HR Analytics

Lecture 3
Previous on HR Analytics…

• What is HR analytics? Why it is important?

• The strategic value and position of HR (analytics).
• What do we understand with the “predictive” part?
• Benefits and challenges of HR analytics
• Different frameworks highlight different components.
• There is no “one size fits all” approach to implementing analytics into your people function,
but it is underpinned by capability, data and technology.
• The importance of ethics and governance.

2
Agenda

• Information sources, how to measure, analytical software

• Big data
• Make a business case

3
Agenda
Learning Objective 3,4,5,6:
• Apply statistics (e.g., descriptive, t-test, regression analysis) to describe specific HR problems (e.g., absenteeism,
individual performance).
• Develop a research design to evaluate the strengths and weakness of various HR practices and/or interventions
(e.g., the effects of leadership development practices on employee performance), with an application of statistical
methods (e.g., descriptive, t-test, regression analysis), and packages (e.g., SPSS).
• Justify – “sell” the results of HR problems to top and line management (e.g., to show the impact, weather beneficial
of detrimental of various HR practices towards employee performance) and provide policy suggestions.
• Develop critical thinking and logic – how to develop testable models from theory and understand how such models
can be tested (e.g., descriptive, t-tests, regression analysis), and what software can be used (e.g., SPSS).

By the end of this lecture you will be able to:

• Explain what is data and information, sources of information and argue which software can be used for your
analyses.
• Synthesize what is big data and what are the key challenges of big data in HR analytics.
• Argue how to make a business case with data and technology.
• Run a simple analyses in SPSS for a specific problem (diversity).

4
HR information systems and data (1)

• Some very important questions before before the analysis process:

• What type of data do you have in your organization already?
• What problems you actually have in the organization?

• Do not limit yourself by what is immediately available in the HR department, even if you are
in HR role.
• If you do so, you might be able to model organizational patterns of behavior and link HR
practices directly to revenue or efficiency.

5
Data, information, knowledge, wisdom

• Data – set of discrete, objective facts (12.8)

• Information – we add a context; a systematically-arranged set of data that has meaning (the
turnover for people in grade 7 and 8 in the IT sector in UK is 12.8 per cent)
• Knowledge – gained through experience and expertise in a particular field. Mostly hidden,
until is shared.
• Wisdom – integrated knowledge—information made super-useful, “intuition” etc

• Information and knowledge enable us to increase efficiency (doing things right), not just
effectiveness (doing the right things).

6
Types of data (1) – Categorical variables

Data type Description Examples

Binary Only two categories Male or female; someone who has left
versus someone still employed; disabled
or not.

Nominal Three of more categories Division or function (e.g., Finance,

Marketing, Operations); country (e.g.,
Slovenia, Netherlands, Germany)

Ordinal Categories have a meaningful order Customer satisfaction scale (very

dissatisfied, neutral, satisfied, very
satisfied); ranked sales position of
employee or team (1st, 2nd, 3rd, etc).

7
Types of data (2) – Continuous variables

Data type Description Examples

Interval Numerical value. No defined “zero” Temperature; year.

starting point.

Ratio Continuous numerical data with a Height; weight; salary; age; job tenure;
defined zero starting point. Ratios of sales figures (currency); % linked to an
scores on a scale also make sense. individual object such as individual, team,
department (e.g., % of the team who
indicated that they were “engaged” or staff
turnover % in each team or division).
Information sources (1)

• What type of data is available to be analyzed will vary greatly depending on the type of
organization.
• Some examples:
Information source Description Examples
HR database (SAP, Oracle) Information describing the Age, gender, education level, role,
employees across a wide range of salary, performance ratings,
HR activities, including personal disabilities, tenure, sickness
details, performance, diversity data, absence, leaver information, country
promotion information. location, team leader, etc.

9
Information sources (2)

• Some examples:
Information source Description Examples
Employee attitude survey data (often Survey results from annual Job-strain level, employee
stored in survey programmes and employee attitude surveys. Either engagement, job satisfaction,
exported to Excel) managed in-house or with an person-organization fit, perception of
external survey provider justice…
organization. Typically containing
employee engagement data. The
information will depend purely on the
question asked.

10
Information sources (3)

• Some examples:
Information source Description Examples
Customer satisfaction survey data Often run by marketing or sales Customer rating on specific services,
(often stored in survey programmes function survey data, often provides branches and customer-facing
and exported to Excel) extremely useful information about employees, customer loyalty,
customer preferences and, when customer preferences, customer
linked to employee profile and satisfaction, likelihood of further
operational data, can help business…
organizations to identify how changes
to employee skills and behaviours, as
well as any operational or process
changes, can have a direct impact on
the customer experience.

11
Information sources (4)

• Some examples:
Information source Description Examples
Sales performance data Information usually owned by the sales Average monthly sales, number of
function, recording details of sales new customer introduced, meeting
performance and revenues. This or not meeting target
information is key to determining
organizational revenues and the
success or otherwise of salespeople
and teams in meeting targets. When
linked with employee data and
operational information it can also help
determine the conditions under which
salespeople and teams are more likely
to succeed.

12
Information sources (5)

• Some examples:
Information source Description Examples
Operational performance data Information or data measuring the Supermarket scan rates, call-centre
successful running of the business. call durations, average number of
Usually referring to efficacy. calls dropped out, average number
of queries resolved on first contact,
time taken to on-board a new
costumer…

13
How to measure? (1)

• But how to measure various attitudes of employees?

• Most of the “things” you would like to measure are not tangible and cannot be measured as
easily as, say, salary or attendance to training.
• Avoid asking one simple question to capture complicated things. For example: “How
engaged are you?”

14
How to measure? (2)

• Questions included need to be good indicators of what you are trying to measure (and they
are reliable).
• Make few assumptions:
1. You want to find out what people are thinking, feeling and doing.
2. You want to measure attitudes and perceptions that you think you are measuring.
3. You want to measure things to take prescriptive actions.
4. You don’t want to waste everyone’s time
with unreliable measures a meaningless
results.
5. You don’t want to waste money on
expensive interventions from meaningless
results.

15
How to measure? (3)

• Your measure should accurately and consistently tap the degree to which employees vary
or differ on that concept.
• A measure has two key ideas:
1. Construct: is the idea that the measure is trying to tap into (e.g., engagement).
2. Indicator: which can be seen as statement describing the construct (e.g., “I am proud of
my job”.), which leads to 1-5 response.
• The questions are used to indicate the level of a construct in the employees’ “head”.
• Keep in mind, one indicator is not enough to measure a construct.

16
How to measure? (4)

• When presenting a measure of a construct you should be able to clearly explain exactly
what the construct entails.
• Engagement: “a positive, fulfilling work-related state of mind that is characterized by vigor,
dedication and absorption” (Schaufelli & Baker, 2003).
• Indicators (some examples):
• Answer on a scale from 1 to 5, where 1 = “strongly disagree” and 5 = “strongly agree”.
• “When I get up in the morning, I feel like going to work” (Vigor).
• “My job inspires me” (Dedication).
• “Time flies when I am working” (Absorption).

17
How to measure? (5)

• Answers will vary among employees (some will say 1, some 4, some 5…).
• A good measure with precise indicators will produce “accurate” variation in the numbers
recorded.
• A bad measure (maybe the wording was complex or confusing) will provide variation based
on how well the person understands the questions. = BAD SCENARIO!
• Practically it is very unlikely that any indicator will indicate or gauge something perfectly. A
very relevant problem in HR.
On a scale of 1 to 10,
how would you rate
your overall job
satisfaction and work-
life balance?

18
How to measure? (6)

X=t+e
X = the degree to which the numbers produced in our survey records vary.
t = true component: an indication of true variation.
e = error component: measurement error.

Bad measure means high error!

19
What is a good measure?

• Should be good = valid. Validity refers to whether the measure accurately represent the
construct intended.
• Different types of validity:
1. Face validity. It evolves around the notion that the indicators actually look as if they
should measure the construct at hand.
2. Construct validity. When indicators actually measure the construct at hand.
3. Discriminant validity. To check if the measurement is not actually measuring some
other construct.
4. Criterion validity. Do the measures demonstrate the appropriate degree of statistical
robustness that tells us they are measuring the construct?

20
Analysis software options (1)

• There is no shortage of statistical software programs:

ALL ROUNDERS
• SPSS (GUI)
• Minitab (GUI)
• Stata (command-line)
• SAS (command-line)
• R (command-line) (free)
• JASP (GUI) (free)
• Excel

21
22
Analysis software options (2)

• There is no shortage of statistical software programs:

SPECIALIZED PROGRAMS
• Structural equation modelling:
AMOS (GUI),
PLS (GUI) (excellent for small samples < 50),
MPLUS (command-line)
• Multilevel modelling:
HLM (GUI),
MLwin (GUI),
MPLUS (command-line)
• Social network analysis:
Pajek (GUI)
UCINET (GUI)
EgoNet (GUI) (free)

23
Big Data

24
Big data (1)

• Big data is often used to describe big sets of data that are so large that traditional methods
for sorting, analysis and sharing data can be inadequate.
• Attempts to move the definition away from the size of the data, to its “smartness” i.e. the
extent to which it is able to provide the material to conduct fine grained analysis that
successfully explains and predicts behaviour and outcomes (George et al., 2014).
• Problems to bring all this data for analysis.

25
Big data (2)

• A simple illustration:
• A company of 10.000 employees has 50 data points (name, last name, gender, date of birth,
position, job history, employee ID, line manager name information/structure, salary,
performance reviewer, bonus payments, etc…) for each employee in a three-year period.
• 10.000 employees x 36 moths x 50 data points = 18.000.000 data points you need to work
with.
• Using Excel with this breadth and depth of data is unlikely to be fully effective, and the actual
analyses will take a lot of time and effort.
• Excel can calculate that 12.5 per cent attrition for an organization is going up, or going down,
or is twice as high as a year ago. But the key question is whether this is good or bad.

Excel: Why using Microsoft's tool caused Covid-19 results

to be lost (bbc.com)
26
 Types of databases (1)

• Types of databases you are likely to come across in organizations:

1. Relational databases (SAP, Oracle, Sybase) store information in tables.
2. Multidimensional databases (Cognos, Business Objects), are also called online analytic
processing (OLAP) databases.

1. HR Example: Imagine a database where each row is an

employee and each column is information about them
(name, job title, department, salary). This helps HR track
and manage employee records efficiently.
2. HR Example: Picture a database that allows HR to
analyze employee performance data by different
dimensions, such as by department, by job role, and by time
period. This helps in identifying trends and making informed
decisions about promotions and training needs.

27
 Types of databases (2)

• Types of databases you are likely to come across in organizations:

3. Object-oriented databases (Workday), which stores data as “objects”.
4. Big data databases (Hadoop) can handle enormous amounts of data.

3. HR Example: Think of an employee

database where each employee is an
object with attributes like name, job title,
skills, and methods like
calculateBonus(). This allows for more
complex data management and
operations, such as tracking career
progression and skill development.
4. HR Example: Imagine storing and
analyzing all the data from employee
surveys, social media interactions, and
performance reviews. This helps HR
gain insights into employee sentiment,
engagement, and overall organizational
health.

28
HR analytics and big data (1)

• Traditional vs big data databases = work together to provide the best possible experience.
• We are a long way to go as we are facing multiple problems:
• End-to-end analytics (Rasmussen & Ulrich, 2015)
• Insufficient data to ask the right questions – the problem of silo mentalities (Angrave et al., 2016) ,yet
you can still run some analyses even if you have very little data following simple
principles:
• The business problem that we are trying to solve;
• The data that is available to ensure that we make the most relevant effective and
financially viable decision or recommendation to the business.
• HRIS problems! Consultancy selling products we do not need, costly to implement,
providing relatively little predictive value. Generic ‘best practice’ approach rather than
deep engagement with the organisation and its needs.

29
Building business cases with
data and technology

30
Intro

• Articulating a powerful business case means that you need to be able to speak the language
that resonates with the business, and the senior leaders who are making that decision.
• Commercial, viable numbers and predictions that make realistic sense.
Stage 1

Stage 2

Stage 3
The The Desired
diagnosis proposed outcomes
solution

31
Stage 1: The diagnosis

• Key components:
• Establish the initiative > identify critical business challenges and process that could be
optimized, and/or opportunists for change.
• Identify external insights and data that can help build some data regarding the issue >
project assessment methodology, review external data that outlines the cost of field
project.
• Identify internal business challenges that will be impacted by the initiative, for example
improving productivity.
• Identify as many details and data point as possible when identify the causes > new
employee engagement survey process > quantify the extent to which each issue
contribute to the lack of engagement.

32
Stage 2: Outlining the proposed solution

• Key here is to explain why the solution you are suggesting can resolve the problem that you
have identified.
• Be as specific as possible about what is required.
• For each step, describe what resources you will need and how they will be allocated,
along with how the step will contribute to the solution > costs/benefits analysis.
• Provide a timeline for project execution and completion for each component of the
initiative.
• Establish the end date and the follow-up procedures and metrics to measure project
success so the senior leaders can understand the anticipated return on the proposed
investment.
• Outline what the senior leaders will have to do as key stakeholders of the initiative.

33
Stage 3: Determining the desired results

• This stage is about demonstrating what the initiative will achieve for the organization in terms of
performance improvements.
• Outline what effect your initiative will have on the original business challenges that were
outlined at Stage 1. Focus on quantifying the anticipated performance improvements.
• Identify the internal impact created by the initiative, outlining the return on investment (ROI)
predictors. These will reflect organizational improvements and efficiencies, with data to
support your views such as productivity data, reduced turnover etc.
• External benefits, if there are any, quantify them with data trends drawings on external
insights. E.g., new recruitment process might improve employee branding, work with
marketing function to outline what the benefits look like in terms of tangible performance data.
• Identify clear takeaways that will resonate with senior leaders in terms of resolving some of
their business challenges and pressure points.
• Outline a picture of what the organization will look like after the initiative has been
implemented > make the intangible more tangible.

34
An example of a business case
Stage 1: Identify and Analyze Stage 2: Plan and Execute Stage 3: Evaluate and Improve
1. Identify Challenges 4. Specify Requirements 8. Demonstrate Impact
• Resources: HR team, training budget •
• High turnover, low productivity Improvements: Higher satisfaction, retention
• Allocation: HR management, training time • Quantify: 20% turnover reduction, 15%
2. Gather Data
• Cost/Benefit: Training vs. reduced turnover productivity increase
• External costs, internal surveys • ROI: Cost savings, efficiency gains
5. Timeline
3. Pinpoint Causes • Months 1-2: Surveys 9. Clear Takeaways
• Career development, management • Months 3-4: Develop training • Resolves key challenges, improves
practices performance
• Months 5-6: Implement
10. Future Vision
6. End Date & Follow-Up
• End Date: Month 6 • Engaged, productive workforce

• Follow-Up: Quarterly check-ins, metrics

7. Senior Leaders’ Roles
• Support: Endorse, provide resources
• Monitor: Ensure alignment, address issues

35
A simplified example of
Stage1/2

36
Comparing ethnicity across two functions in a UK bank

• A big UK bank would like to test if there are significant differences between the means of
BAME (Black, Asian, or Minority Ethnic Employees) prevalence in teams from two functional
groups.
• They believe that there is a connection between the BAME prevalence the teams may be
depended upon the function.
• Rumor is that BAME individuals are clustered into the particular profession because they feel
that others are closed-off to them and some profession in the organization do not make
BAME individuals very welcome.
• The HR department has found a recent study done by Race for Opportunity that suggests
that banking is one of the professions that BAME individuals often felt was close to them.

37
Comparing ethnicity across two functions in a UK bank

• Business problem: The head of the HR department put you in charge to check if there is
a difference in the prevalence of BAME and proportion of male teams in Sales vs
Professional service (Product development, Marketing, HR, Finance etc).

38
Comparing ethnicity across two functions in a UK bank
1. DepartmentGroupNumber (Team/unit key identifier)
2. GroupSize (Number of employees in the unit)
3. PercentMale (Percentage of the unit made up of males)
4. BAME (Percentage of the unit made up of Black, Asian, or Minority Ethnic Employees)
5. NumberTeamLeads (Number of team leaders in the unit – in general this organization has a team leader for every
seven or so employees)
6. NumberFeMaleTeamLeads (Number of teams leads in the unit who are female)
7. Location (Geographical location: 1 = Central London, 2 = Greater London, 3 = Rest of UK)
8. LondonorNot (Geographical location recoded to 1 = Central London or Greater London, 2 = Rest of UK)
9. Function (Function, 1 = Sales staff (costumer-facing people) or 2 = Professional Service (non-customer-facing
people)
10. EMPsurvEngage_1 - EMPsurvEngage_1 (Proportion of the unit who answered favorably – with either “agree” or
“strongly agree” on a 1 to 5 “strongly disagree” to “strongly agree” response scale – to each of nine employment
engagement questions: I work with full intensity in my job; I try my hardest to perform as well as I can in my work; I
concentrate hard in my job; My mind is always focused on my job; I strive very hard to complete my work well; I
would recommend this organization as a place to work; My job inspires me; Time flies at work; I am proud of the
work I do)

39
Comparing ethnicity across two functions in a UK bank

11. EMPsurvEngagement (Average unit/team score for the nine engagement questions)
12. EMPorgIntegrity1 – EMPorgIntegrity5 (Proportion of the unit who answered favorably - with
either “agree” or “strongly agree” on a 1 to 5 “strongly disagree” to “strongly agree” response
scale - to each of five employee ‘organizational integrity’ questions: My organization listens to
staff; My organization does the right thing; My organization treats external stakeholders with
respect; My organization encourages staff to point out questionable practices; We all do the right
thing in this organization)
13. EmpSurvOrgIntegrity (Average unit/team score for the five organizational integrity questions)
14. EMPsurvSUP1 – EMPsurvSUP4 (Proportion of the unit who answered favourably - with either
“agree” or “strongly agree” on a 1 to 5 “strongly disagree” to “strongly agree” response scale - to
each of four “supportive supervisor” questions: My supervisor listens to me; My supervisor cares
about my well-being; My supervisor helps ensure we get things done; My supervisor is there
when he/she is needed)
15. EmpSurvSupervisor (Average unit/team score for the four ‘supportive supervisor’ questions)

40
Continu
ous DV Explores whether two
(Interval Categoric groups of data differ in
or Ratio al-Binary Independent some dimensions. E.g., are
– and IV Group sample t-test males more stressed than
someti A versus females? Are females more
mes Group B engaged than males?
Ordinal*
)

41
 If Levene’s F-test statistic produces a significance value of greater than 0.05,
then we can assume that the functional groups have similar patterns of spread
or variance in team BAME prevalence, or they have ‘equal variance’ (that
they are not significantly different in their patterns of variance). However, if
Levene’s F-test statistic produces a significance value of less than 0.05 (p<0.05),
then the functional groups that we are comparing can be considered to have
very different variation in team BAME prevalence; the functional groups
1 do not have ‘equal variance’.

42
Based on the results, what can you suggest to the organization?

• Few options to start your thought process:

• As HR professionals, the next step might be to recommend interviewing BAME
individuals in the Sales teams about their experiences. Depending on the outcome and
assuming that you find some evidence of potential bias, you might recommend
interventions such as reviewing approaches for recruitment, retention, and development
of BAME individuals in the Sales function, or diversity training such as unconscious bias
and inclusive leadership.

43
Summary

• Information sources, how to measure, analytical software

• Big data
• Make a business case

44