On October 2nd, Peter Beeusaert of SD Worx and Pieter Van Bouwel of Python Predictions shared our collaboration on Intuo.io’s Talent Enablement Conference. In this presentation, they explained how some of the most powerful data science applications have clear value in an HR context.
They discussed a recent joint use case ‘How data science can segment your employees to propose differentiated absenteeism treatment’ to demonstrate how bringing together data science and HR experts can improve and facilitate HR decision making.
4. STRUGGLE TO GENERATE MAX VALUE OUT OF DATA
Metrics
Benchmarks
Scorecards
Surveys
HR Reporting
Added Value
HR Intelligence Maturity
Source: Investing in People, Wayne Cascio & John Boudreau, Dec 8, 2010, FT Press, adapted by SD Worx & Python Predictions
In HR, there seems
to be a ‘wall of
Boudreau’ – most
companies stop at
HR reporting and do
not engage in data
science yet
5. TEXT ANALYTICS
So which typical data science applications
could work in an HR context? We could
analyse textual information, for example
to flag issues in employee satisfaction or
classify CVs according to skills
6. RECOMMENDER SYSTEMS
We could use recommender systems to
make personalized suggestions of
relevant trainings for every employee
7. PREDICTIVE ANALYTICS
Or we could predict specific events, such
as future absenteeism (see our previous
case with SD Worx here).
8. HOW TO GET THERE
Metrics
Benchmarks
Scorecards
Surveys
HR Reporting
Added Value
HR Intelligence Maturity
Correlation
Causation
Prediction
HR Analytics
Source: Investing in People, Wayne Cascio & John Boudreau, Dec 8, 2010, FT Press, adapted by SD Worx & Python Predictions
So we should probably move
beyond reporting, to analytics
9. SEGMENTATION
We’re sharing a concrete customer case
around segmentation – in terms of
absenteeism, we should not treat every
employee as if he / she is similar to every
other employee. Segmentation helps to
understand differences in absenteeism
10. Average 'direct
cost' of
absenteeism in
Belgium:
€ 1.000
per employee per
yearAbsenteeism is of no interest to anyone:
Not for the employer, but certainly not for the employee!
ABSENTEEISM IN BELGIUM
The past 10 years, total absenteeism increased by 40%
4.15%
5.81%
2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
+40%
11. CLASSICAL APPROACH
Drawing up an absenteeism policy
Company X
❑ Notification procedure
❑ Notification call
❑ Medical certificate
❑ Contact while absent
❑ Control doctors
❑ Returning chat
❑ Reintegration
❑ Industrial doctor
❑ Reintegration agreements
❑ Frequent absenteeism
interview
❑ Absenteeism coach
❑ Follow-up interview
❑ Signal call
❑ …
Most
companies
have an
absenteeism
policy (‘one
size fits all’)
14. ETHICS
● Surveys: inform and explain the goal to employees
● Individual information will NOT be shared to managers or
the end customer
● Information is anonimised
● Items got an ethical/legal check ➔ even if they are relevant
from an analytical point-of-view, they are removed if there’s
a nogo!
Of course there are some crucial
ethical implications in an HR context
15. SEGMENTATION
In segmentation, we aim to
create groups of employees that
are similar to each other, but
where the groups are different
We have analyzed the
employees in a specific role in
three companies in the same
industry
16. EXPLORATORY Final Segments
Project definition
Collect data
Build clusters
Profile clusters
Business review
1
2
3
4
IMPORTANT TO HAVE A GOOD METHODOLOGY
Segmentation
is an iterative
process
17. FACTORS THAT IMPACT ABSENTEEISM
Workload Engagement
Performance &
Talent reviews
Employee
Characteristics
Absence in
the past
Employee
Survey
Manager
Survey
Payroll
Data
The data we used in the project
19. PROFILING EXAMPLE CLUSTER X
19
Physical charge
Pleasure & energy
Quality of work
Too heavy Acceptable
Low High
Bad Good
An example of
how we profile
the resulting
segments
21. HOW TO ALLOCATE EMPLOYEES?
Can we assign new hires or employees without survey info to
one of the clusters?
Absence
Evaluation
Clustering 4
??
??
?? ????
22. DEFINE RULES
Find key questions that can classify employees
Absence
Evaluation
SURVEY
❑ Personality: politeness
❑ Work motivation
❑ --------------------------------
❑ --------------------------------
❑ Physical workload
❑ --------------------------------
❑ Dutch speaking
❑ --------------------------------
❑ --------------------------------
❑ --------------------------------
❑ Job flexibility is important
❑ --------------------------------
❑ --------------------------------
❑ ...
Low
absence
High
absence
Example: ‘because I can choose my own hours’
We will show the detailed profile of one
segment that suffers from absenteeism
23. Reason for this job?
I like it!
Good work/life balance
I can choose my own schedule
I can work close to home
Speaks Dutch
V
V
V
V
Acceptable time on the road?
I always have care for my children
PROFILING SEGMENT: ‘BECAUSE I CAN CHOOSE MY SCHEDULE’
I speak with
I’m in job content
Politeness
W o r k i n g speed Work precision
For this segment, it is
crucial that they can
choose their own
working schedule, since
they don’t always find
care for their children
pride of my work
24. Strategic level
Operational level
Tactical level
The segmentation forms the basis of
an action plan, that has benefits on
different levels
Monitoring size
and evolution of
the segments
Prioritizing and
designing solutions
for each segment
Helping teamcoaches
to understand these
differences
25. KEYS TO SUCCESS Cooperation between analysts and domain experts
Project definition is crucial
Realise the model’s power and limitations
Complex algorithms are not the key to success
Data and ethics can go hand in hand
Not data volume, but the right data
26. BREAK DOWN THE WALL
Time to break down the
‘wall of Boudreau’ – we
hope we inspired you to
make more use of data
in an HR context