HR GO recruits AI in its search for more engaged job candidates

Family businesses often have interesting stories to tell about their origins and UK-based recruitment company HR GO is no exception. On a cold December evening back in 1957, with London engulfed in heavy ‘pea-souper’ fog, a young temporary secretary was permitted by her boss to leave work early in order to make her commute home to Kent.

It was only later that evening that Betty Parkinson realized with horror that, had she caught her usual 4.56pm train from Cannon Street to Dartford, she would have been caught up in the Lewisham rail crash, in which that train collided with another and a bridge collapsed, causing the death of 90 people and injuring over 170 others.

Damaged train lines meant that Parkinson couldn’t return to her job in London, so she found a new one, closer to home, working at a paper mill in Dartford. When the paper mill needed more staff, she set herself up as a recruitment agent for the firm, initially working from home. Her early success apparently fuelled her ambition as an entrepreneur and she went on to found Parkinson Staff Bureau, later known as PSB Recruitment.

Thoroughly modern digital business

These days, the company now known as HR GO is a very different place, albeit with Betty’s son Jack at the helm as chairman and her grandson John leading group IT. It has an annual turnover of £90 million, more than 30 offices and some 300 staff. It’s also well on its way to becoming a thoroughly modern digital business in how it interacts with both clients and with job seekers. In particular, its website attracts some 20,000 visitors per month, who can browse available opportunities and submit applications, making this a key revenue stream for the company.

But converting visitors into applicants is crucial, as marketing director Sharon Wilding explains:

Over the past two or three years, we’ve put a lot of effort and investment into our website. We could see that it was growing nicely, but at around 7%, our conversion rate just wasn’t high enough. We spent a lot of time looking at the analytics, to find out what we could pick up from what Google was telling us about why that might be.

We worked with an SEO agency to examine different options there, looking at people’s journeys through the website - but we couldn’t see anything obvious, nothing really stood out as a problem. We had regular meetings with our web development team at Space Between. It was a problem we kicked around each month, but what we were really looking for were new ideas on how we could address our conversion rate.

It was the team at Space Between, also part of HR GO plc, that came up with that new idea - a two-fold approach based on biometric testing and the use of Ascend, a conversion-rate optimization product from AI company Sentient. As previously covered at diginomica, in the case of lingerie firm Cosabella, Ascend works on the idea of ‘evolutionary algorithms’, helping companies to reach the ‘best’ version of their website, in conversion terms, by testing large numbers of design variants on a live website audience.

Unlike with standard A/B testing, the AI capabilities of Ascend can quickly identify those design variants that work best for visitors and combine them, rapid fire, with other successful variants for further, iterative testing. What’s more, Ascend is able to do this very quickly, ‘cross-breeding’ a very large number of design variants to reach a superior end result.

Biometrics and beyond

But first came the biometric testing: job candidates were invited to Space Between’s offices in order to try out the existing site, while their responses to the challenge of navigating this were monitored through eye-tracking, facial recognition, galvanic skin response (GSR) and heart-rate monitoring.

The results held valuable clues to where the HR GO website could be improved. They suggested, for example, that HR GO users were uneasy with the way they were being forced through the conversion tunnel, as opposed to being gently guided through the process. The biometric testing sessions also indicated that there were a lot of visual clues on the website that were distracting users from the decision to apply for a particular opportunity.

These clues generated many ideas about design changes, according to Space Between’s co-founder and director Luke Frake - but too many to test through standard A/B testing. Sentient Ascend, by contrast, allowed the agency’s team to test individual components and variants to see which changes - and which combinations of changes - were most effective. The technology was thus used to search through almost 1,100 possible design changes and determine the best performers over a four-week trial. Says Frake:

One of the reasons we used Sentient Ascend on this is that we can test our ideas very iteratively, so if we thought people were having trouble understanding the obvious route to reach the next stage in their journey, we could try to support that process with different ideas. Ascend will run through those changes individually and then start combining them together, so we can see the impact we get when tests are combined. For us, that means we can do a lot more, a lot quicker.

Sharon Wilding admits she was surprised by the results. Many of the changes that reaped the best results were pretty subtle, she says. Where the company thought it was being short, snappy and clear, it found that a more gentle approach was more effective.

Just staring at our screens, trying to figure out what might work or not work, we would never have come up with some of combinations suggested. But this has shown that it’s not always what you think looks good to visitors, or what you think will work for them, that really counts.

What really counts, of course, are the results - and HR GO has seen an uplift in conversion rate from 7% to 18% since it has made changes suggested by its testing with Ascend. There’s more work to do, says Wilding, because the company wants to keep pushing at that conversion rate,

...but looking further ahead, we’re starting to think about bigger changes and about testing them out this way based on what we’ve learnt already. Ultimately, we’ve got a clearer understanding now of what our website needs, and more flexibility around the different elements of it - so new features for filtering jobs that are displayed, for example, and for general registration of candidates, both seem like good candidates for testing.

People want and expect to apply for jobs quickly and easily, but for us, it’s just the first step in a more personal conversation we want to have with them. This way, we have the opportunity to get their attention and engage with them early on in their job hunt.