viability of hr analytics
TRANSCRIPT
The Viability of Data Analytics in Human Resources
Jonathan Gunter
SCS 2943 – 010
December 14th, 2015
Introduction – The Evolution of Human Resources
Over the last century, the model for business operations was predicated almost entirely
on sales and production. Workers were seen as expendable components of the overall
production process, carrying out mechanical tasks and having little or no autonomy in
performing their duties. Due to these working conditions the human resources agenda was
really an afterthought, being more commonly referred to as the ‘Personnel Department’ as
opposed to ‘Human Resources’. This division would function primarily as the administrative arm
of the business where responsibilities included payroll, benefits, and policy writing.
In the last thirty years however, technology has radically shifted the nature of work. The
emergence of the Internet in the 1990s led to faster and easier access to information, greater
global connectivity, automation of many administrative or tedious tasks, and self-service
capabilities. Economic drivers shifted from production-based to knowledge-based, and
companies placed a greater emphasis on the intangible skills that prospective candidates
possessed. This resulted in the “war for talent” as competition for top talent increased
exponentially. This struggle still carries through even today as the largest age demographic, the
“Baby Boomers”, progresses towards retirement. This demographic age group comprises over a
third of the current working population1, and as they phase out of the workforce the incoming
population of post-baby-boom workers will not be large enough to compensate for their
departure. As a result, there will be a dearth of critical talents and skills that companies will be
fighting for tooth-and-nail.
1 Population by sex and age group, Statistics Canada (Sept 2015). www.statcan.gc.ca/tables-tableaux/sum-som/l01/cst01/demo10a-eng.htm
It is due to the contributing factors mentioned above that progressive companies are
beginning to embrace Human Resources as a strategic business partner. Pressures have begun
to mount on the HR function, to pivot away from the traditional activities of facilitating new
employee orientations and writing handbooks and policies to instead be more active
contributors to business strategy; finding efficiencies through people in order to push the
bottom line. In facing new pressures, HR leaders need to adopt new ways of thinking. As
mentioned by Deloitte HCM Principal Josh Bersin, some executives will often make critical
business decisions based solely on experience or intuition2. Instead of relying on personal
judgment, a greater emphasis needs to be placed on tapping into the ever-expanding volumes
of available data. However, the issue with this approach is that few companies know how to
properly implement an analytics initiative.
Potential Barriers to the Viability of HR Analytics
Insufficient Understanding of Data & Analytics
One of the primary barriers with HR analytics is that many HR departments perceive
themselves to already be analytical simply by producing dashboards or scorecards that depict
headcount totals or turnover rates. While metrics on the efficiency of HR activities (i.e. time to
hire) can be useful in getting a sense of the current state of affairs, it provides marginal value or
insight for business leaders and doesn’t mean much to operations outside HR3. In a study by
Bersin by Deloitte, 86% of organizations focus their efforts on ‘reporting’, choosing to provide a
description of past activities and not perform any analytical modeling to predict future positive 2 Developing Advanced Talent Analytics: Why It Matters to CFOs, Bersin by Deloitte, Wall Street Journal (Sept 2015)3 Maximizing the Impact and Effectiveness of HR Analytics to Drive Business Outcomes, Scott Mondore, Shane Douthitt and Marisa Carson, Strategic Management Decisions, Vol. 34, Iss. 2 (2011)
or negative outcomes4. Of the 14% of respondents that were deemed to be “mature” analytic
organizations, 80% were found to have improved recruitment practices due to their effort in
measurement and analytics5. Further to that, Deloitte discovered through their studies that
analytic organizations were twice as likely to believe they were consistently selecting the right
candidates and delivering a strong leadership pipeline, and three times as likely to believe they
were efficiently operating HR6. Lastly, data-driven HR organizations were found to generate 30%
higher stock returns than the S&P 500 over the last three years7.
For other HR departments that understand the need to transition from ‘reporting’ to
‘analytics’, most either don’t know where to start or apply analytics resources to the wrong HR
activities. To gain some momentum, HR can obtain guidance from their colleagues in Finance to
get a sense of how to progress their analytic capability. As mentioned by Human Capital
strategist Jac Fitz-Enz, finance professionals have moved up in recent years from the perception
of simply being accountants to supplying valid and valued strategic analytics to business
executives8. Finance became a more valued function to the business as financial capital grew to
be more important than physical capital in a service-based economy. As a result, the Finance
function evolved to develop generally accepted accounting principles (GAAP) that were
accepted across industries. These principles allowed senior Finance professionals and leaders
to develop valued analytics specific to their own businesses9. The Human Resources function is
4 High-Impact Talent Analytics: Building a World-Class HR Measurement and Analytics Function, Bersin by Deloitte, (Sept 2013).5 Developing Advanced Talent Analytics: Why It Matters to CFOs, Bersin by Deloitte, Wall Street Journal (Sept 2015)6 HR Technology 2015 – Ten Disruptions: Ignore Them At Your Own Peril, Josh Bersin (Dec 2014), Slide: “Analytics Drives Huge ROI”.7 Ibid.8 The ROI of Human Capital, Jac Fitz-Enz (2011), Ch. 1 (Human Leverage), pg. 13.9 Ibid.
now in a similar situation, where human capital is coming to the forefront of business strategy.
Since most CFOs have an extensive view across the organization, they can provide valuable
perspective on deficient areas of the business and where HR should focus their analytic
efforts10.
Insufficient Statistical Skills & Tools in the HR Department
A misconception is that even if HR leaders are able to pinpoint areas of opportunity for
applying analytics, they suffer setbacks due to gaps in statistical talents and tools available
within the HR department. A study by the Aberdeen Group revealed that the biggest problem
with workforce analytics is having people who know how to properly analyze numbers. The
study reported that 44% of companies reported a lack of people resources that understand how
to interpret and analyze data, and transform the data into actionable insights11. However this
should not become an obstacle when attempting to implement an analytics initiative, since
statistical skills can be captured from different sources. Companies can search the external
market for statisticians or actuaries who specialize in data modeling, and couple their skill set
with the domain knowledge possessed by HR professionals. The two sides can work in tandem
in order to build a data map to understand key performance indicators (KPIs) in a human capital
context, what sources of data are available and what data attributes should be pulled in to the
analytic study. If conducting an analytical search externally proves too costly, companies can
leverage analytical talent within the company, such as financial analysts, to provide guidance on
how to properly manipulate and analyze sets of data. Finance will often possess the sharpest 10 Developing Advanced Talent Analytics: Why It Matters to CFOs, Bersin by Deloitte, Wall Street Journal (Sept 2015).11 Four Most Significant Barriers to Producing Workforce Analytics, Michael Moon, HCM Essentials – Aberdeen Group (Oct 2015)
analytical skills, and these professionals can assist HR in understanding how to link talent data
with revenue, profitability, and other operational business data12.
Another commonly perceived barrier is a lack of statistical tools and technology available
to properly analyze HR data sets. In the aforementioned study by the Aberdeen Group, nearly
35% of companies insist they don’t have tools or software to assist with analytics13. The truth of
the matter is that HR can perform data modeling at little to no cost, using a combination of
simple analytic techniques and open source statistical software. For example, correlational
analysis can be a useful tool for uncovering relationships between talent data attributes. The
downside however, is that uncovered relationships can merely be coincidental and lead to
invalid conclusions drawn from the data. Alternatively, a more advanced modeling technique
like multiple linear regressions can be used to measure multiple predictors simultaneously, and
then select the variables that have the strongest relationship with the outcome variable. This
technique can be more effective than correlational analysis, since it quantifies and prioritizes
the best individual predictors to an outcome14. The most complex but effective way to measure
multiple talent variables at once is through Structural Equations Modeling (SEM), which takes
into account multiple independent and dependent variables all at once to determine cause and
effect relationships. As opposed to correlational analysis, the cause and effect relationships that
can be uncovered from SEM results in much more conclusive evidence that is less likely to be
12 Developing Advanced Talent Analytics: Why It Matters to CFOs, Bersin by Deloitte, Wall Street Journal (Sept 2015).13 Four Most Significant Barriers to Producing Workforce Analytics, Michael Moon, HCM Essentials – Aberdeen Group (Oct 2015).14 Predictive Analytics for Human Resources, Jac Fitz-Enz & John R. Mattox II (2014), Ch. 6 (Predictive Analytics in Action), pg. 101.
dismissed by senior leadership when sharing findings15. Furthermore, talent data collection
brings a portion of measurement error with it. While correlation assumes that everything is
measured without error, which can be problematic, SEM can correct for measurement error
using reliability assessments16. The downfall of structural equations modeling is that it can be
very complex, and therefore requires specialized software to perform. Fortunately there are
modules available in SAS statistical software which can be used, or even the lavaan and “sem”
packages available in “R”, which is freely available statistical software17. The models that HR
builds however, will only be as effective as the data they possess.
Sourcing Data & Data Quality
Data quality is a critical component to analytics, especially for a function like HR that
already has tenuous ties to executive leadership. If analytical findings are produced from
inaccurate or messy data, senior leadership will dismiss the findings and HR will lose credibility
as a result. In any analytics study, a significant investment in time and effort should be
dedicated to data acquisition and cleansing. While it isn’t glamorous work, it is nonetheless
pivotal to achieving successful analytic results. Since a considerable amount of effort needs to
go into data cleansing efforts, it’s important to distinguish which data attributes are most critical
to the analytics exercise and start improving data quality on these particular variables18.
15 Maximizing the Impact and Effectiveness of HR Analytics to Drive Business Outcomes, Scott Mondore, Shane Douthitt and Marisa Carson, Strategic Management Decisions, Vol. 34, Iss. 2 (2011)16 Ibid.17 About SEM Software. http://www.structuralequations.com/software.html18 Developing Advanced Talent Analytics: Why It Matters to CFOs, Bersin by Deloitte, Wall Street Journal (Sept 2015).
Another important barrier to consider in relation to data quality is the number of
dissimilar data sources that need to be gleaned from for useful information. These sources will
often comprise different data dictionaries and will belong to different data owners or stewards
across different functions. Michael Moon of the Aberdeen Group would refer to these different
sources of data as “information silos”, where the data is static and housed in different parts of
the company19. Even for analytic initiatives that exclusively reside within HR, this can still prove
to be a cumbersome exercise due to different systems used for applicant tracking, payroll,
compensation and benefits, employee administration, and learning and development, just to
name a few. According to a Bersin by Deloitte study, about 1 and 3 organizations have ten or
more HR systems20. Data that does not flow freely though different channels of the company is
not being utilized to its full value, so it’s critical to have the appropriate systems development
resources available to install the necessary integration programming to open up the channels
between these silos21.
HR departments that can successfully navigate through the potential barriers mentioned
above offer themselves an opportunity to deliver true value to their clients in the business. The
next section will cover some specific case studies within HR, where data analytics has proven to
be an effective tool and has delivered compelling insights that could trigger transformative
action to business operations.
19 Four Most Significant Barriers to Producing Workforce Analytics, Michael Moon, HCM Essentials – Aberdeen Group (Oct 2015).20HR Technology 2015 – Ten Disruptions: Ignore Them At Your Own Peril, Josh Bersin (December 2014)21 Four Most Significant Barriers to Producing Workforce Analytics, Michael Moon, HCM Essentials – Aberdeen Group (Oct 2015).
Applications of HR Analytics
Recruitment & Selection Analytics
In a meta-analysis study written in the Harvard Business Review, researchers discovered
that human judgment can often have its pitfalls during the recruitment and selection process.
The authors argue that while experienced managers and executives are very effective at
identifying the core skills and competencies required for a specific role, their intuition fails them
in the actual selection decision since they place too much emphasis on inconsequential
factors22. The study discovered that algorithms were 25% more effective than human selection
decisions, outperforming humans in selecting above-average performers through variables such
as supervisor ratings (algorithms were 7% more effective), number of promotions (4% more
effective) and ability to learn from training (11% more effective)23. Another study by the
National Bureau of Economic Research reviewed over 300,000 hires from 15 different
companies, and compared the tenure of employees who had been hired based on the
algorithmic recommendations of a job test with that of candidates who had been selected by
humans24. The job test included a wide array of questions concerning technical and cognitive
abilities, personality traits, and cultural fit25. After running the answers through an algorithm,
the predictive model classified the candidates into three groupings: green for high potential
employees, yellow for medium potential and red for the lowest rated potential. The conclusion
was that the algorithm worked, and that green hires stayed at the job 12 days longer than
22 In Hiring, Algorithms Beat Instinct, David Klieger, Nathan Kuncel, & Deniz Ones, Harvard Business Review (May 2014).23 Ibid.24 Machines Are Better Than Humans at Hiring the Best Employees, Rebecca Greenfield, Bloomberg Business (Nov 2015).25 Ibid.
yellow hires, which stayed 17 days longer than red-classified hires26. The sentiment reached
based on the findings of these studies is that humans that make the selection decisions are
often overcome by a number of different biases, which leads them astray when attempting to
select the best candidate. The objective is not to replace recruiters with machines as there are
still important cues that humans can pick up on that machines cannot. However, HR
professionals and business line managers need to overcome their distrust of algorithms and
instead embrace it as a useful tool to supplement the hiring decision.
Retention Analytics
Employee retention is an area that presents one of the greatest cost-savings
opportunities within HR. When factoring in recruitment costs, training costs, and productivity
losses, it can cost companies and estimated average of six to nine months’ worth of salary to
replace an employee, and this cost increases the longer an employee stays27. However when
looking at a combination of employee attributes from internal HR systems and external data
that can be found on social networking sites, organizations have the potential to collect
extensive retention risk data on their employee base. This includes aggregated retention risk
profiles by employee, as well as identifying the underlying drivers of retention and attrition28.
In an example from Deloitte, their human capital consultants worked with a global
company in China to improve employee retention with their sales workforce. The initial belief
from middle management was that increasing compensation would improve retention rates.
However after conducting an analytics study, it was discovered that there was minimal 26 Ibid.27 Can an Algorithm Prove You won’t Quit Your Next Job? Rebecca Greenfield, Bloomberg Business (Nov 2015).28 HR Technology 2015 – Ten Disruptions: Ignore Them at Your Own Peril, Josh Bersin (December 2014).
correlation between compensation and turnover29. Instead Deloitte identified that time in role
and supervisor tenure were better indicators of turnover, and as a result recommended that
development opportunities, job rotations, and career discussions be held with employees that
were identified in the model as “high risk”30. The result was a significant drop in turnover over
the next six months, which allowed the company to reach their sales growth targets in the
region31.
Another company on the leading-edge of people analytics is Google, led by VP of HR
Laszlo Bock. Mr. Bock implemented what has been dubbed a “Three Thirds” structure to his HR
team, where one third of the team comes from a traditional HR background, one third has
backgrounds in business strategy consulting, and the last third comes from an analytical
background and possess Masters Degrees and PhDs in data and analytical science32. The
philosophy adopted by Bock’s HR team is a data-driven one, where every decision revolving
around people management is supported by algorithmic or data-based evidence33. Included in
their decision-making practices is a retention algorithm that allows Google to proactively and
successfully predict which employees are at greatest risk of leaving. This not only provides early
alert warnings to management on risks of attrition, but additionally permits the use of
personalized retention solutions by identifying which employee attributes carry the most weight
in the reason for leaving34.
29 Developing Advanced Talent Analytics: Why It Matters to CFOs, Bersin by Deloitte, Wall Street Journal (Sept 2015).30 Ibid.31 Ibid.32 Building a New Breed, Michael O’Brien, Human Resource Executive Online (Oct 2010).33 How Google Is Using People Analytics To Completely Reinvent HR, Dr. John Sullivan, ERE Media (Feb 2013).34 Ibid.
Workforce Management & Planning
In addition to their advanced retention analytics, Google has done an effective job of
using analytics to understand the core competencies and behaviours of great managers and
leaders. When Google initially started out, there was a common belief among leadership that
middle management was not critical to business operations. As a result of this belief, they
disposed of this line of management entirely. When they quickly realized that this belief was
false, they turned to data as a means of understanding manager effectiveness35. After analyzing
a substantial amount of internal performance data, they were able to prove that great managers
had a statistically significant influence on critical measures such as attrition, productivity, and
team performance36. Based on this analysis, Google took the next step of classifying “good
managers” and “struggling managers”, and conducted qualitative interviews of both groups.
Using text analysis they were able to identify common traits of good and struggling managers,
and use this information to build a semi-annual performance evaluation for their managers.
Included in this evaluation framework was an alert system that allowed Google to proactively
identify “good” and “struggling” managers, and provide remedies for struggling managers
through training and support37.
Another great example of using analytics to optimize workforce planning came from a
long-standing energy company managing its largest growth in headcount in its 130-year history.
Black Hills Corp. doubled its workforce to over two thousand employees after a significant
acquisition. In evaluating their workforce, Black Hills ascertained that a substantial portion of
35 Is HR Going to the Geeks?, Bernard Marr, LinkedIn (March 2015)36 Ibid.37 Ibid.
their workforce was aging and close to retirement. This created a significant risk for a talent
deficiency, as the company forecasted that they could lose over 8,000 years of experience in the
next five years38. To prevent the massive dearth in talent, Black Hills used workforce analytics to
calculate the number of expected retirements per year, the types of workers required to replace
them, and where to source those workers from in their recruitment activities. The result from
this aggregated analysis was a workforce planning summit that devised and prioritized nearly 90
action plans to address the shortage in talent39.
Conclusions and Analysis
There are countless other business case studies that can be drawn from to depict the
effective use of HR analytics. However the common link between these examples is one of the
core problems that continues to plague HR activities, which is that too much of people-based
decision-making is carried out with little or no data and excessive reliance on personal
experience and intuition. Other enabling or supporting functions in the business such as
Finance or Marketing actively use data and analytics in nearly all of their consulting activities
with the business. Employees are often a company’s greatest asset, typically representing
around 60 percent of corporate variable costs40. If this is the case, why is there so little evidence
based decision-making used when casting critical HR decisions? An article from Mondore,
Douthitt, and Carson best articulated the approach that HR leaders need to take to not only be
more analytical, but better positioned to be an effective partner to the business41:
38 Change Your Company with Better HR Analytics, Mick Collins, Harvard Business Review (Dec 2013).39 Ibid.40 How Google Is Using People Analytics To Completely Reinvent HR, Dr. John Sullivan, ERE Media (Feb 2013).41 Maximizing the Impact and Effectiveness of HR Analytics to Drive Business Outcomes, Scott Mondore, Shane Douthitt and Marisa Carson, Strategic Management Decisions, Vol. 34, Iss. 2 (2011)
1. Calculate the Return-on-Investment (ROI) in everything that they do.
2. Give evidence-based advice on how to drive the business from a people perspective.
3. Take accountability for a portion of the organization’s financial health.
4. Show results (HR effectiveness) and not just HR activity completion (HR efficiency).
5. Create an HR strategy that has direct impact on the bottom line.
In summary, the cliché of “HR earning its seat at the table” is overused and still dubious in many
cases. Effective business leaders in the C-suite make decisions based on good data and insights.
It’s time for HR to start doing the same, and this starts with asking the right questions. However
the right question does not start with HR, it starts with the business. The most basic questions
that should serve as the starting point are “what problems does our business face today?” and
“what people-based priorities can support the correction of that problem?” This is not to
suggest that algorithms completely replace HR professionals, but instead for HR employees to
become more adept at understanding and analyzing data. By having effective data analysis and
HR domain knowledge working in tandem, it can lead to more effective and evidence-based
decision-making that will assist HR in taking that next step to becoming an effective business
partner.
References
1. Population by sex and age group, Statistics Canada (Sept 2015). www.statcan.gc.ca/tables-tableaux/sum-som/l01/cst01/demo10a-eng.htm
2. Developing Advanced Talent Analytics: Why It Matters to CFOs, Bersin by Deloitte, Wall Street Journal (Sept 2015)
3. Maximizing the Impact and Effectiveness of HR Analytics to Drive Business Outcomes, Scott Mondore, Shane Douthitt and Marisa Carson, Strategic Management Decisions, Vol. 34, Iss. 2 (2011)
4. High-Impact Talent Analytics: Building a World-Class HR Measurement and Analytics Function, Bersin by Deloitte, (Sept 2013).
5. HR Technology 2015 – Ten Disruptions: Ignore Them At Your Own Peril, Josh Bersin (Dec 2014), Slide: “Analytics Drives Huge ROI”.
6. The ROI of Human Capital, Jac Fitz-Enz (2011), Ch. 1 (Human Leverage), pg. 13.7. Four Most Significant Barriers to Producing Workforce Analytics, Michael Moon, HCM
Essentials – Aberdeen Group (Oct 2015)8. Predictive Analytics for Human Resources, Jac Fitz-Enz & John R. Mattox II (2014), Ch. 6
(Predictive Analytics in Action), pg. 101.9. About SEM Software. http://www.structuralequations.com/software.html10. In Hiring, Algorithms Beat Instinct, David Klieger, Nathan Kuncel, & Deniz Ones, Harvard
Business Review (May 2014).11. Machines Are Better Than Humans at Hiring the Best Employees, Rebecca Greenfield,
Bloomberg Business (Nov 2015).12. Can an Algorithm Prove You won’t Quit Your Next Job? Rebecca Greenfield, Bloomberg
Business (Nov 2015).13. Building a New Breed, Michael O’Brien, Human Resource Executive Online (Oct 2010).14. How Google Is Using People Analytics To Completely Reinvent HR, Dr. John Sullivan, ERE
Media (Feb 2013).15. Is HR Going to the Geeks?, Bernard Marr, LinkedIn (March 2015)16. Change Your Company with Better HR Analytics, Mick Collins, Harvard Business Review
(Dec 2013).