Actionable Predictive Analytics and Implementation — the Next Big Thing in Talent Management

Every leader wants to know what is the “next big thing” in talent management? Well in my book, it is the forward-looking talent management approach known as predictive analytics. If you are unfamiliar with the term, predictive analytics are simply a set of decision-making metrics or statistics that alert or warn decision-makers about upcoming problems and opportunities in talent areas like recruiting and retention. Predictive analytics are clearly superior to traditional HR metrics, which simply tell you what happened last year.

What happened last year is unlikely to be an accurate indicator of what will likely happen this or next year. For example, last year with high unemployment rates and a weak economy, turnover rates were low. But it would be a fatal assumption to assume that those low turnover rates would continue in an improving economy.

Without Action, Big Data Is a Big Waste

But even predictive analytics have limitations, because it turns out that providing decision-makers with large volumes of data and information does not automatically result in better talent management decisions. Even predictive analytics can be labeled as “so-what metrics” because they don’t excite or alarm the reader. Consider “actionable predictive analytics” which add several factors (i.e. cost and recommended action factors) that increase the likelihood that decision-makers will take some action after reviewing the analytics. Remember that that is the goal, to increase the speed and quality of talent management decision-making as a result of providing the right amount of information, in the right format at the right time. Perhaps an example illustrating the difference between the three different categories of metrics would be appropriate:

Historical metric — last year’s corporate turnover rate was 8%.
Predictive analytic – “as a result of a drop in the regional unemployment rate, there is an 86% chance that the turnover rate in this job family will dramatically increase from last year’s 8% up to 12% within the next six months and up to 16% within 10 months.”
Actionable Predictive Analytic — an actionable analytic adds a cost element to the standard predictive analytic “We project that this 100% increase in turnover will reduce your group’s productivity over the next 10 months by 17% resulting in a reduced output value of $812,000.” It also adds a “recommended action” component “the recommended action is to implement personalized retention plans for the top performing 20% in this job family; they cost $2,000 each to develop and have a 89% success rate.

The Top 10 Benefits of Using “Actionable Predictive Analytics” in HR

Traditional HR metrics are overly simplistic in that they merely report what happened last year. Much like telling you who won the Super Bowl last year, historical metrics don’t add as much value as telling you six months in advance who will likely win the Super Bowl this year. Analytics are superior because they analyze past and current data and reveal patterns and trends. If you are trying to sell your leadership on switching to analytics, below you’ll find a list of the top 10 factors that make “Actionable Predictive Analytics” superior.

  • Time to prepare — predictive analytics tell you what is about to happen, so that you have time to prepare a plan to mitigate the damages or even avoid the upcoming problem altogether. On a positive note, they can also warn decision-makers about upcoming talent opportunities, like an upcoming period of reduced talent competition in the recruiting marketplace.
  • An opportunity to be strategic – part of the very definition of being strategic is to be forward-looking, so implementing predictive analytics can demonstrate to executives that you are acting strategically by forecasting, assessing risk, and preparing for the future. And because other business functions have been using them so long, joining them will make you appear more businesslike and building internal partnerships with already successful users can make implementation much easier.
  • Become aware of shifts in historical patterns – predictive analytics let you know which of the past historical patterns will remain steady or will no longer hold true for the future. This information allows decision-makers to shift their focus toward those changing patterns.
  • Understanding why those shifts are happening – the best predictive analytics identify not just which factors are changing but also why they are changing, so decision-makers can implement solutions that best fit the root cause of the problem.
  • Becoming aware of new relationships in talent management — predictive analytics can reveal new relationships between talent factors that did not exist in the past. For example a policy shift requiring part-timers to work under 30 hours per week may have unintended consequences in that it could now significantly hamper recruiting and increase turnover problems in jobs that had no past history of recruiting or retention problems.
  • Increasing the odds that decision-makers will act – because actionable predictive analytics “advise” decision makers on the estimated costs of upcoming problems and which actions have the highest probability of solving the predicted problems, decision makers are more likely to actually take action when they are learned about an upcoming problem.
  • “Time to the answer” may be reduced – making decisions fast is also central. Because most predictive analytics approaches are more integrated and comprehensive than traditional HR metrics, the time that it takes to get an answer to a decision maker’s query may be reduced dramatically.
  • They can allow you to model – advanced analytics processes allow decision-makers to develop a model which allows them to try out different alternatives and to vary the constraints and the assumptions in order to see how the results would change. Such models or if-then scenarios can excite decision makers, while at the same time helping to avoid major errors through pre-testing.
  • Provide a competitive advantage – if your talent competitors don’t develop predictive analytics, the predictions provided to your decision-makers can provide your firm with significant talent and business advantages.
  • They can cover all critical talent areas — predictive metrics can provide heads-up alerts in all of the important talent areas including recruiting, retention, leadership, performance management, and internal movement.

Predictive Analytics Are More Common Outside of Talent Management

Predictive analytics may be new to you because they are in fact relatively rare within talent management, but they have been around for decades in the business world.

The most common example is weather prediction. Predictive analytics allow businesses and farms impacted by weather to prepare for upcoming weather events. Hurricane predictions for example have become amazingly accurate as a result of the use of “big data” and statistical approaches which predict upcoming storms. Predictive policing is becoming more common where analytics help police departments know in advance where and when crimes are most likely to occur in their city. Predictive analytics have recently become extremely popular in the area of consumer behavior, where they have been used to predict future shopping behavior and changing patterns. The insurance industry gets the nod for the longest history of use with predictive analytics they have used to identify patterns of illnesses and accidents.

Within talent management, Google has excelled, producing predictive analytics in hiring, leadership, and retention. In the retention area, Google learned to use a combination of seven different factors to predict which employees were most likely to leave (in some cases, before the employee actually realized it themselves). In other cases, Sprint used analytics to predict which new hires were likely to quit and Cisco once used predictive metrics to identify which struggling new hires were likely to succeed over the long term.

Note: In next week’s follow-up article on ERE.net, “Implementing Actionable Predictive Analytics In Talent Management,” I will describe the components of high-impact “actionable predictive analytics” that encourage managers to act on upcoming talent management problems and outline the functional areas of talent management that predictive analytics can cover.

Implementing Actionable Predictive Analytics in Talent Management

I’ve been espousing the need for predictive metrics in HR for over 20 years, so I am pleased that more talent leaders are now beginning to realize their value. Unfortunately, most of what is written on the subject tends to be very general and instead what is really needed are some how-tos and some implementation tips.

In my first article on the subject, I covered the benefits of predictive metrics and the need to add actionable components, so that predictive metrics drive action and actually improve people-management decisions. In this article I will outline those actionable components and highlight the specific areas where you might need predictive metrics.

The First Essential Step — Including the Elements That Make Predictive Analytics Actionable

In my experience, most existing predictive analytics in talent management (including those that come from software providers) come up short because they don’t drive decision-makers to actually take action as a result of your predictions.

In order to have an impact, predictive analytics need to influence the “alerted” individual to take the appropriate action. And unless the analytic is designed correctly, the odds of it driving action can unfortunately be quite low. Instead, I recommend “actionable predictive analytics,” which are designed to provide additional information that drives managers to actually act, so that an upcoming problem is minimized. I sometimes call it an OMG analytic or metric because it is designed to get a managers immediate attention (Oh My God) and to drive action.

The Key Elements of “Actionable Predictive Analytics”

Actionable predictive analytics should include most of these 14 basic elements. They are:

  1. A red, yellow, or green light indicator — providing the manager with an easy to see visual red, yellow, or green light symbol lets them know immediately if this metric requires their immediate attention or action. Alternatively the problems highlighted by the metrics can be prioritized from A+ down to C.
  2. The name of the problem or opportunity revealed by the metric — putting a recognizable name on the problem or opportunity allows the decision-maker to instantly recognize what category the problem falls under. For example recruiting, retention, internal movement, or performance management.
  3. The economic impact in $ – perhaps the most important element is the potential dollar impact of the problem or opportunity. Since the language of business is money, alerting managers to the business or revenue impact of an upcoming problem or opportunity is essential to getting their attention. For example, the upcoming increase in turnover will likely impact 4 percent of revenue or $2.5 million. The cost of delayed action or the cost of doing nothing should also be quantified.
  4. Benchmark comparison numbers – after you quantify the problem (i.e. the turnover rate will increase from 6 percent to 12 percent) you must also provide comparison data so that the decision-maker can know how severe their problem is compared to benchmark numbers like, last year, today, the best, and the worst in the industry and the firm.
  5. When it would likely occur – general warnings about upcoming problems usually result in little or no action. What will spur action is to attach a time period to the problem. For example, the turnover rate will begin to increase in the first quarter of next year and it will peak by June. In addition to predictive metrics, report real-time metrics, so that decision-makers know what is happening today (as well as in the immediate future).
  6. Where will it occur – a blanket warning covering all managers, business units, or regions will generally not create a sense of urgency. However if you specify who will be most impacted by the problem, you increase the chances of getting an effected managers attention.
  7. The probability of it occurring – including the probability that the predicted problem or opportunity will actually occur will allow leaders to gauge the likelihood of it actually happening.
  8. The corporate goal that it impacts — most managers will increase their attention on any problem if that problem directly impacts one of the stated corporate goals, like revenue, sales, customer satisfaction, market share, etc.
  9. Include a trendline – in many cases, including a linear trendline can quickly get a decision maker’s attention. So consider also including a simple “hockey stick” graph to visually show the upcoming growth rate of the problem.
  10. The cost of doing nothing or delay — if the problem is likely to get exponentially worse if action is delayed, including the dollar cost of that delay or inaction will likely spur action on the part of the manager.
  11. Include the root cause  managers are more likely to take action if they know why a problem is occurring. Showing that you know the “root cause” of the problem will help to encourage managers to act. Knowing the root causes will also make it more likely that the decision-maker will select the right solution.
  12. Highlight the recommended actions – simply making managers aware of a problem won’t resolve it unless they know what actions to take to avoid or minimize it. List the top recommended actions along with their probability of success and how long before they will have an impact. For example, personalized retention plans have a 78 percent chance of success within our firm and they can have an impact within 30 days. Following the fire prevention model, managers should also be made aware of “sprinkler actions” which should be taken to minimize damage immediately after a major problem occurs.
  13. The accountable individual — because the decision-maker may require additional information or help, list the internal individual expert who is accountable for the problem and where they can get additional information.
  14. A just-in-time alert — in some cases, decision-makers will for a variety of reasons fail to take action in advance of an upcoming problem. Send them a “just-in-time alert” which gives them a “heads-up reminder” just before the problem or opportunity arrives.

The Second Essential Step — Ensuring That Your Predictive Metrics Can Be Scanned in a “Snapshot” Format

A visual display may also be necessary. Sometimes an easily scannable visual “snapshot display” makes it easier for managers to see the problem. An example of the snapshot display is provided at the top of this post.

The Third Essential Step — Where and How You Present Your Analytics Also Impacts Whether They Will Be Read

If you expect your decision-makers to change their behavior and to take immediate successful action, your predictive analytics need to be presented in a format that increases the chances that they will be seen and read by decision-makers. If you want to ensure action, make sure that they are:

  • Included in standard reports — almost no one external to talent management wants to read a standalone HR metric report. So if you want your metrics to be read, fight to have them included in standard monthly and quarterly business reports.
  • Easy to scan – time how long it takes to scan and read each individual metric and the overall set of metrics, to ensure that it meets the available time of the reader.
  • Prioritized — change the order each time so that the metrics that require immediate action appear first.
  • Only a handful of metrics are provided – because executive attention spans are limited, executives should be surveyed to identify how many and which metrics they really want to see. Often five is the limit for strategic metrics.
  • Drop-down information is available – when the reader wants it, more detailed information like definitions and formulas need to be readily available.

The Fourth Essential Step — Don’t Forget Cause and Effect

Although it is more difficult because it must be prearranged, the ultimate proof that something works is cause and effect, which can be demonstrated by using a split sample with a control group. Yes, it is the same method used to prove that drugs are effective. It can be adapted to talent management. For example, you can split a team randomly and then train a portion of a team (while holding back training from the control group), to see if there is a performance impact from the training. Many think that this approach costs more, but the opposite can be true if using the approach stops you from using talent approaches that seem to be working (using correlations) but that prove not to work under cause-and-effect testing.

The Fifth Essential Step — Select the Areas Where You Need Actionable Predictive Analytics

Once you realize the value of predictive analytics, the most obvious final question is, Which areas should you develop individual metrics? There is no standard set of recommended predictive analytics because every firm has different needs and problems as well as different data sets that can be used to predict upcoming problems and opportunities. In order to give you an idea of the possibilities, I have included a list of the talent areas where predictive metrics could have a significant impact. In the world of predictive analytics, they are often called “queries.” The most powerful ones are listed first under each talent management area.

Recruiting-related predictive metrics

  • Position openings — which jobs (and when) will need to be filled as a result of corporate growth.
  • Skill needs — when and how the future skill and experience requirements for the firm will change.
  • Performance level needed – what the average performance level that will be needed in new hires for each major job family will be.
  • Identifying the elements of hiring success – a hiring algorithm that can successfully predict the characteristics of applicants and interviewees who will perform successfully on the job. A separate hiring algorithm for successfully hiring innovators and college students may also be needed.
  • General talent availability — predict upcoming talent shortages and surpluses in the marketplace. This would include the local unemployment rate because it impacts the availability of talent.
  • When direct competition will increase – predicting when ramped up hiring by competitor firms will make it more difficult for your firm to successively hire top candidates.
  • Talent opportunities — predicting when competitor firms are likely to reduce or freeze competitive hiring and when they are likely to lay off top desirable talent. Also, when targeted individuals might reenter the job market.
  • Changing candidate expectations — when and how the expectations of our target candidates will shift.
  • Changing sources – predicting where and when source effectiveness will shift, so that using those new sources will make us more effective in attracting “not-looking” prospects and active candidates.
  • Boomerangs — predicting which former top-performing employees are likely to want to return to your firm.
  • Referrals – identifying which employees are most likely to be able to make quality employee referrals.
  • Acqui-hire targets – identifying which “talent rich” firms will be available for purchase or merger.
  • Employer brand strength — predicting when and why our employer brand strength will increase or decrease, compared to others.
  • Contingent workers  when and where will more contingent workers be needed.

Retention-related predictive analytics

  • Upcoming turnover by job — in which jobs, and when, will there likely be upcoming turnover.
  • Upcoming turnover by individual — which specific employees are most likely to leave, and when. Also which of our employees are most likely to be targeted for poaching by competitor firms.
  • Changing turnover causes and actions — what the future causes of turnover will be, and the effective counter actions for each.
  • Upcoming retirements — predicting in which jobs that upcoming retirements will occur.

Internal movement related addictive analytics

  • Need for redeployment — where and when internally will the firm have a shortage/surplus of talented teams that must be moved internally. Individuals who are soon to become “overdue” for internal movement should also be identified.
  • Ready for promotion — identifying who and when individual employees will be ready for promotion.

Leadership-related predictive analytics

  • Identifying the elements of leadership success – an algorithm is needed that successfully predicts the characteristics and skills of high potentials and actual leaders that will allow them to successfully lead. Also an algorithm is needed for identifying the characteristics of individuals who should be put on the succession plan.
  • Leadership availability — predicting where and when there will be upcoming internal leadership shortages or surpluses. Predicting when key leaders will need to be replaced.
  • Ready for leadership — identifying the employees who are ready for leadership development or actual leadership positions.

Productivity-related predictive analytics

  • Productivity problems – predicting where and when individual and team performance, productivity, and/or innovation will begin to slack off. Also identifying the barriers to productivity and innovation.
  • Upcoming performance issues – predicting in what areas will excessive absenteeism, excessive sick leave usage, sexual-harassment, engagement, safety, low morale, error rate, or other discipline issues likely occur.
  • New-hire performance – predicting which new hires are projected to be failures and successes after six months on the job.
  • Predicting burnout — identifying individuals who are approaching job burnout.
  • Likely to be released — identifying which individuals are likely to be released for performance issues and when, so replacement recruiting can begin.
  • Needed layoffs — predicting the areas where high labor costs or a surplus of talent will likely require layoffs.

Training-related predictive analytics

  • New skill sets – when current skills and training will become obsolete. What new jobs will require brand-new training programs.
  • Learning speed – when organizational learning speed will decline.
  • Learning leaders — identifying who in the organization is learning, creating knowledge, and sharing knowledge.

Compensation- and benefits-related predictive analytics

  • Who is underpaid — which groups of employees will reach underpaid status compared to market rates (and when). What the future labor costs will be for each major job family.
  • Excess OT — where in the organization excessive overtime usage will occur.
  • Benefits – when current benefits will cease to aid in attraction and retention.

Labor relations-related predictive analytics

  • Labor issues — predicting the where and when of upcoming union activity, grievances, and strikes.

Miscellaneous

  • Competitor actions — predicting competitor actions in the talent management area, as well as their reactions to our own talent management actions.
  • Organization design — when and where the current organizational and facility design willbecome obsolete.
  • Technology substitutes – forecasting areas where technology should be considered as a substitute for labor.
  • Speed issues – when and what areas organizational speed will need to be increased.
  • SWOT – forecasting strengths, weaknesses, opportunities, and threats that the firm will face in the talent management area.

Final Thoughts

The basic goal of predictive analytics is to get everyone in talent management to develop a “forward-looking mindset,” where everyone is focused on identifying and acting on upcoming potential problems. C onsider predictive metrics as a kind of radar that scans the horizon for upcoming talent management problems and opportunities. Obviously an accurate view of the future requires predictions based on data rather than personal hunches or speculation. But success also requires that any analytics and metrics that are used are designed in such a way so that they drive action and that those actions result in better people-management decisions.

About Dr John Sullivan

Dr John Sullivan is an internationally known HR thought-leader from the Silicon Valley who specializes in providing bold and high business impact; strategic Talent Management solutions to large corporations.

Check Also

The Only Metrics That Matter – Business Impact Metrics (A comprehensive example covering the business impact metrics for the compensation function)

The only HR metrics that execs don’t consider to be dreaded “so what metrics” are …