- 1. This document contains information and data that AAUM
considers confidential. Any disclosure of Confidential Information
to, or use of it by any other party, will be damaging to AAUM.
Ownership of all Confidential Information, no matter in what media
it resides, remains with AAUM. AAUM Confidential Analytics for
HR
2. - 1 - Corporate profile Founded by IIT Madras alumnus having
extensive global business experience with Fortune 100 companies in
United States and India having three lines of business Prof Prakash
Sai Dr. Prakash Sai is professor at the Department of Management
Studies, Indian Institute of Technology Madras. He has wealth of
international consulting experience in Strategy Formulation Puneet
Gupta Puneet spearheads the IFMR Mezzanine Finance (Mezz Co.), is
strengthening the delivery of financial services to rural
households and urban poor by making investments in local financial
institutions. Padma Shri Dr. Ashok Jhunjhunwala Dr. Ashok
Jhunjhunwala is Professor at the Department of Electrical
Engineering, Indian Institute of Technology Madras India. He holds
a B.Tech degree from IIT, Kanpur, and M.S. and Ph.D degrees from
the University of Maine, USA. Analytics Appropriate statistical
models through which clients can measure and grow their business.
Competitive Intelligence Actionable insights to clients for their
business excellence Livelihood Services ranging from promotion of
livelihoods, implementation services, livelihood & feasibility
studies. Key Focus Areas in Advanced analytics and Predictive
analytics Product geniSIGHTS (Analytics/BI), Ordo-ab-Chao (Social
Media) More than 25 consulting assignments for Businesses &
Govt orgs Partnership Actuate, IIT Madras, TIE and 3 strategic
partnerships Dedicated corporate office at IIT Madras Research park
since 2009 Aaums office, IIT Madras Research Park 3. - 2 -
Competencies in Advanced analytics Build appropriate statistical
models through which clients can measure and grow their business.
Expertise in Digital Media Finance/Insurance Retail Entertainment
Human Capital Government organizations Research & training
Competitive assessment Competitive intelligence Provide actionable
insights to clients for their business excellence. Expertise in
Business Entry Business Expansion Market research Livelihood
Perform livelihood services ranging from promotion of livelihoods,
implementation services, livelihood and feasibility studies.
Expertise in Government organizations Non Government organizations
Corporate with livelihood focus Research 4. Objective Metrics
definition Analysis & Key insights Predictive modeling Way
forward 5. - 4 - Objective Client management approached Aaum to
develop comprehensive metrics to be standardized for its clients.
Aaum team developed standard metrics that could be rolled out
across clientss customers to derive appropriate corrective actions
and preventive actions performed interaction analysis for key
metrics to derive a holistic understanding developed predictive
models for some very useful parameters predicting employee
productivity based on leave parameters derived insights based on
the performance of the RO. 6. Objective Metrics definition Analysis
& Key insights Predictive modeling Way forward 7. - 6 - S.No.
Metric Definition Interpretation 1. a. Productivity in days Total
number of working days of the employee/ total number of expected
working days This metric will range from 0 to 1. The closer the
value is to 1, the higher the productivity. 2. Average leave day
Average of the difference between leave reporting date and leave
availed date. Ideally this metric should be closer to zero. A
higher deviation implies the gross indiscipline 3. Swipe
Indiscipline (including OD) No. of days not swiped/ Total working
days (including OD) This metric will range from 0 to 1. Value
closer to 0 implies a good scenario. 4. Swipe Indiscipline
(discounting OD) No. of days not swiped/ Total working days
(excluding OD) This metric will range from 0 to 1. Value closer to
0 implies a good scenario. 5. Swipe OD discipline No. of OD
swiped/Total number of OD This metric will range from 0 to 1. The
closer the metric is to 1, the better it is. 6. OD indiscipline No.
of rejected OD/Total number of OD This metric will range from 0 to
1. Value closer to 0 implies a good scenario. 7. Regularization
Rejected (RR) No. of rejected regularization/Total no. of
regularization requested This metric will range from 0 to 1. Value
closer to 0 implies a good scenario. Metrics definition 8. - 7 -
Metrics definition cont S.No. Metric Definition Interpretation 8.
Leave Discipline Total no. of leave accounted/ Total absent days
This metric will range from 0 to 1. The closer the metric is to 1,
the better it is. 9. Leave affinity Total days of leave
taken/Frequency of leave This metric defines the no. of leaves per
installment. 10. Effort Variance (Actual hours worked-Ideal hours
to be worked) / Ideal hours to be worked A positive value of the
metric indicates over productivity. A negative value indicates
under productivity. 11. Attrition No of people leaving the
organization/Total count An organization would always wish
attrition to be close to Zero 12 Forming No of people leaving the
organization with in 20 days/Total count Forming cost A metric that
qualifies these severity of an employee leaving with in 20 days
Norming No of people leaving the organization with in 90 days/Total
count Norming cost A metric that qualifies the situation where in
the employee leaving between the period 20 days - 90 days
Performing No of people leaving the organization after 90
days/Total count Performing cost A metric that qualifies the
situation where in the employee leaves the organization after 90
days Attritionchildcosts 9. Objective Metrics definition Analysis
& Key insights Predictive modeling Way forward We have come up
with Spread metric to qualify the dispersion of the data. The
metric in comparison with the central tendency (mean or median)
measure throws lights on how well the data is represented at unit
level versus the overall metric. 10. - 9 - Metric 1. Productivity
in days Region Unit Pay grades Departments Main effects N, Spread
N, Spread N, Spread Productivity at location 4, 5 as well as
productivity at pay grades 7,8 and T are lower. The closeness of
the spread and productivity metric of location 4 reveals that the
metric best mimics the overall productivity metric. The spread
metric of the printing dept, which is 2.8 times greater than the
overall spread, and its low productivity implies the presence of
many influential observations with a very high dispersion values
(outliers) has brought the productivity level significantly down.
11. - 10 - Metric 1. Productivity in days Main effects Location Age
group Experience N, Spread N, Spread Shift Spread Spread Location
Productivity of freshers as well as those between age 18 -25 are
much lower. Productivity at location 2 is much higher than any
other location. The spread metric of Gen shift which is 2.79 times
higher than the overall spread in comparison with its productivity
metric which lies in sync with the overall productivity implies the
presence of few outliers which tend to bias the metric. . 12. - 11
- Metric 1. Productivity in days Interactions Location-Pay grade
Department-Pay grade Productivity of T grade employees in Unit 5
and Seamless department depict a relatively lower productivity. The
closeness of the spread and productivity metric of T grade
employees in Location 5 to the overall values shows how closely the
data mimics the overall data. . N, Spread N, Spread 13. - 12 -
Metric 2. ALD Main effects Region Department Age Experience
Accounts, Admin department and experienced category of 1-3 yrs show
maximum indiscipline in ALD 14. - 13 - Metric 2. ALD Main effects
Unit Pay grade Head office, Pay grade 1, T and 8 are the most
critical segments as far as leave reporting discipline is
concerned. 15. - 14 - Metric 2. ALD Age-Experience Unit-Pay grade
Location-Pay grade Interaction effects reveal that employees in
Head Office and regional team with pay grade 4 and Corporate and
Strategy with pay grade 2 are the segments where leave report
indiscipline is very high. Location-Pay grade Interactions 16. - 15
- Swipe Indiscipline (including OD) Swipe Indiscipline (including
OD) Swipe Indiscipline (including OD) Swipe Indiscipline (including
OD) Region Departments Pay grade Metric 3. Swipe Indiscipline (+OD)
Unit Main effects Location1, Printing and Seamless departments, Pay
grade T exhibit high swipe indiscipline while Marketing department
exhibits low swipe indiscipline. 17. - 16 - Shift Location Age
Experience Metric 3. Swipe Indiscipline (+ OD) Main effects Swipe
Indiscipline (including OD) Swipe Indiscipline (including OD) Swipe
Indiscipline (including OD) Swipe Indiscipline (including OD) The
younger age group displays higher indiscipline as compared to the
older proportion. Employees with 6months-1 year experience, Shift C
and Location 1 are other categories displaying high swipe
indiscipline. 18. - 17 - Metric 3. Swipe Indiscipline (in OD)
LocationPay grade Dept-Unit Experience-Pay grade Age-Experience
Interactions Pay grade-Age Further drill down to interaction
effects reveal, swipe indiscipline is particularly noticeable in
Admin & Support department of Location 5, youngest age group of
pay grade 7 and employees with experience 1-3 of grade T. 19. - 18
- Region Pay grade Department Unit Metric 4. Swipe Indiscipline (-
OD) Main effects Head office and regional team, pay grade 4 and IT
department show high swipe indiscipline. 20. - 19 - Age Shift
Location Experience Main effectsMetric 4. Swipe Indiscipline (- OD)
The middle age group of 25-45, employees with experience ranging
from 6months to 3 years, Headoffice shift and employees belonging
to Location 2 depict high swipe indiscipline. 21. - 20 - Unit-Pay
grade Age-Pay grade Age-Location Dept-Pay grade Metric 4. Swipe
Indiscipline (- OD) Interactions However interaction effects reveal
that Head office and regional team employees of pay grade 4, IT
department employees with pay grade 5, middle aged proportion with
pay grade 4 and experience of 1-3 years are the critical categories
in terms of high swipe indiscipline. 22. - 21 - Region Pay grade
Department Metric 5. Swipe OD Discipline Head office & regional
team show good swipe discipline for OD, while Location 3 and 6
employees depict low discipline. Main effects Unit 23. - 22 - Age
Shift Location Experience Metric 5. Swipe OD Discipline Main
effects The swipe OD discipline decreases with increase in age.
Location 6, Gen shift, and 3-5 years of experience are the critical
segments in terms of swipe OD discipline. 24. - 23 - Region Pay
grade Department Unit Metric 6. OD indiscipline Main effects
Location 4, Commercial department and pay grade 7 are the segments
depicting high OD indiscipline. However marketing department and
Location 2,3 show low indiscipline. 25. - 24 - Age Shift Location
Experience Metric 6. OD indiscipline Main effects The younger
proportion of 18-25 age group and employees with experience 1-3
years are highly undisciplined in terms of OD. Corporate shift and
Location 4 exhibit the same trend. 26. - 25 - Region Pay grade
Department Unit Metric 7. RR Main effects Location 6, pay grade 8
and Seamless department are the segments which show high
regularization rejections. 27. - 26 - Age Shift Location Experience
Metric 7. RR Main effects Employees belonging to the age group
above 55, and experience with 5-10 years depict higher rejection
rate. 28. - 27 - Metric 8. Leave discipline Main effects Region Pay
grade Department Unit Grade T, Location 4 and Printing are the
critical segments that show poor leave discipline. 29. - 28 -
Metric 8. Leave discipline Main effects Age Shift Location
Experience Interestingly the leave discipline improves as age and
experience increases. Shift C and Location 4 show high leave
indiscipline. 30. - 29 - Unit-Pay grade Experience-Age
Experience-Pay grade Dept-Pay grade Metric 8. Leave discipline
Interactions Interaction effects reveal that Trainees as well as
the freshers maintain very low leave discipline compared to other
segments. 31. - 30 - Metric 9. Leave affinity Main effects Region
Pay grade Department Unit Leave affinity increases with decrease in
pay grade. Seamless and Tubeline show greater leave affinity. 32. -
31 - Metric 9. Leave affinity Main effects Age Shift Location
Experience Leave affinity decreases with increase in age. The
frequency of leave taking is high for the freshers. Employees with
5-10 years of experience, Shift C and Location 1 exhibits the
maximum leave affinity. 33. - 32 - Metric 9. Leave affinity
Interactions Unit-pay grade Pay grade-Experience Age
group-Experience Dept-Pay grade However interactions do not reveal
any interesting patterns. 34. - 33 - Metric 10. Effort Variance
Main effects Region Pay grade Department Unit Pay grade 7, Location
5 and Engineering department are the overstretching segments in
terms of working hours. Whereas the segments Management, Head
office & regional team and Admin & Support are under
performing in terms of the same. 35. - 34 - Metric 10. Effort
Variance Main effects Age Shift Location Experience The Effort
Variance increases with increase in Age and when the experience is
over 3 years. The critical segment here is the HeadOffice that
shows a negative Effort Variance 36. - 35 - Interactions Admin
& Support department of the Head office and regional team
depict very low effort variance as compared to the other segments.
However tying productivity and effort variance together, it was
noticed that the departments of Printing and Seamless which
exhibited low productivity in terms of days has actually
overstretched themselves to arrive at positive effort variance
implying that these departments are not actually underperforming.
Metric 10. Effort Variance Unit-Dept Productivity Categorized 37. -
36 - Metric 11. Attrition Main effects Region Pay grade Department
Unit Shift Location 6, Pay grade 7 and Departments Laminator and
Printing show very high attrition rate. 38. - 37 - Metric 12.
Attrition Child Costs June July Aug Sep Oct Nov Dec Jan June July
Aug Sep Oct Nov Dec Jan Printing Seamless Attrition costs are
benchmarked at a base level and compared over a period of time.
This graph illustrates month wise split of the attrition child
costs incurred by the two critical departments Printing and
Seamless. For printing department, the child costs are very high in
the month of June. Performing costs are nil for both the
departments in the month of August. 39. Objective Metrics
definition Analysis & Key insights Predictive modeling Way
forward 40. - 39 - Predictive modeling using Random Forests
Prediction is based on a sound understanding of the client business
and deriving insights from various data sources to explain the
underlying characteristics. We have built Random forests to develop
predictive modeling. Random forests are state of the art analytical
techniques that constructs several decision trees to arrive at
variables with predictive intelligence 41. - 40 - Pr_Day TRUE <
0.2 0.2 0.75 0.75-0.90 0.90-0.98 >0.98 < 0.2 4 0 0 0 0 0.2
0.75 0 181 0 0 0 0.75-0.90 0 0 331 0 0 0.90-0.98 0 0 0 131 0
>0.98 0 0 0 0 4 S,No. row.names %IncMSE IncNodePurit y 1
Department_new 7.7744527 0.784334305 2 ALD 9.4599482 0.897245267 3
SCorporate 0.4160107 0.073057995 4 SGen 9.0660538 0.340773951 5
SHeadoffice 3.8128584 0.070116793 6 SShift.A 9.1486368 0.346673313
7 SShift.B 9.3715632 0.410356133 8 SShift.C 6.9114632 0.337520495 9
SWeeklyOff 0.9108635 0.136731588 10 Not.swiped.including.OD
22.0522982 1.769871655 11 Not.swiped.when.present..TWDays. OD
1.2986346 0.038448390 12 Swiped.OD..OD.All 2.2035039 0.123827886 13
Rejected.OD..Total 2.8027272 0.095028156 14 Experience 9.7536757
0.638688390 15 Status 0.0000000 0.002368345 16 Pr_Effort.Variance
9.7623909 0.546977854 17 RR.Metric 2.8510718 0.154144889 18
LD.Metric 42.9440334 3.486118889 19 LAFF.Metric 12.2612422
0.715462753 20 Age.group 8.2907674 0.297730402 Predicted values
against the true classification Predicting the productivity of the
employee About 20 variables where chosen to construct a random
forest model on productivity parameter. Productivity is converted
to a factor variable consisting of 5 levels. i.e. < 0.2, 0.2
0.75, 0.75-0.90, 0.90- 0.98, >0.98 We have constructed a new
random forest model based on the high importance score value on the
previously constructed random forest. Cross segment ratio = 1
Inferences Our new random forest model indeed showed up 100 %
accuracy in classifying the productivity parameter correctly. The
variables retained in our model are Department_new
Not.swiped.including.OD,Experience LAFF.Metric LD.Metric This model
can be used to predict the productivity of the employee 42. - 41 -
Predictive modeling Prediction involves integrating various data
sources with a good sample strength. Previous slide demonstrates a
scenario of predicting productivity of the employee based on his
leave patterns, department and swipe discipline. The model is
specific to this particular department as the solution is optimized
for this data set Poor sample strength and unavailability of
dataset became a severe bottleneck in modeling some of the very
useful metrics e.g. Attrition! A few more possibilities are
discussed in the following slides and can be further explored with
the availability of data 43. - 42 - Objective: Attrition is a
plaguing problem in any company or industry. With intensive data
mining, it is quite possible to build models to better understand
the attrition patterns at employee level, department level, unit
level and organizational level Case illustrations: Predicting the
attrition at employee, unit, organizational levels Explanatory
variables: Information on the below mentioned datum collected over
a period of 3 -5 years. 1. Socio-demographic Age, sex, marital
status, location, education, place of schooling, previous
employment information. 2. Interview Short listing criterion,
position, pay grade, department, region, etc offered at the time of
the joining. 3. Time sheet Department, pay grade, unit, region,
location, shift, DOJ, RO details, productivity, swipe patterns,
etc. 4. Leave Leave patterns, regularization pattern, outdoor
pattern, affinity, ALD, RO details. 5. Performance Appraisal
scores, remarks, achievements, recognition, issues/concerns,
negative feedback, pay grade history, bonus info, project specific
info, promotions, etc. 6. Learning and Development Trainings
undergone, skill sets developed, etc. 7. Exit interview Reason,
issues, strengths, etc. Organizational benefits: Predicting
employee tenure based on his/her characteristics. Forecasting
attrition effects at project, department, unit & organization
levels. Root cause analysis and arrive at corrective/preventive
actions to bring down the attrition rate in critical departments
Can further be delved to determine the effectiveness of other
departments. e.g. Effectiveness of the training department.
Modeling approach: Structured data mining approach will be adopted
based on underlying characteristics of data. Logistic regression,
neural networks, support vector machines, decision trees, random
forests, etc would be administered to arrive at predictive models.
44. - 43 - Objective: Analyzing the past staffing, efforts at
project level, unit level, employee level and developing the future
requirements for existing/new projects based on appropriate
forecasting techniques. Case illustrations: Forecasting project
specific staff requirements Explanatory variables: Information on
the below mentioned datum collected over a period of 3 -5 years. 1.
Project Project detail, milestones, staffing efforts, etc. 2.
Customer feedback/satisfaction scores Customer remarks, issues,
suggestions, etc. 3. Time sheet Department, pay grade, unit,
region, location, shift, DOJ, RO details, productivity, swipe
patterns, etc. 4. Leave Leave patterns, regularization pattern,
outdoor pattern, affinity, ALD, RO details. 5. Performance -
Appraisal scores, remarks, achievements, recognition,
issues/concerns, negative feedback, pay grade history, bonus info,
project specific info, promotions, etc. Organizational benefits:
Forecasting staff requirements, understanding of lean/stressed
periods Equipping the departments/organization with the necessary
staffing requirements well in advance. Remove inefficiencies, delay
in the project delivery. Modeling approach: Structured data mining
approach will be adopted based on underlying characteristics of
data. Logistic regression, neural networks, support vector
machines, decision trees, random forests, etc would be administered
to arrive at predictive models. 45. - 44 - Objective: Training
programs are very important in building role based skill sets
(project specific) as well as the behavioral skills sets.
Organizations spend significant resources on this front but return
on their investment needs to be analyzed for building a successful
training & development. Case illustrations: Improving the
effectiveness of the organizations development & learning
programs Explanatory variables: Information on the below mentioned
datum collected over a period of 3 -5 years. Costs
Training/learning cost, resource requirement, etc. Project Project
detail, milestones,staffing efforts, customer score/feedback,etc.
Time sheet Department, pay grade, unit, region, location, shift,
DOJ, RO details, productivity, swipe patterns, etc. Leave Leave
patterns, regularization pattern, outdoor pattern, affinity, ALD,
RO, etc. Performance Appraisal scores, remarks, achievements,
recognition, issues/concerns, negative feedback, pay grade history,
bonus info, project specific info, promotions, etc. Learning and
Development Trainings undergone, skill sets developed, etc.
Organizational benefits: Analyzing the effectiveness of a training
program, performance of the employee before/after the training in
actual projects, project performance (customer satisfaction scores)
Identifying the employees in need of the training program.
Identifying the kind of training program that would benefit the
employee and the organization. Cost-benefit analysis. Time and
efforts spent on the training program really worth it? Modeling
approach: Structured data mining approach will be adopted based on
underlying characteristics of data. Logistic regression, neural
networks, support vector machines, decision trees, random forests,
etc would be administered to arrive at predictive models. 46. - 45
- Objective: Profiling of candidates to suit the specific roles and
requirements of the organization with learning's from the past.
Case illustrations: Streamlining the hiring process with learnings
from the past Explanatory variables: Data set covering the below
mentioned datum collected over a period of 3 -5 years. 1.
Socio-demographic Age, sex, marital status, location, education,
place of schooling, previous employment information. 2. Interview
Short listing criterion, position, pay grade, department, region,
etc offered at the time of the joining. 3. Time sheet Department,
pay grade, unit, region, location, shift, DOJ, RO details,
productivity, swipe patterns, etc. 4. Leave Leave patterns,
regularization pattern, outdoor pattern, affinity, ALD, RO details.
5. Performance Appraisal scores, remarks, achievements,
recognition, issues/concerns, negative feedback, pay grade history,
bonus info, project specific info, promotions, etc. Organizational
benefits: Identifying if an employee is most suited to his role in
the organization. Identifying other possible avenues in the
organization where his skills would be most suited. Modeling
approach: Structured data mining approach will be adopted based on
underlying characteristics of data. Logistic regression, neural
networks, support vector machines, decision trees, random forests,
etc would be administered to arrive at predictive models. 47. - 46
- Questions/Feedback? Contact us 01 N, 1st floor IIT Madras
Research Park, Kanagam road, Chennai 600113 Tel :` +91 44 66469877,
Fax:+91 44 66469877 Email: [email protected] Twitter:
AaumAnalytics, Web: www.aaumanalytics.com Facebook:
http://www.facebook.com/AaumAnalytics LinkedIn:
http://www.linkedin.com/company/aaum-research-and-analytics-iit-madras
Aaums office at IIT Madras Research Park About Aaum Aaum Research
and Analytics founded by IIT Madras alumnus brings in extensive
global business experience working with Fortune 100 companies in
North America and Asia Pacific. Incubated at IIT Madras Incubator
ecosystem with a focus on researching and devising the
sophisticated analytical techniques to solve the pressing business
needs of corporations ranging from finance, insurance, HR, Health
Care, Entertainment, FMCGs, retail, Telecom.