One of the most important changes in business practice over the last century has been the use of numbers and analysis to improve productivity. My (admittedly vague) grasp of industrial history is that the start of this movement came with the notion of piece work and production lines to improve the efficiency of manufacturing – see Taylorism and The Model T-Ford for more details. More recently we have started to apply similar disciplines to drive productivity improvements in the fuzzier areas of business – e.g. sales and marketing effectiveness, return on investment analysis of new IT projects and most recently innovation accounting from Eric Ries and the lean startup movement.
Up to now, however, human resources has largely escaped the disciplines of analysis and measurement. Google is one of the most data oriented companies on earth and also has enough people to draw statistically significant inferences from staff surveys and attrition rates so it is perhaps unsurprising that they are in the vanguard of change in this area.
There was an interesting article on Slate yesterday which described just how they are applying data to improve HR outcomes. Perhaps the best example comes from 2007 when Google’s People Operations team noticed that the attrition rate amongst female employees was too high. Their first step was to drill into the data to better understand what was driving the problem. It turned out that the issue lay with recent monthers who were leaving twice as often as other women. Reasoning that improving their maternity leave plan might help, Google updated its industry standard offer of twevle paid weeks off in California and seven weeks outside to five months paid leave which the mother can split into chunks and take when she wants, including before the birth and when the baby is older. Following the change attrition rates amonst new mothers dropped to the average for all female employees, i.e. a 50% improvement. The change is analysed as a win for Google because the increased costs were matched by recruitment savings and staff happiness increased as measured by Googlegeist, a lengthy annual survey of employees.
More recently they have hired social scientists to run experiments to help them better manage the firm and they now have hard data on the best ways to achieve a range of outcomes including getting people to contribute more to their pension (implore them to save an unreasonably large amount), implementing management hierarchy (good middle managers make people happy), delivering pay increases ($1,000 salary increases are preferred to $2,000 bonuses), to optimising the interview process (four interviews is enough to form a solid view of a candidate).
Most companies have too few people to run HR experiments for themselves, but they can learn from the work that Google is doing. Learning from other people’s experiments is, of course, different from doing your own and the applicability of their conclusions needs to be carefully thought through, but as you have read, there is some useful stuff coming out already. I will keep an eye out for more.