P 01 advanced_people_analytics_2016_04_03_v11

Predictive Analytics World for Workforce | April 4-6 2016 | San Francisco

1

HR Science Vishwa Kolla Head of Advanced Analytics | John Hancock Insurance

Predictive Analytics World for Workforce | April 4 – 6 2016 | San Francisco

meets

2

In God we trust.

All others – please bring me data

- W. Edwards Deming


3

Vishwa Kolla Head of Advanced Analytics

John Hancock Insurance, Boston

MBA Carnegie Mellon University

MS University of Denver

BS BITS Pilani, India

Advanced Analytics

CoE, Maturity Model

Customer /

Workforce Analytics

(entire value chain)

Machine Learning

Scoring Engine

Optimization

Simulations

Forecasting & Time

Series

• 15+ Years

• John Hancock Insurance

• Deloitte Consulting (Industries –Insurance,

Retail, Financial, Technology, Telecom,

Healthcare, Data)

• IBM

• Sun Microsystems

Business Analytical (Math, Stats)

Technical (Programming)

Expertise

Experience




4

What are your firm’s biggest assets?



5

Finance

Finance will probably say Product



6

Marketers

Marketers will probably say Customers



7

Finance Marketers

In reality, it is your employees



8

To win in the marketplace, first win in workplace

IQ (1x)

EQ (2x)

RQ (5x)

Products

Sales

Customer

Experience

Productivity


9

Lack of direction, not lack of time, is the problem.

We all have 24 hour days

- Zig Ziglar



10

Where should we focus?

Acquire Nurture Retain



11

Retention is where most initiatives start



2x - 6x less

expensive to retain

than to hire


12

Acquisition kick starts the journey



Cost of a bad hire is

3x that of a good

hire (read cultural

damage)


13

Nurture is often the less investigated of all areas



Productivity gains

are highly

correlated to

engagement

14

Often, the best solution to a

management any problem is

the right person

- Edwin Booz


Home Run

Total Potential

• Height

• Weight

• High School / College Home Runs

• Home Park Layout

• History Of Home Runs

Running Back

Draft Potential

• 40-Yard Dash Time

• Total Rushing Yards in College

• Total Touchdowns in College

• # of Heisman Trophies /

Championships University Earned

Historically

Shots Blocking

Potential

• Height

• Arm length

• Hand size

• Position


15

Applying AA in acquisition is natural in sports


Target

Variable Predictors

Performance

Measurement

is intrinsic

Large sample

sizes

Demonstrated

Value

https://thenounproject.com/term/euro/61335

Predictive Model Build Process


16

Applying AA in acquisition can get expensive

9 – 36 mos.; 4-5 ppl.; $1-2 M; Repeat for BU / Function


Challenges

Time consuming

Expensive

Limited

Measurability

Amount of

customization

Cumbersome

Abandon

Collect Data Score Interview

Predictive Models Nudges

Verify

Build a

6 Person

AA team Predictive Analytics World for Workforce | April 4-6 2016 | San Francisco

17

Applying A can be practical (for a small shop)


Life

Insurance

Advanced

Analytics

(AA)

Experience

Levels

Insights from Work History

Insightful Summary

Succinct Project description

Use of quantification

Consistent structure

Buzz : Real work ratio

Job change frequency

Time at any position

Progression history

Proximity to Workplace

Capability index

Tier 1 / 2 / 3 Schools

Insights from Work Product (Resume)

Spacing between sections

Appeal of layout

# Punctuation issues

# Grammatical errors

# Positive words

# Theme repetitions

Insights from Open Ended Questions # Impactful Initiatives

Fit in Analytical spectrum

Listening index

Coach-ability index

Work Ethic

Role –

Leader /

Talker /

Thinker /

Doer

Personality

18

The Inspired Inspire


Engagement Campaigns

• Star Performer Appreciation

• Star Group Activity

• Quarterly Outing

• Annual Holiday Party


19

Engagement improvement needs re-thinking

1 2 3 4 5

Performance

Nu

mb

er

of

Pe

op

le

Improved

Engagement Challenges

• Not enough lift

in scores

• Not timely

enough

• Not relevant

enough



20

It is all about who we interact with & how much


Management

Consulting

Hours

on

Project

Individual, Peers

How much, how long

Time of day, month, year,

entry / exit from project

Where (home | away)

Performance history

Time-off(s)

Life stage

Personality

Project

Size, Budget, Duration,

Location, Timing, # of

functions, expenses

Synthetics

Definition and comparison

to peers

Senior / Executive

leadership to Staff ratio

etc.

Non

Professional

Service

Industries


21

Network analysis gets to the heart of the issue


Actors

Interaction(s)

Number

Speaking time

Average speech segment length

Variation in speech energy

Variation in movement

Self perceived dominance

Actors

Interaction(s)

Number

Time of interactions

Length of message

Attachments

Subject categorization

Number of conversations

Data Collection

Identify Unit of

analysis

Curate (Collect,

De-identify,

Cleanse)

Merge

Repeat each

time period


22

Nurture & Retention are big data problems

Internal Data (90%)

External Data (10%)

Data Engineering (Create Longitudinal View) Predictive Models

Profiles on

variety of

dimensions

Engagement

Index

Likely to get

promoted

Likely to attrite

Customer

Project

Point in Time

Snapshot

What data should I keep?

1Q Look back

2Q Look back

3Q Look back

4Q Look back



23

Insights from use case were eye opening


Data collection timeframe defines action ability

period

Life / career stages make recommendations counter

intuitive – e.g., travel

High burn projects were good (for younger

population), and with time-off

Network effect (individuals consistently on projects

comprised of more stars had higher risk of attrition)

Blogging is good

Some voluntary attrition is good

Relevant Data Set

24

Getting to the finish line involves careful planning

and execution

Core Inputs (Model Build)

Historical Data

Raw Data

Additional Inputs (Test)

Modeling Data Set

Core Inputs

(Model Build)

Additional Inputs (Test)

Valida

te

Test Train Relevant Data Noise

Da

ta P

art

itio

nin

g

Da

ta E

xtr

ac

tio

n

Da

ta E

ng

ine

erin

g

Ap

ply

Filt

er

Ru

les

Da

ta A

gg

reg

atio

n

Predictive Model Build Scoring Engine Development Live Scoring Engine

Ev

alu

ate

Fin

al M

od

el E

qu

atio

ns

Ro

ll o

ut

to P

rod

uc

tio

n

Data

Integration

Model

Integration

Systems

Integration

Real – time Scoring Engine Development

Service Layer Development

UI Engine QC Engine

Business Objective – Any Predictive Model

1

2

Un

i -V

aria

te A

na

lysi

s

Bi-V

aria

te A

na

lysi

s

3

4 5

Problem

Definition

Model

Strategy

Data

Engineering

Model

Build

Model

Implementation

& Governance

1 2 3 4 5

01/## Current



25

Closing


Advanced People Analytics is for real and not “entirely” hype

Work closely with Business

Prioritize Process over immediate Purpose

A structured process is critical

There is no pixie dust

QC every step along the way


26

THANK YOU! Predictive Analytics World for Workforce | April 4 – 6 2016 | San Francisco


27

Credits

The Noun Project

Humanyze

MIT Media Lab


P 01 advanced_people_analytics_2016_04_03_v11

Data & Analytics