A former Netflix data scientist has used the video-streaming company’s movie recommendation algorithm to develop a predictive system to let employers know when they are about to lose their most talented and productive workers.

Workday Talent Insights is a new app from Human Resources software creator Workday, and aims to use a variety of criteria to predict the departure of hard-won and hard-nurtured staff. It also provides options that companies can consider to retain valuable employees.

The software considers factors such as number of job functions, interval between promotions and length of tenure to estimate ‘corporate restlessness’. It cross-matches this information with job postings at major placement sites such as indeed.com to determine the current supply-and-demand state of the market in relation to that particular worker and his or her function within the company.

WTI was developed, ironically, by former Netflix employee Mohammad Sabah, an acquisition for Workday when his previous employer Identified was acquired by the company last February, and is based on the Netflix movie recommendation engine, which analyses a customer’s viewing history and makes recommendations for other products based on it.

Workday’s senior VP of Technology Products Dan Beck says “It’s not just a predictive model of what could happen next, but adding context, so you can then give something more prescriptive, like a recommendation,”

Opinion With an upturn in the labour market over the last 14 months, it has been a while since projects of this nature have hit the headlines. Interestingly it addresses a current concern about the perceivably unintelligent associations that many algorithms make between what we have done and what we might like to do next.

Amazon pioneered customer history analytics with a recommendation engine widely praised for its relative intelligence and profitability years before Big Data and granular analysis occupied developer zeitgeist. But advertisers seeking to precision-target users with deep analysis are beginning to encounter resistance, as evidenced by a study into the ‘creepiness factor’ of bespoke ads, conducted by a researcher at Ithaca college.

The problem seems to be one of imagination, or the habit of algorithms to look for repetitive patterns in scenarios where the variables (such as the economy, or the circumstances of the individual) change so often that one would arguably need 50-70 years of historical data to cover most of them, and correlate that insight with individual behaviour.