presents a new challenge for the data-driven approach, but psychometric tests and the pitfalls that accompany them are not new, and there are ways to make employee performance data more reliable.
“If the data is collected over a long enough period of time, it becomes very difficult – albeit not impossible – to fool the system in the same way. Another crucial step is to measure a whole class of information. An interview is just one snapshot, and can in many cases be gamed in much the same way as a psychometric test”, says Symes.
But to many, the idea of having your diet, exercise regime, productivity and efficiency at various tasks actively monitored by your employer will be a disturbing one.
Companies already monitor a variety of feeds including CCTV, email metadata and file transfers in order to protect their intellectual property, but carrying out the same – or more invasive – practices to decide who stays in their job and who doesn’t has rather a different feel to it.
“The ‘big brother’ element is certainly a very interesting and very important consideration. One of the pilots we’re running at the moment has caused some concerns among employees, but I would absolutely want to know if there were elements of my professional life that I could improve”, says Symes.
That might all be very well for Symes, but I wager his is not the majority view. The marketing industry has already learned to its cost what happens when you let the data-driven approach run wild, not stopping to ask about intrusivity and privacy. A safer, albeit more challenging approach is to extract more value from existing data.
LinkedIn: let the data do the work
One early success story is that of LinkedIn. Doubtless many of you reading this article will be all too familiar with having scores of “people you may know” forced upon you like a rogues gallery of marketing types upon logging in to accept a new connection, but of substantially more interest to recruiters is its new function: “people you may want to hire”.
As George Anders writes over at Forbes, “LinkedIn has created algorithms that might do the sorting even more nimbly. The result: a digital cheat-sheet for recruiters”.
The tool, available to its most generous customers (an annual subscription for an individual costs over £3,500) uses a company’s existing candidate shortlists to search for other potential hires with similar characteristics.
As a customer interacts with these suggestions, LinkedIn’s algorithms learn the recruiter’s preferences, and can filter future recommendations to better suit their perceived priorities when selecting a candidate.
Perhaps the same shift from targeted advertising to a social media analytics centred approach that we are beginning to see in marketing, will be replicated in recruitment before an HR analytics project becomes the new Target pregnancy story.
Do you share Symes’ view that having your professional life closely monitored will make you a better employee, or is this intrusive approach the wrong way to close the recruitment data gap? Have your say below or join in the debate on Twitter either with me directly @jburnmurdoch or our official account @Guardiandata