People & Strategy Summer 2017 Vol. 40 Issue 3 - 40
Presentations and discussions that focus on the key data
points will generally be more efficient and compelling than
those that wander through every nuance. Thus, it is typically useful to limit sharing results to key items related to the
insight yielded by the analysis, and save the detailed analysis
for an appendix.
To return to our hiring example, perhaps our analyst used
AI to identify the key predictive variables of successful employees over a three-year period, as well as conducting follow
up focus groups. Perhaps the artificial intelligence engine
identified graduation from one particular school as being an
important predictor of success. Follow up focus groups with
new hires from that school, as well as the managers of those
new hires might reveal that this particular school offered a
class in exactly the skills required for the business.
Building talent strategies on
data-based insight will deliver
significant business value. Doing this
effectively requires more than simply
letting an unsupervised artificial
intelligence engine loose in your HR
information system, however.
Step 6: Taking Action Based on That Insight
Developing insight is a necessary but not sufficient condition
for the true purpose of a talent analytics project: driving
action. All of the steps up to this point have been designed to
lead us to this critical step.
To the extent that the purpose of talent analytics is materially affecting the operations or success of an organization, a
talent analytics team cannot ethically consider its work complete at the end of the insight phase. Insights are only useful
when they can be used to inform a decision or action. To this
end, talent analytics teams are well served to include or closely
partner with individuals who are experts in more traditional
HR skills such as change management.
A key risk for talent analytics functions is in delivering
insights that are interesting, but not actionable. For example,
random forest analysis may yield a reasonably precise estimate
of attrition risk. As a research project, this may meet the first
five criteria of our model, culminating in an insight. However, the method may simply flag risk without articulating the
drivers of that risk. This situation would require additional
research into the risk factors for the high attrition risk individuals. Otherwise, it might lead to inefficient interventions that
waste resources and may have a marginal effect on attrition. In
organizations with relatively newer talent analytics functions,
such projects can create a sense that talent analytics is not of
practical utility, in turn putting investment and influence for
the function at risk.
To return to our hiring example where we learned that
one school produced the best employees due to the skills
PEOPLE + STRATEGY
training available, we can now recommend two actions. First,
the firm should focus hiring on the school that is currently
producing the skills needed. Second, the firm could work
with that school as well as others to offer a similar curriculum
in additional systems, thereby raising the overall quality of
the talent supply. This expansion could also include things
like supporting the skills program in at least one school with
a higher representation of underrepresented minorities as a
way to support diverse hiring and social good.
Step 7: Measuring Results to Determine Whether Your
Action Was Effective
Although taking action is the goal of a talent analytics project, the work is not truly finished at that step. The work is not
done until we verify that the action delivered the intended
Measuring outcomes to ensure success is important for several reasons. Most obvious, such measurement can help fine
tune processes to improve outcomes, or identify areas where
the interventions are not delivering sufficient improvement.
Additionally, documenting the delivered business results from
a complete talent analytics project (through implementation)
is important to ensuring continued investment and opportunity for talent analytics projects. Finally, ongoing or periodic
measurement helps identify areas where the impact of an
intervention may be changing over time. An intervention that
was initially very successful may decay over time such that a
refresh on the project is necessary to continue to deliver the
An effective and relevant business question in step one is
essential to good evaluation and ongoing measurement. The
business problem, data, and method will determine the best
schedule for follow up evaluation and ongoing measurement.
Some questions, like impact of the design of a recruiting website on resumes submitted, might be measured in a day. Some,
like the effectiveness of high potential programs in preparing
senior leaders, might take several years.
Let's return a final time to our hiring example. The actions
were to increase hiring from a specific school, and to expand
the skills class to additional schools. The follow-up measurement on the same-school hiring could be evaluated fairly
quickly, perhaps 30 or 60 days. Determining whether expanding the skills program to additional schools would likely take a
full year, and more likely three years to add the class to curriculum, have students take the class and be hired, and have new
hires perform for a significant amount of time. Thus, each
specific action will require a thoughtful follow-up schedule.
Talent analytics is a powerful way to create power and
impact for your talent function. Building talent strategies on
data-based insight will deliver significant business value. Doing
this effectively requires more than simply letting an unsupervised artificial intelligence engine loose in your HR information system, however. Following the steps outlined in this
article, you can move your organization toward better strategy
execution and greater success.
Alexis A. Fink is general manager, talent intelligence and analytics
at Intel Corporation. She can be reached at email@example.com.