HR by the Algorithm: Playing by the Numbers






One of the hottest areas in our run-for-the-numbers race right now is predictive analytics – algorithmically-based formulas that might predict such important things as who will be a top performer or who is likely to leave the organization.

Rather than looking for a crystal ball, the intent in these algorithms is to use mathematics (referred to as “science” by the vendors) to provide HR with the data points for better decision making. Let’s look at one popular example today: predicting employee voluntary attrition.  This feature is increasingly prevalent now for two reasons: one, attrition is important to organizations worldwide in a tight market for skills, and two, it is generally easier to “predict” based on the kind of data that HR generally has available: time in one job, duration without promotion, reputation of the manager with his or her subordinates historically, length and time of the commute to work, and the like. 

Just this morning I listened to one solution provider, Ultimate Software, talk about its retention predictor and upcoming high-performance predictor. Likely these can be valuable tools for many. But maybe we should play out conceivable outcomes of the predictive philosophy.

Let’s look at two possible though perhaps unlikely ramifications for the “Predictive Economy” as it applies to the workplace: misunderstanding the algorithms on which these predictions are made and the replacement of people in the decision-making process through over-reliance on numbers.

Solutions that predict behavior, success, retention or leadership abilities are by definition fairly complex: they must crunch data from many sources to ascertain a conclusion such as “Sally is a likely high potential performer.”  An HR professional or indeed, Sally’s manager, will be delighted to see this fact highlighted in Sally’s profile – but will either stop to consider on what basis the conclusion was made? Will anyone in an organization take to time to figure out how a software-provided algorithm works and evaluate the validity of the data used by the program to support such a conclusion within his or her specific environment?  If a formula is a “predict the difference between a C and an A player 75% of the time,” is it meaningful to you? And are you comfortable with looking at a computer-derived conclusion and making an employee decision on it? 

But let’s say the algorithms are faultless: we can predict to a one who will succeed, who is a good team player, who will leave the organization, and who will be a great leader. We can replace the “people part” of HR and training because we will be able to calculate good hires, automatically assign appropriate learning to them because we know what they will excel at, algorithmically assign them to teams because we know their social behaviors, automatically plot out promotion paths and salary increases – and eliminate the decisions and indecision that often surrounds the employee acquisition and optimization processes.  Human intervention not needed.

Now, not to sound overly curmudgeonly, I do like scientifically-derived data about almost everything—especially people. But as we move into the hyper-analytic movement in human capital management, let’s just make sure we really know what we looking at in those algorithmically created conclusions that can affect our employees lives.



This publication contains general information only and Deloitte is not, by means of this publication, rendering accounting, business, financial, investment, legal, tax, or other professional advice or services. This publication is not a substitute for such professional advice or services, nor should it be used as a basis for any decision or action that may affect your business. Before making any decision or taking any action that may affect your business, you should consult a qualified professional advisor.

Deloitte shall not be responsible for any loss sustained by any person who relies on this publication.

As used in this document, "Deloitte" means Deloitte Consulting LLP, a subsidiary of Deloitte LLP. Please see for a detailed description of the legal structure of Deloitte LLP and its subsidiaries. Certain services may not be available to attest clients under the rules and regulations of public accounting.

Copyright © 2014 Deloitte Development LLC. All rights reserved.


Christa Degnan Manning

Vice President, Solution Provider Research Leader / Bersin, Deloitte Consulting LLP

Christa leads technology and service provider research for Bersin, Deloitte Consulting LLP. In this role, she helps businesses align their workforce support strategies with appropriate third-party software developers, service partners, and governance models. A 25-year technology industry veteran, Christa’s expertise assists organizations in creating functional capabilities and employee experiences that increase productivity, engagement, and workforce efficiency. Frequently cited by business and trade media, she has presented market research on business to business trends, leading practices, and expected challenges and benefits at industry and user conferences globally throughout her career. Christa has a bachelor of arts in English from Barnard College, Columbia University, incorporating studies at University College, University of London. She also holds a master of arts degree in English from the University of Massachusetts and has completed executive coursework on business metrics at the Wharton School, University of Pennsylvania.

2 thoughts on “HR by the Algorithm: Playing by the Numbers

  1. Great blog, Katherine. As the Principal Data Scientist at Ultimate Software, I think you are 100% correct in suggesting that there has to be trust between the predictors and those using them. At Ultimate, we do share with our customers what makes up the scores, but we also strongly encourage our customers to not get too wrapped up in it. The goal is not to "raise scores" – it is to alter the day-to-day relationship between the manager and the employee so that the employee, manager, and company all benefit.

    Predictive analytics are merely predictions, which can be right or wrong, all with some degree of accuracy. (and ours are quite good!). But no matter how good a manager thinks he/she is, the underlying characteristics found in data will always be able to provide additional insight that the manager did not expect. These data points can alert a manager, especially one that is more focused on themselves and their own daily tasks, to turn their attention to a specific employee in a specific way.

    But it is also equally important to understand that nobody’s predictive analytics will never contain "all" of the data about a person. So I do not believe it will ever be possible, or applicable, to ignore the "people" side.

    For example, an employee may miss a great deal of time dealing with a sick parent. If "number of weekly hours worked" is a contributing attribute, then that employee may receive a "likely to leave" prediction, which wouldn’t be true. So a good manager would take no action at all. A bad manager would apply some generic action that could actually cause the employee to leave.

    But the vast majority of cases do not have something special like this. So the predictors can make a huge difference. Our customers are taking advantage of the predictors, and they are seeing substantial ROI in their bottom lines.

    So the best guidance I can give is to let the data be the driver to identify the people of interest, but then utilize both the data and the "people" side to take appropriate action/non-action. This is the easiest path to successfully leveraging predictive analytics.

Leave a Reply