How Bulb swapped game-based assessments for the real job and lifted post-hire performance by 20%.

Bulb’s contact centre hiring had a problem most companies never discover, because most companies never check. When the team analysed their game-based assessment tool against actual outcomes, they found the scores weren’t just uncorrelated with job performance. They were inversely correlated. The better a candidate performed in the abstract puzzle games, the worse they tended to perform in the actual job. The tool at the front of the funnel was systematically favouring the wrong people.

The damage compounded downstream. Hiring managers had worked out that the scores didn’t mean anything, so they ignored them and asked for CVs instead, which put recruiters back to manually screening applications the assessment was supposed to have sifted. Time to hire stretched. And in a high-churn contact centre environment, where every leaver means a backfill, the recruitment team was working twice as hard to stand still. Some candidates enjoyed playing games in a hiring process; others found it alienating, and felt a puzzle game had nothing to do with whether they could help a customer on the phone. On that point, the data agreed with them.

The brief to ThriveMap was direct. Replace the games with an assessment that actually predicts performance in the role: a job demanding clear communication, attention to detail, and real problem-solving, none of which a game measures in context.

The assessment was designed from a site visit and a workshop with the people who knew the role first-hand: energy specialists, pod leaders, recruiters, and Bulb’s talent and diversity specialist, deliberately drawing on a range of perspectives. That work produced an Ideal Candidate Profile covering three things: capability (can they do the job: communicating clearly, resolving problems, making the right choices), commitment (do they want the job), and environment (is the reality of the role — fixed schedules, strict targets, hours on the phone — right for them).

From that profile, ThriveMap built a Realistic Job Assessment: a day in the life of the contact centre role, scoring candidates on the behaviours the job actually runs on. The assessment took over the sift entirely, removing the manual CV screening burden, and hiring managers got candidate analytics showing each recruit’s strengths and weaknesses, so onboarding started informed.

The headline result was the one the game-based tool had been moving in the wrong direction: post-hire performance improved by 20% across cohorts assessed by ThriveMap, with better-aligned assessment translating into lower attrition. Recruiter time moved the same way. Phone screen time halved, and with the manual sift gone, time to hire fell 85%, letting the team fill critical positions faster and spend the reclaimed hours with the candidates most likely to succeed.

The unusual part of this story is what happened after go-live. Bulb asked ThriveMap to check whether the assessment was still aligned with the vetting happening at their assessment centres, and ThriveMap ran the statistical analysis on its own product. The validation confirmed the core: candidates scoring in the top band were significantly more likely to be hired at assessment centre than average scorers, and validation checks showed high scorers were five times more likely to be rated Good or Excellent once in the job.

But the analysis didn’t stop at self-congratulation. It also flagged that a small number of the assessment’s own questions showed no clear relationship with outcomes, and recommended reviewing them, with removal on the table if more data confirmed it. It flagged the same about parts of Bulb’s assessment centre, identifying which exercises drove decisions and which added little.

This is the point Bulb’s story makes better than any other: an assessment is not a finished product, it’s a prediction that should be tested against reality and refined. Game-based tools had failed Bulb precisely because nobody had checked whether the scores meant anything. ThriveMap’s assessments are built to be checked — validated against real hiring and performance data over time, with the willingness to cut what isn’t working, including its own content.