Employee feedback provider Culture Amp has claimed to be able to offer insights to employers into who among their workforce is likely to hand in their resignation, as well as when and why, based on a collation of workplace data from some 2.5 million people worldwide.
“A huge part of our mission at Culture Amp has been making people science accessible to more people, and inside more companies, than anyone else. Focusing on scale and accessibility has meant that we have an incredible amount of data to draw upon,” said Didier Elzinga (pictured), the company’s CEO and co-founder.
“This, in turn, is allowing us to give our clients the ability to peer around the corner so they are then able to predict outcomes and obstacles based on data and experience, and take the required action to achieve better results.”
Chief scientist Jason McPherson said that, traditionally, employers rely on an exit interview, which often does not provide the full breadth of reasons why an employee leaves, and of course, comes too late for the company to take action to retain them.
“Beyond the exit interview, little is known or documented around why an employee or groups of employees are choosing to leave an organisation at various stages of their tenure,” he said.
“Our new dashboard provides a predictive forecast of a company’s employee turnover. Our proprietary algorithms sense when an employee is going to leave an organisation and uncover the reasons why.”
How does the technology work?
In a blog post on the company’s website, attributed to product marketing manager Toby Roger, Culture Amp said that its machine learning process has been fed a range of data sets, which then calculates the most likely scenarios.
“There are some signals to predict employee turnover that are well known and as you might expect. For example, answers to the employee engagement survey question ‘I see myself at [company] in two years time’,” it said.
“Subtler signals — like shifting patterns over several engagement surveys — act together in a more sophisticated way to indicate whether people may leave. The proprietary algorithm takes into account a range of obvious and subtle signals to sense turnover risks.”
According to the post, British company Auto Trader was able to reduce staff turnover by 9 per cent using the technology.
Culture Amp has been asked to outline the accuracy of its predictions, and whether it is transferable across national boundaries and businesses of different size.
How accurate are the predictions?
My Business was curious about the accuracy of these predictions, and whether they are applicable across all countries and businesses of all size, and so put these questions to Culture Amp.
“Our underlying machine learning technology predicts not just if someone is going to stay or leave, but rather what probability they have of leaving within a specific period of time,” a spokesperson replied.
“When we performed a retrospective analysis of over half a million people who left six months after completing an engagement survey, we saw that the turnover risk is around 80 per cent accurate in identifying who is likely to leave and when. These insights are shared on a group level and are one component of how we can provide turnover insights.”
Do the results differ between different countries?
“If a company has a region or location recorded for each of their employees, these groups of employees can be identified and potentially surfaced through the Culture Amp platform,” the spokesperson said.
“We make specific predictions for geographies where they are flagged to us. We find other factors more predictive of risk than geography, but we adjust for it by considering the historical trend in a company and the regions it operates in.”
Does it maintain the same level of accuracy for small businesses as it does for large corporations that have more data to input?
According to the software provider, “company size is an important indicator of employee turnover”.
“We handle this in two ways, firstly within the model there is a company size effect. Analysis reveals that small companies are three times more likely to experience employee turnover than large companies.”
“Secondly, we anchor the turnover predictions to the recent history of turnover within each company. This ensures that the predictions for each company are in scale to their particular circumstances.”