Your number one customers are your people. Look after employees first and then customers last.

—Ian Hutchinson

In today’s complex, global, and extremely competitive world, talent is top of mind for most CEOs. While concerned about acquiring and retaining the best and brightest, executives are also worried about nurturing, engaging, and advancing their star performers. As simpler jobs get offshored and the demand for highly skilled individuals continues to rise, competition for the critical remaining pool of talent is fierce. Additionally, millennials bring up another set of challenges by setting new employee preferences for health and environment, purpose and community, and hard work with balanced lifestyles. As a result, the talent puzzle seems to be as multifaceted as it has even been.

Perhaps it is not surprising that retention and attrition may be the biggest challenges of all. In a recent Deloitte survey, over 70 percent of respondents expressed “high” or “very high” levels of concern about retaining critical talent over the next 12 months, along with 66 percent who had the same concern about retaining high-potential talent.¹ Robin Erickson, PhD, Vice President of Talent Acquisition Research at Bersin, reminds us that employee retention is a top concern of most business and HR leaders now that the global economy is improving. “High-performing employees can look for jobs passively just by having their job histories on a social networking site. And, according to Deloitte’s 2014 Global Human Capital survey, 83 percent of the organizations surveyed do not believe they are ready to respond.”

TRADITIONAL APPROACHES ARE FAILING

Many traditional human resources (HR) methods for understanding and improving retention (or attrition, its antonym) are typically reactive and slow to produce results. They cater to a past generation and associated organizational structures where employee movement was slow and laborious. They typically consist of blanket (not individually targeted) retention campaigns aimed at attacking the latest attrition driver identified from someone’s gut feelings acquired over many years of experience. One example is an across-the-board compensation increase for every employee by the same amount, when variations and customizations (by level, department, skills demand, etc.) are typically needed. The focus for these programs also often seems to be on a handful of high-demand top personnel with the most critical skills, with little to no attention given to the rest of the employee base that has the potential to impact retention ratios in a more major way.

Clearly, the aim of traditional approaches can be uncertain and slow in action, so their return on investment (ROI) is extremely difficult to quantify. Exit interviews may contain crucial information, but their confidentiality and low frequency hinder their use in an efficient manner, and they are used only anecdotally. It is, therefore, no surprise that many of these traditional methods are quickly becoming obsolete and ill-suited to an increasingly mobile workforce in which individuals can easily surf the web for job postings tailored to their published skills on social media sites, get e-mail notifications of open positions by their social media network, or respond to direct mobile text messages from various recruitment sources.

WHAT IS RETENTION, ANYWAY?

The inability to calculate an ROI has hindered the understanding of what employee retention truly is. Erickson continues, “Retention has long been believed to be a squishy theoretical concept when it truly is not. Retaining employees requires a combination of real and concrete things that include career growth opportunities, fair compensation, regular performance feedback, learning and development opportunities, focus on employee engagement, and trust in leadership—ultimately, retention can be measured by regretted voluntary turnover, whether your employees stay or go.”

More and more, we have observed sophisticated companies, both big and small, move toward advanced analytics and Big Data to help them understand their retention issues, quantify them, and become more agile in solving them. These companies are seeking analytics not to help them identify a silver lining, but rather to be used as a tool to continuously fine-tune their understanding of retention issues and, almost in real time, to propose nimble and novel solutions to curtail retention. From machine-learning techniques and building algorithms that continuously learn behaviors to the use of Big Data to integrate as many multidimensional facets of the individuals as possible, ROI for these types of HR retention solutions has become cutting-edge and quantifiable.

WHAT YOU NEED AND HOW IT WORKS

The building blocks of retention analytics solutions are similar to those of any other analytics initiative. They consist of:

• Data: This is the first building block of any analytics initiative and needs to contain instances where employees were retained and others who have attrited. This is so the model can be trained to recognize the differences, discern and learn the patterns of each, and be able to make accurate predictions. The more varied the data sources (i.e., capturing every angle of the employee, business, working conditions, economic climate, etc.), the better the model accuracy is likely to become as it has more information to draw upon for its analysis. Of course, data is to be manipulated and used with care, and most data manipulation leading practices are in the public domain today.
• Intelligence: Creating insights from the first building block, intelligence is about extracting, interpreting, and acting on the data. This involves people as well as methods and processes.
  • People are needed to develop the crunchy questions, interpret what the statistics are suggesting, and decide what to do with this knowledge. In the context of retention analytics, various subject matter experts are needed: data wranglers and statisticians, retention specialists and end users, HR team leaders and executives, and other peripheral HR practitioners who can help in designing the appropriate actions to take to avoid attrition by individuals, in addition to crafting these solutions to fit within current everyday work flows.
  • Statistical methods and routines are needed to process, cleanse, and assemble the data, quantify correlations and data signals, and perform validation testing. Leading practices in statistical modeling are in the public domain and need not be of extreme complexity to yield real and meaningful results.
• Technology: This is software to support the data and intelligence building blocks, and can range from basic database management tools to house the data (e.g., Microsoft Office tools, text files, etc.) to the most customized statistical packages for data mining and predictions (e.g., SAS, SPSS, or R).
  • We have noticed that technology can be an accelerator but not necessarily a differentiator. In other words, the sophistication of the technology used need not be of highest complexity to yield real results. However, care should be taken so that whatever technology is used doesn’t create new burdens for the end users.

These blocks come together to build the foundation of retention analytics solutions. Data, which contains mountains of information and insights, needs to be organized and mined for insights by intelligent tools and people. Once those insights are extracted, interpreted, and acknowledged, business strategies need to be carefully identified by HR retention experts and put into action. Only once put into action can these solutions realize value by reducing the retention ratio and deterring employees from leaving.

THE BUSINESS CASE

Beyond the fight for the critical and small pool of talent that is increasingly sought after, there is a real economic driving factor behind this move toward sophistication. Losing an employee means at the very least a transfer of knowledge and experience to, perhaps, a competitor, as well as a real loss of productivity, and a shift, even if momentary, of personnel morale. Beyond the well-known costs of searching for replacements (social media adds, headhunters, etc.), finding a suitable candidate (interviews and trial periods), training (learning curves and training costs), and finally hiring (hiring costs, relocation packages, sign-up bonuses, etc.), turnover can quickly become a large business cost to a company. According to Josh Bersin, principal at Bersin, “the cost of losing one employee is 1.5 to 2.0 times the person’s annual salary. And if you are talking about a senior executive or highly skilled individuals this cost is quickly increasing beyond twofold!” That is because the pool of people who could replace this individual is small. Therefore, more investment needs to be made to find, attract, hire, and train these individuals than in the case of a lower-level position that can easily be filled.

Knowledge Is Power—or Is It?

But identifying who is at risk of leaving and the reasons why is only half the battle. What to do then? Are these reasons always actionable? Will they impact attrition decisions? What if they are not actionable at all?

If the reasons for leaving are specific enough (i.e., at the individual level, recent, and combined with other close factors), they can be influential. If overwork with excessive travel is leading an individual to think about leaving, then a close look at the individual’s career and project assignments with tailored changes (e.g., local project and a rest period) may be enough. If model drivers are pointing to a superstar who draws significant competitor curiosity through social media and recent performance data suggests a beginning of disengagement (e.g., as measured via survey collections), then compensation coupled with highly customized career and leadership development may be worth the investment (e.g., consideration for fast track, project assignment with top management, top leadership training, or shadowing). See Figure 11.1.

But even if the model drivers seem not to be actionable (e.g., an individual who wants a complete change of career path, or one who wants to stay home and start a family), even in these instances knowing these moves before they occur can give the organization some lead time. These organizations are able to anticipate the change and prepare for it more smoothly than if they did not know about it. The about-to-leave individual can, in fact, help to find a suitable replacement, assist in training firsthand, and help avoid the gap in productivity as well as any emotional toll the change may take on the team.

DEPLOYING RETENTION ANALYTICS . . . PIECE OF CAKE?

With the implementation of any business analytics solution, the secret is often in how well it is being deployed rather than how complex and advanced the tools or methods are: “The main pitfall I have seen,” says Houston, “is people and organizations getting extremely excited about this new shiny tool that produces real-time scores and cool visualization dashboards, while not spending enough time to work with the business community and end users to understand how this will help them do their jobs more effectively. Too often, these solutions ultimately are given to folks with not enough thought around best integration, and end up adding to one’s daily workload versus alleviating it. And with today’s fast world of quick apps and everything at our fingertips, we all have even less patience and attention span to tolerate add-on work.”

According to a study from Bersin by Deloitte,² only 5 percent of Fortune 1000 companies use predictive analytics in HR. Even if the concept of predictive modeling is still new in HR, other areas of businesses such as finance, manufacturing, and marketing have embraced it. In fact, these industries have been relying on predictive modeling to proactively reach out to customers at risk, target the most profitable customers, optimize their supply chains, and improve their overall customer relationship management.

Approaches and techniques similar to those used in the aforementioned industries could be applied to help anticipate employee attrition. Predictive models provide HR teams and managers with actionable insights to proactively put in place initiatives and strategies to address company business objectives such as:

• What are your most important talent issues?
• Which employees are at risk of leaving your organization? When and why?
• What is the profile of employees most likely to leave?
• What is the risk to the organization if employees leave?
• What is the current pulse of your employees? And how is it trending?
• What are the main drivers of attrition for your organization? How do they map by various subsets of employees?
• Who are your star performers and critical employees? Which ones are most likely to leave in the next three to six months and why?
• What is the risk/impact of this attrition?

HOW TO IMPLEMENT PROACTIVE TALENT RETENTION MODELS

The key to successful analytics lies in the “implementation” or “deployment”—this is where we have seen some companies succeed while others have failed. The most advanced and complex predictive model will realize little to no value if it ends up sitting on a shelf. Therefore, let’s turn to the ingredients of successful analytics implementation. As discussed in the previous chapters, predictive analytics has proven to provide actionable insights in anticipating outcomes such as who will click, who will vote, who will buy, who will convert, and who will lie. Applied to HR, predictive analytics could help to anticipate who will quit, when, and why. To do this, you will need the data, the quant or data miner, and a statistician or business analyst to build an appropriate attrition model needed to provide you with the key drivers and actionable insights to address your talent attrition.

As a quick overview, predictive models encompass two types of variables in addition to an equation (link function) that bring the two together:

1. Input variables, also called independent variables, are factors that you include in the model in order to test and assess their relationship with the outcome event. Input variables also provide their impact in predicting the outcome.
2. Output variables are the elements you are seeking to predict. In this case, it would be the attrition status: 1 = attrition and 0 = no attrition. These may also be known as dependent or response variables.

The goal of the predictive modeling exercise is to build the quantitative relationship between input and output variables based on past learnings as encompassed in the data. Basic but market-leading statistical assumptions and methodologies are to be used to build this relationship. It is generally enough to use these basic statistical techniques as most of the lift may be realized with simple regressions or decision trees. The iteration process is similar to any statistical exercise in other industries and applications.

Predictive models also tell how strong each variable is in terms of explaining why an employee will leave based on statistical tests. Figure 11.2 illustrates an employee attrition model that can be explained using three major data sources as discussed in Chapter 2: company data, publicly available talent data, and labor market data.

Predictive models applied to the aforementioned data help to determine the statistical impact of every data element, whether it is company human resources information system (HRIS) data, labor market data, or publicly available talent data.

Companies we spoke with during our research for this book mentioned not having access to all the data or having access to just part of it. In the next main section, we will explore what types of variables should be included in employee attrition predictive models in order to assess their relevance and impact.

DATA FOR TALENT ATTRITION PREDICTIVE MODELING

As noted in JP’s interview with Chidambaram from Pfizer, it is important to remember that there is no one size fits all when it comes time to discuss predictive model drivers in HR. Each company has its own issues, from culture to employer and employee value propositions. Today, thanks to Big Data, we have access to more talent data than ever before. For the scope of this book, we will look at four major sources of data to test in the model:

Internal data includes data from HRIS compensation data learning and development training system applicant tracking systems.

Company data includes financial performance, company revenue, growth, number of customers, company brand, company social media scores, company reviews and rankings such as those on Glassdoor and opinion sites, and company online reputation.

Labor market data includes data from a vast array of sources that cover trends such as:

• Gross domestic product (GDP) by industry location and company size.
• Unemployment rate by industry location and company size.
• Cost per hire and turnover rate by industry location and company size.
• Company stock market trend indicator.
• Stock market indicator (e.g., Dow Jones Industrial Average, Standard & Poor’s 500).
• Supply and demand data (is your company in an industry with high demand for its skilled workers such as STEM fields?).
• Online job postings (analysis of all job postings available online).
• Regional and global financial risks.

Publicly available talent data is the most important piece of talent data and is a complete game changer because this data 10 years ago was not available. Open talent data or publicly available data includes the digital footprint that your talent leaves on the web, on social media sites like Twitter, Facebook, or LinkedIn and on niche sites such as GitHub or Stack Overflow.

According to some studies, this publicly available talent data is linked to:

• Profile updates tracking
• Profile photo changes
• Having or not having a profile on social media
• Education level
• Years of experience by occupation
• Social media posts, updates, and mentions
• Followers and number of following influencers
• Groups
• Influencers

In some industries, this external data can explain 50 percent of attrition. This means that half of your employee attrition could be explained by this type of information. Now you see what we mean when we say that this data is a game changer! It holds statistically significant insights and explains voluntary talent attrition.

Table 11.1 provides a summary of the most common variables that could be assessed in your model.

PUTTING YOUR EMPLOYEE ATTRITION FINDINGS TO WORK

It is important to keep in mind that there is also not a one-size-fits-all approach when it comes to addressing churn, so prioritization should happen based on roles such as:

• Revenue-generator roles
• Customer-facing roles
• Mission-critical, non-customer-facing roles

Once you get actionable insights from your predictive models and can identify high-value employees who are at risk of quitting, the reasons why they might leave, and the time frame, one approach we suggest is to leverage strategies used in other industries to prevent this outcome. To optimize your action plan and strategies, the attrition score should be combined with the lifetime value (LTV) of the employee (for more information on these topics, see Chapter 10).

You should calculate each employee’s lifetime value (grouped into high, medium, or low segments, as discussed broadly in Chapter 10) and his or her risk of leaving, also called attrition score (grouped into high, medium, or low attrition segments) using predictive models and suggested recommendations for retention strategies.

In the end this would look like Figure 11.3.

THE SEGMENTATION STRATEGY OF TALENT RETENTION MODEL INSIGHTS

To effectively leverage the predictive insights from your attrition model, we recommend tying the attrition score to the employee’s LTV score. When it comes time for strategic discussions with your business and executive team, leverage the quadrant segmentation approach to help you develop your human capital strategy for each quadrant defined in Figure 11.4.

The resulting quadrant approach with its four segments provides HR and managers with the following five benefits that help them to:

1. Identify employees who are more likely to quit by LTV segment.
2. Identify when they would quit.
3. Identify the patterns that lead to voluntary attrition.
4. Apply those patterns to the employee population to identify employees likely to quit.
5. Develop incentives to retain desired employees and put together succession planning for medium- and low-performing employees.