@CIPD_Events | #HRanalytics17
Sponsored by:
Using People Analytics to Solve
Complex Business Issues
Jordan Pettman
Nestle
CHF 89.5 billion in sales in 2016
328,000 employees in over 150 countries
418 factories in 86 countries
Over 2,000 brands
1 billion Nestlé products sold every day
Nestlé at a glance
16.11.2017
Enhancing quality of life
and contributing to
a healthier future.
Our purpose
16.11.2017
The Nestlé story
2011
2012 2010 1990s 1970s 1947 1929
2000s 1960s 1980s
1867
Henri Nestlé
1905
1866
George H. Page Anglo-Swiss Condensed
Milk Company
2014 1938
16.11.2017
Nestlé
Globally Managed
Businesses
Zones
Markets
SIMPLE COMPLEXITY
16.11.2017
Ensure that:
1. All Nestlé businesses have access to and the capability to work with People Analytics to drive better people decisions
2. Analytics are useful and actionable in the business
3. Any Nestlé business can engage with People Analytics
OUR CHALLENGE
PEOPLE ANALYTICS @ NESTLÉ
Supporting the execution of the people
strategy through tools and frameworks
incl. Strategic and Operational People
Planning
Measuring the impact of HR intervention
and megatrends
Using statistical models and reasoning to
address top workforce concerns
Measuring progress against strategic
priorities through dashboards and
automation.
Managing the compliance with and
governance of change process to Nestlé’s
People Data Standards
16.11.2017
Global Standard People Dashbaords
Globally (191 countries) available self
service People Dashboard Catalogue
• 3 Reporting structures (Global,
Geographical Zone & Market)
• X 5 Categories: Headcount, Diversity &
Inclusion, Talent Management, Hiring and
Turnover.
= 105 Dashboards
Standard Dashboards are available by
Dimensions:
• Location (and type), Business, Function,
Job Level and Managerial Status,
Employee Group and Tenure
16.11.2017
SAP WebBI:
Illustrative Charts
PMEI is a research-based index based on
findings from our engagement survey,
leadership competencies, and external
research.
PMEI highlights actionable behaviors that
positively impact our workforce on
enablement, performance, and retention.
People Manager Effectiveness Index
Linking Engagement and Business Data for leadership development in Nestlé USA
The Legacy of Leaders Program is a peer-to-
peer discovery process for managers led by 2-
3 high performing leaders from NUSA.
This PMEI-based program showcases best
practices from NUSA’s high performing
leaders around team culture, managing
change, coaching, and performance
management.
STRATEGIC IMPERATIVE
Developing leaders at all levels is critical to our success at Nestlé USA and as
such is a key element of our HR operational strategy.
Legacy of Leaders
Statistical relevance of PMEI behaviors
Following the validation work done
by NUSA, Nestlé Global
Headquarters conducted an analysis
with:
• 700 leaders from 189 countries
representing all businesses and
categories
• 2016 completely anonymized
engagement survey data
• Outcome variables from 2012-2016
Validation of PMEI
This study showed that the highest
scoring leaders outperform their peers in
many important areas, including:
• Business growth and financial performance
• Gender diversity
• Development of key talent, more likely to
be on succession plan
• Retention
Key Findings
What this means for Nestlé: We can tailor leadership development programs based on a
targeted and specific set of behaviors that are correlated with business objectives.
Nestlé’s Reward Journey to Maturity
Creation of
Operational
C&B (2009)
Creation of Total
Rewards Policy (2012)
Creation of Operational Manual
to help C&B Managers to
Interpret TRP (2012)
Creation of Maturity
Profile (2011)
Creation Induction
Program to Teach New
C&B Managers (2010)
Evolution of Nestle
Salary Review Mgmt
Tool NSRM (2013)
Argumentation for
better PE
Distribution (2013)
Creation of Total
Rewards Toolkit (2014)
Living
Wage
Project (2014-16)
Nestle Total
Rewards System 2015-2016
Creation of
TR
Capabilities
(2016-2017)
Pay Analytics Toolkit
(2017)
Total
Rewards as
Strategy
Topic (2015-
Present)
Creation of
Teaching Site
for Managers
(2014-2016)
Measuring Reward Maturity
• Rewards Maturity has positive correlation with
• Performance distribution levels
• Leadership Effectiveness scores
• Rewards scores
• Profitability
Drug treatment studies
Active agent
Placebo
Survival time
Formal definitions
- Time: ‘What is the start of follow up time?’
- Event: ‘How do we define a fail?’
SURVIVAL ANALYSIS
CAN WE PREDICT WHO IS LIKELY TO LEAVE THE COMPANY?
All employees will ultimately leave, so the question is not whether, but when? Survival analysis generates conditional probabilities – the hazard rate – the instantaneous risk that
employees will quit during a given time interval.
all relevant events in the life of an employee
Demographics
age
gender
education
commuting
distance
marital
status
past work
experience
Events during employee’s career
promotion
performance review
international
assignment sabbatical leave
training
salary increase
External influences
business
performance
external
opportunities
team dynamics
Input for statistical analyses
- Mantel Haenszel log rank test
- (Extended) Cox Proportional Hazards
model
- Survival trees and random forests
SURVIVAL ANALYSIS
Sample of Night’s Watch Rangers, Stewards and Builders in Westeros
N = 335
A (NOT SO) REAL EXAMPLE
ID Name Gender Origin Age (at T0) Marital Status Training
Period Position
Intelligence
Score Recruitment Encounters
Size of
Enemy
Army
Event
Reason
for
Leaving
T1 T2
105 S T Male The Reach 24 Single 3 Steward 120 Forced by
father 2 20 000 0 NA 0 12
105 S T Male The Reach 24 Registered
Partnership 3 Steward 120
Forced by
father 4 40 000 1
Become a
Maester 12 22
106 O Y Male White Harbor 31 Single 10 Builder 90 To avoid
punishment 1 20 000 0 NA 0 12
106 O Y Male White Harbor 31 Single 10 Builder 90 To avoid
punishment 2 40 000 0 NA 12 20
106 O Y Male White Harbor 31 Single 10 Builder 90 To avoid
punishment 4 60 000 0 NA 20 32
106 O Y Male White Harbor 31 Single 10 Builder 90 To avoid
punishment 6 100 000 0 NA 32 39
A (NOT SO) REAL EXAMPLE
Non-parametric estimator of
the survival curve
The fraction of employees that is still
with the company for x amount of
months after joining
28
A (NOT SO) REAL EXAMPLE
Comparing subgroup differences
A (NOT SO) REAL EXAMPLE
𝜆 𝑡 𝑋𝑖 = 𝜆0(𝑡) ∙ 𝑒𝛽1𝑋𝑖1+⋯+𝛽𝑝𝑋𝑖𝑝
where 𝑿𝑖 = 𝑋𝑖1, … , 𝑋𝑖𝑝 are the realised values of the covariates for subject 𝑖.
Position
Reference: Builder
Location Reference: King’s Landing
Marital Status Reference: Married
Steward: 0.91
Winterfell: 0.42
Dragonstone: 0.40
The Reach: 0.80
Baseline Hazard Single: 1.22
Interpretation: the hazard rate for
Rangers is 4.1 times higher than
for Builders
Ranger: 4.10
to other techniques (e.g. logistic regression)
- Flexible:
Can handle time varying covariates as the characteristics of an employee can change
during the follow-up period
- Competing events:
Can model cause-specific hazards
SURVIVAL ANALYSIS
A (NOT SO) REAL EXAMPLE
• Independent Censoring: the subjects who are censored at time t should be representative of all the
subjects in that subgroup who remained at risk at time t with respect to their survival experience.
– This assumption is fulfilled when the end of follow-up period is defined.
• Proportional Hazards: the survival curves for 2 strata must have hazard functions that are proportional
over time (i.e. constant relative hazard).
• Stratifying
• Interaction with time
Risk of turnover within 12, 24, 36 months
Identifying dominant themes in a vast array of documents
Each document is a mixture of topics and each word is
attributable to one of the document’s topics
Example: Free text in Exit surveys
TEXT ANALYSIS
Identifying dominant themes in a vast array of documents
Fictional results
Data: free text answers to the question «Why
did you decide to leave the company?»
Result:
Topic 1: Family reasons
Topic 2: Career advancement
TEXT ANALYSIS
PAY ANALYSIS
HOW DO WE MONITOR FOR FAIR PAY PRACTICES
Pay equity analysis checks the current payroll against the compensation philosophy.
Interactive, web-based application that enables local C&B to investigate Pay Equity through linear
regression.
Create applications directly from R
No web development skills needed
Users can interactively look at results from the analysis
PAY ANALYSIS
Input year and market of
interest
2
Take a look at the descriptive
plots of the data
3
Decide which variables
you want to include in
your model
4
Interpret the
results
6
Check the quality
of the model
5
Template file is provided
1
performance
pay
job level
seniority
work experience
PAY ANALYSIS
Which variables should be included in the regression model?
Control variables Independent variable Dependent variable
Accounting for these is this correlated with this outcome?
Aimed at improving the predictive ability
of the model
Diagnostic plots guide the user to decide
whether the model is adequate
QQ-plot checks for normality of the error terms
PAY ANALYSIS
PAY ANALYSIS performance
performance
reference high
Predictor Estimate 95% Confidence Interval
High Performance 1.013 [0.981; 1.047]
High performers earn 1.013 times what the reference category earns.
When controlling for the other variables in the model, the difference in pay
between high performers and the reference group is not statistically
significant.
Merci
Questions?
Feedback?
Get in touch!
: @jl_p
: linkedin.com/in/jordanpettman