From ending staff churn to optimising your lunch queue, why numbers are an HR professional’s best friend
Predicting crime before it happens is the stuff of science fiction. At least it was in the Tom Cruise blockbuster Minority Report. Set in the year 2054, it saw spidery robots flanking leather-clad cops as they snatched wannabe murderers from their homes on the say-so of “pre-cogs” who could “see” the future.
But like so many once-dystopian ideas, the concept of previewing crime no longer exists solely on the big screen. Data-mining technology developed on the mean streets of urban California is now bring used in the decidedly less glamorous setting of towns such as Dartford, where software called PredPol tells police officers when and where wrongdoers are likely to strike, visualised as pink boxes on a giant map of the county.
Behind the numbers lies an algorithm busily churning up-to-the-minute data on crimes, offending patterns and lawbreakers’ habits. When a would-be mugger in north Kent is searching out a victim, Kent Police aim to mobilise officers to be in the right place at the right time. During a three-month pilot, levels of street violence fell 6 per cent.
Big data – the mass of information generated by organisations, their customers and employees, and the way it is crunched, cut up and correlated – is decidedly more than a passing fad. With so many businesses reporting genuine insight and cost savings from rethinking the way they use metrics, talk of a big data revolution is more than hyperbole. The CIPD’s forthcoming work with CIMA on human capital metrics (see page 5) is a further acknowledgment of its importance.
No strict definition of “big data” exists, but Nick Prudhoe, director of technology at HR software experts Ceridian, says: “For me, it’s a data set that’s either too large for a single system to process, or too disparate as well.” That suggests expensive complexity (the disappointment of many large, legacy software systems has turned some HR directors off the concept of data) but John McGurk, CIPD learning and talent development adviser, says: “There’s enough power on a desktop PC if you’re prepared to invest in capability.”
Big data has already allowed Tesco to collect the most detailed set of information about customer habits ever assembled. The same principles let retailers such as Netflix and Amazon pre-empt what you might like to buy before you’ve even entered a keystroke. And now it’s HR’s turn to be at the forefront.
A fifth of all HR roles advertised through People Management’s recruitment channels specifically mention the interrogation of data as a requirement. McGurk feels there is considerable scope for people metrics to become as sophisticated and significant within organisations as marketing and customer data.
The early adopters are relatively unsurprising: while Google has moved away from recruitment algorithms, it still works through huge amounts of employee data to optimise its working environment. Formula 1 giant McLaren, meanwhile, has applied the same sort of rigour it uses in racing metrics to managing its staff. But even the more traditional industries are being shaken up.
“Unstructured data is where the gold is,” says Prudhoe. That means taking demographic information and allying it to data from operations, marketing and other functions to do everything from improving recruitment strategies to revealing patterns of sick leave. So what can data help you with?
Find out who’s likely to leave – and how to keep them
Every employer fears an all-out talent exodus in the current climate, where wages remain in recession mode even as economic indices take a modest upturn. Big data can be used to understand which staff might leave, and whether they’re the ones you want to keep, says Andy Campbell, Oracle’s human capital management director.
“Analysing the results of exit interviews might tell you people are leaving because they were offered more money elsewhere,” says Campbell. “But it is a blunt instrument.”
He says qualitative data can offer more insight: “For example, by looking at that person’s previous three performance reviews and comments from their peers, you might be able to gather that they’re a bit hacked off. They weren’t being given certain opportunities or they might have wanted to work on projects but were unable to.” As a result, you can target initiatives (whether pay-related or broader) at the most disillusioned and valued staff members.
Understand why your staff are underperforming
Big data doesn’t just have to exist in a database. Ben Waber, head of consultancy Sociometric Solutions and author of People Analytics, is pioneering the use of “social sensing technology” in the workplace – by watching people at work (through the use of “tagging” equipment) he can glean the sort of insight pure numbers never give you.
Waber already knew from his work at MIT’s Media Lab that where and how people take their breaks has a surprisingly substantial effect on performance. So when he was asked to monitor staff at an online travel agent with concerns over performance, he quickly noted two distinct groups of employees: data sets showed one would consistently relax at tables of four people, even when the make-up of the table varied from day to day. The other group consistently had 12 people.
Analysis confirmed that the people in larger groups were much more productive. This was because they were talking to people in different parts of the company, giving them a broader view of what was going on in the rest of the business and how they fitted in.
“We saw that near one door all the tables had four seats and near the other, all the tables were larger, with 12 seats. People were just sitting at the same tables,” says Waber. “If you’d gone to the senior team and said ‘have you thought about how big your tables are...?’ they would have thought you were crazy.”
Know where employees should be – and why
Proper workforce planning has been saving companies (particularly retailers) big money for years. But the advent of advanced software takes the concept into new realms: Ceridian’s Prudhoe says it’s possible to work out exactly how many staff a business will need and when, even adjusting for sickness absence and other variables.
“One major retailer we work with has 90,000 employees and we have access to the data provided when they swipe in and out of work with their entry cards,” he says. “That’s about eight swipes a day, which is about 40 million transactions a year.”
Such metrics on “hours worked” can be combined with payroll, sales figures, the fixtures for the coming year’s football matches and even weather records to reveal opportunity trends – both around staff (sick leave is more likely in certain weather conditions) and customers.
After crunching its data, one employer in the charity sector found that low engagement was not the main reason a certain number of its staff were calling in sick on Mondays. The data, from surveys and behavioural analysis, revealed that staff were so stressed by an ongoing change management programme that they just couldn’t face coming in after the weekend. So the employer beefed up its training for managers around leading change.
Settle the remote working debate for good
The concept of allowing employees to work from home is acknowledged as best practice for enlightened companies. But judicious use of data can shed some light on where remote working doesn’t work – and what to do about it.
Waber was asked to examine the productivity of remote workers versus office workers in a software business, comparing the amount and quality of work the two groups did as well as the level of email communication.
He found that remote workers took 32 per cent longer to develop computer code than those in an office. “Communication and collaboration, and therefore understanding, were seriously reduced for remote workers,” he says. Email records showed that remote workers sent an average of eight messages to solve a sample problem compared with 38 for their office-bound colleagues, who were consequently able to finish their work sooner.
Waber concluded serendipitous conversations were the missing factor: “Remote workers weren’t bumping into people in the corridor or at lunch. Interactions where someone might say ‘Oh, I should email you about that thing you’re working on’ weren’t happening.” His research doesn’t undermine the case for remote working, but it does point to some of the methods that can make it more effective – for example, using enterprise-wide communication systems to share knowledge, and ensuring remote staff have regular, meaningful face-to-face contact.
Show the holes in your recruitment processes
What do your successful hires have in common? Big data may show much of what you thought you knew about recruitment is based on supposition.
Deloitte used data to examine the effectiveness of a large employer’s recruitment processes to see how they panned out in reality. The result: certain steps in the process had an inverse relationship with good hiring.
“The employer had developed a candidate assessment centre with exercises that scored people according to observed behaviours,” says Laurence Collins, a director in Deloitte’s workforce analytics team. “We found that candidates who scored highly – performing well against the framework developed by the organisation – were actually more likely to be an unsuccessful hire in terms of performance.”
Further investigation revealed that the selection exercises responsible had been developed by a former employee who had used an out-of-date frame of reference. “The world had moved on, new systems had come in and the team needed to be more collaborative,” says Collins.
The downside to data is often portrayed as an over-reliance on numbers at the expense of old-fashioned intuition and communication skills. A less-voiced but more pressing concern is sheer information overload. If you’re getting started on a data project, it’s essential to define your outcomes and ask which questions are most important to have answered before you delve into the metrics.
It’s a mantra Preyal Dewani, head of people strategy and programmes at mobile phone giant O2, says she keeps returning to as the company examines its “employee performance life cycle” to understand when it might see a drop in effectiveness, overlaying age and length of service with information about engagement and performance to identify the best times for intervention strategies.
The processes don’t have to be complicated. When Kevin Hough, group head of resourcing at insurance firm LV=, wanted to target recruitment campaigns more effectively, he asked what was making people work in a certain area – in particular how far they were travelling (and would be prepared to travel). “The data was there but it was sat on a system, so we put it into Geomaps,” says Hough.
This relatively simple overlaying of information graphically demonstrated future areas of opportunity when hiring, as well as which grades of staff were prepared to commute the furthest, helping demonstrate ROI in recruitment advertising and target future campaigns on certain geographic areas – which weren’t necessarily the ones the business had anticipated, says Hough.
Big data can be revelatory in such circumstances, but it has its limits. The dangers of “correlation equals causation” (assuming that because data shows an outcome you also know why) are well-acknowledged. Waber points out Facebook data has shown that people with a high IQ like curly fries: does that imply that eating curly fries will make us smarter, he asks?
These perils are inherent in the trend for employee sentiment analysis, which takes techniques honed in customer interactions on social media into the workplace. By looking at what’s being said (primarily on intranets and other internal communications systems), it can identify patterns of disaffection before they become more serious. Campbell is one of a number of experts who fear it cuts out context and can lead to rash judgments.
And that’s where the all-important human angle comes in. HR, as the gatekeeper for sensible people management, needs to own the use of employee data metrics, says McGurk: “Anyone can learn to do it. You shouldn’t be able to avoid analytics in HR.” It starts, he advises, by asking which issues need to be addressed, before looking at the information that is available to solve them.
“Have critical questions about data integrity. Ask things like: ‘Where is our data? Can we put our hands on it? How standardised is it? How complete is it? And ask people what they think of it.”
Collins points out that HR has been using large-scale data analysis for longer than many practitioners realise – decades, in fact – with the function “quite reliant” on it to manage reward, pensions and executive compensation.
The future is as limitless as the terabytes of data your organisation churns through every year. But though number-crunching may seem daunting, time and again the conclusions big data reaches are pleasingly everyday.
Waber points out that the famous Googleplex campus of the Californian search giant is built on data-driven insights into how to maximise employee performance, right down to the length of the lunch queue – having five people ahead of you is enough to make you start talking to the person next to you but not so many that you will turn away and come back later.
As he points out: “If you talk to someone during lunch, you are more likely to talk to them about work later in the week, which has a positive effect on overall performance.” Sometimes you wonder whether big data isn’t much more than common sense…
----
Case studies and workshops on big data and HR analytics feature at the CIPD Annual Conference. cipd.co.uk/cande/annual
The CIPD HR Analytics Conference will take place in February 2014 cipd.co.uk/cande/hr-analytics