c



c
 

 
A Markov Analysis Attempt for Attrition -Rate Prediction and
Stabilization

WHITE PAPER




In a competitive arena, the advantage is taken by the first-mover ± and for an

environment where The Rule-Of-Three predominates, it is not just the first
mover, but the fast -mover who has it all.
Every organisation, no matter how stable its quality and people processes, is
bound to fall prey to the silent warfare of the fast-movers ± which I would
£

prefer to call a

 

  and Hertzberg¶s ³Satisfiers´ are today¶s
HR nightmare ± because nothing seems to work!

Thus today, HR as a strategic partner in any organisation has lots to do in

terms of metrics, HR analytics, prediction of trends and quantifying Human
Capital measures.
Since attrition is one of the main problems for any organisation struggling to
retain its expertise and knowledge base, an analytical approach to the same
would also help in prediction and necessary remedies.

This paper aims to draw on the recent HR trend of referring to the employee
as an ³internal customer´ and therefore assumes that manpower attrition is
similar to customer switching problems in case of products, thus has used
Markov Analysis as an Operations Research technique to predict attrition,
and therefore form a basis for manpower planning.

This white paper is aimed at a greater scope of having more thought

provoking ideas in the HR Analytics arena and within its limited scope here,
suggests an OR model as part of manpower inventory planning in general.






 



º ?
 
 ?

? !"

#! $  %?

 ? !    &?

% ? !  ! 

 '?

& ? !$     
  (?

) ? !c "$
  ºº?

' ? 
 
 º%?

*   !  ! º&?

º ?
 º)?
º 
 

?
The Attrition Warfare

One of the greatest strategies of War had been the strategy of attrition warfare, defined in
military terms as ³a strategy of warfare that pursues victory through the cumulative
destruction of the enemy¶s material assets by superior firepower.´
Metrics like body counts and terrain captured measure the progress of battle. On the
opposite end of the spectrum is maneuver warfare. All warfare involves both maneuver
and attrition in some mix. The predominant style depends on a variety of factors such as
the overall situation, the nature of the enemy and most importantly, on attackers¶
capabilities.

Though this paper deals with attrition with respect to the War for Talent in Corporate
arena, the strategy involved is the same ± and even the terminologies quite similar ± if
³body count´ can be a parameter to measure effectiveness of attrition warfare, then in
corporate recrutiment strategies the similar parameter would perhaps be ³acceptance to
offer ratio´ (from the attacker¶s perspective).

Human Resource professionals are under increased pressure from a different kind of a
a


 ± the silent firepower of attrition which causes no less harm to Human
capital assets, as compared to ³the enemy¶s material assets´ as in the definition above.

The concept of what has been stated above can be put into a simple model as shown
below. (fig1.)

++,-+

-.

.+1+

-+/. -0/- -

fig1
/-

The pressure of competition from the environment and the evolution of strategy are self-
explanatory in the above figure. The point to note here is the extent of the impact, which
involves hitherto soft issues like culture and people, and this is the origin of strategic
human resource focus, the war for talent and the need to garrison the human resource
capital as one of the strategic parameters.
  !"

#! $ 
?
APQC (American Productivity and Quality Centre) has made several recommendations to
raise awareness of the problem of knowledge attrition, which include

1.???????Identifying a burning platform or issue related to knowledge loss

2.??????? Looking for windows of opportunity through champions who are willing to try out
knowledge retention approaches.

AQPC has categorized --2+3/-.-0-4-+-2.+ .

This includes

1. // 2+3/-.- ± This includes management practices, values, respect for

hierarchy, and decision flows.?
2. !4/2+3/-.- ± this includes the organization¶s journey from the day it was
founded till the present?
3. ++/ 2+3/-.- ± this includes technical, operational, process and client
information?

A more careful look at figure 1 indicates that there seems to be some good amount of
convergence with respect to AQPC¶s definition of the three types of knowledge and the
model given in figure 1 ± particularly the fact that corporate attrition warfare is all about
gaining (through head-hunting, strategic recruiting, internal job offers, etc) human assets,
who bring along with them the three kinds of knowledge, and thereby attack the very
strategic base of the organization.

Thus from the attacker¶s point of view, depending on which type of knowledge it needs
from the competitor, the recruitment strategies are also sorted out accordingly. It is
evident therefore, that attrition rate among junior employees (2-4 yrs) would be higher for
the functional knowledge part ± associated with technical and operational processes.
At higher levels, the attrition warfare would be more for gaining historical knowledge
(business portfolio changes down the years, etc) and cultural knowledge from the
competitors.

From the organization¶s point of view, the counter strategy is to predict attrition ³zones´
which depend on the criticality or type of knowledge that is at important to the
organization, and thereby evolve plans to counter loss of human assets from those
positions.

Once we realize this, the next step is to come out with concrete plans to prevent attrition,
which can only be forecast using data and trends available. Some of the world¶s best
practice organizations have tried capturing data to predict attrition on the long run, and
done that in different ways.

? ?
? ?


  !   
?
 º ++2+3/-.-,+.-,-+ 5446!4/+//2+3/-.- 

From the attackers¶ perspective, one of the parameters to measure effectiveness of

corporate attrition warfare might be ³acceptance to offer´ ratios. But from the perspective
of the organization that has to cope up with this ever-growing problem, the problems
associated are larger.
Attrition is a pain area in any organization that intends to have a knowledge management
ii[2]
system in place. In a famous article , attrition (through normal retirement or through
resignations) has been discussed as one of the pain areas in the field of KM, because
vacancy of a position might be easier to fill in through the proper people-sourcing
approaches, but filling in the knowledge gap is not. This is particularly in context of a
tough economy where the concept of all-size-fits-all is no longer working, and vacancy of
a position by attrition is basically vacancy of a knowledge-base, and this vacancy in
knowledge base cannot be filled in by  person.

This is precisely what is referred to as 2+3/-.-, which most organisations today

are grappling to capture and retain. This closely pertains to what AQPC referred to as the
// + !4/ 2+3/-.- , in addition to the Individual or Proprietary
knowledge that goes off without being codified and migratory, and therefore is never
assimilated in the organisation as invisible knowledge. This can be exemplified better
through the typical knowledge-cycle of an organisation as shown below, originally by
Takeuchi and Nonaka:

6- c.
7   2+3/-.- 2+3/-.-

‘
  

  4-- +48/-
2+3/-.- 2+3/-.-

   ! 

The problem can be aptly stated through examples from the corporate world itself ±

Corning, which had been experiencing knowledge loss through the large scale
retirements through 1990¶s estimated that it lost around 2000 years of cumulative years of
experience as a result of a retirement package offered in 1998 ± and this exemplifies loss
of knowledge due to planned retirements alone ± here we are talking of r




, where an employee may walk into the office any morning to place his
resignation letter and walk off with the competitor ± not just creating a vacancy, but taking
some of the most vital knowledge quantum from the company to it¶s competitor.
However, organisations even with established knowledge management practices have
not been able to come up with any substantial measure to check this knowledge loss, and
therefore an indicator of failure in capturing tacit knowledge bases.
 


  ++// #-+-4#446++/2+3/-.- 

The problem is more acute depending on the industry and the demographics of the
employees too, as in call centres. Here the knowledge drain is at a different level, and it
corresponds more to AQPC¶s definition of ++/2+3/-.- .

Though it is a known fact that high turnover rates drain the cost effectiveness of call
centres, unfortunately little is being done about it.
iii[3]
In the article ³Reducing Call Centre Turnover´ , managers in call-centres normally tend
to look only at advertising costs, interviewing and training costs etc, but overlook the vital
costs associated with attrition.
Merrill Lynch attempted to find out costs associated with call-centre attrition ± which came
out to be around $9m per annum for a company with 1000 employees, and annual
revenue of $100m.

This shows that retention alone can significantly bring up the bottom-line for a call-centre.

Organizations tend to spend huge sums of money on recruitment, for web-postings, job
fairs, ads, employee referral bonuses, etc, and end up with 50% employees leaving
before reaching any level of proficiency.

Proper testing and screening, training, introduction of the apprenticeship scheme,

aptitude testing (10%), realistic job previews (8%), structured behavioral interviews (3%)
can help prevent attrition by percentages shown in parenthesis.

According to the Forum Group, 65% of the external customers leave due to internal
reasons alone (45% for poor service quality, 20% due to lack of attention) ± thus internal
attrition can devastate call-centre effectiveness if not tackled properly.
iv[4]
Shown in the table below are the typical turnover rates of call centres.

MEDIAN (%) AVERAGE (%) HIGHEST

(%)
Part time inbound 20 33.6 300
Full time inbound 19 26 252
Part time outbound 15 35.5 480
Full time outbound 10 21.3 210
÷



%  !  ! 



Organisations across the world and operating in different industry segments have tried to
find out means to measure business loss through attrition.

Schlumberger, for example, understands how important it is to link its knowledge sharing
techniques with its HR processes: the oil industry faces an attrition rate of 44% by
v[5]
2010.

Pfizer also takes preventive measures to combat knowledge-drain and promote better
knowledge transfer through its six-step knowledge retention process.

Best practice companies, according to AQPC, should conduct a thorough audit to

determine what knowledge is worth capturing. Stated in another way, this would also
indicate the ³critical positions´ in the organization, which can create a substantial problem
to the company incase it is vacated under competitor attack.
The table below shows the practices that are followed by these organizations to collect
vi[6]
data related to attrition:

Siemens Corning World Northrop Xerox Best

Bank Grumman Connect Buy
Internal networks      
Interviews      
Videotaping      
SME directory      
Repositories      
After action project      
milestone reviews
Mentoring programme      
Knowledge maps      
Recruiting strategy      
Retention strategy      
÷


The importance for including the various ways companies worldwide are collecting data
on attrition would be clearer in the subsequent sections.
vii[7]
A Hay Group survey reveals that what people want most is to feel that their careers
are moving forward.
In their survey, ³The retention dilemma: Why productive workers leave and seven
suggestions for keeping them´, reveals that employees leave because of disillusionment
of the company management¶s direction, and because of under-utilization.
Two of the seven things Hay Group identified as ³attrition-preventing´ are clearly related
to training ±

1.???????Measurement of soft skills ± because gaps exist when the companies say they value
their people, and do something else
2.??????? Fight attrition with smart training ± taking a longer term perspective in training and
development as a retention tool.
The relationship between job satisfaction and attrition as surveyed by Hay Group is
shown as follows:

46+3 /0--+446-
0
     
 

  

*se of my skills and abilities 83 49 34
Ability of top management 74 41 33
Company has clear sense of 57 27 30
direction
Advancement opportunities 50 22 28
Opportunity to learn new skills 66 38 28
Coaching and counseling from 54 26 28
one¶s own supervisor
Training 54 36 18
Pay 51 25 26
÷


However, few organisations have been able to tackle attrition in spite of using various
types of data-gathering instruments as shown in table 2.

Thus the problem is perhaps somewhere else.


&  !$     
 ?
In order to understand this, it is important to question the very validity of the data that is
given by the employees ± it is only common sense that an employee would not reveal the
correct reason for leaving the company at some point of time ± thus any action taken by
the organization to prevent attrition by altering the factors as mentioned above does not
have any effect, r
 .

& º -8/-,

- 08/-, 6 - / 6 -  6, + + 4- 5 - /
:*;
9+.- - 

We have seen above, that inspite of a great number of efforts, and the availability of a
number of instruments for collecting reasons as to why people are leaving, an
organization is really not being able to do much about attrition ± the primary reason of this
could be the validity of the data.
As to why employees would not/might not give the correct response to an attrition survey
stems from the social exchange theory (Dillman, 1978). According to this, there is a social
exchange between the survey interviewer, who desires information possessed by the
respondent, and the respondent, who decides how much information to convey. Dillman
posits that the respondent participates because the act of participation is expected to
bring rewards that exceed the cost of participation. These rewards might include
monetary payment, but more importantly would include intangible rewards that, to some
extent, can be influenced by the design and implementation of the survey.

Dillman argues that the willingness of an individual to participate in a survey depends

critically on the degree of trust that the expected terms of the social exchange described
above will be fulfilled.

The social exchange model described above can be translated into an economic model
and, in its translated form, can be used to help generate some empirically testable
hypotheses about the determinants of survey participation, validity of the response and
the data.

This paper only outlines the theory, leaving it for future scope of research on the subject.

According to social exchange theory, the individual's willingness to participate in a survey

depends on a comparison of the benefits and costs of participation to him. Let the
individual's utility function be given by
R R
* it = * (Lit, Yit) + Eit ««««««««««««««««(1)

where

*R (Lit, Yit) is the utility the individual receives from leisure, L it, and income, Yit.
Eit is the psychic value the respondent expects to experience by participating in the
interview,
Eit = 0 if the individual does not participate.
The individual's money budget is

Yit =Vit + wit Hit + pit «««««««««««««««««(2)

where
Vit is nonlabor income,
w it is the market wage rate,
Hit are hours of work
th
Pit is a respondent payment for participation in the t wave of the survey.

The individual's time budget,

T = Hit + Lit + lit, «««««««««««««««««««««..(3)

is the sum of hours of work, hours of leisure, and time spent on the interview.

The individual obviously chooses his labor supply independently of the survey interview
2
by maximizing Equation (1) subject to Equations (2), (3) and Eit = lit = Pit = 0 . This choice
is described by the labor supply function Hit = H(wit, Vit). Substituting the labor supply
function and the time and money budgets into Equation (1), the individual's utility function
is given by

*itR = *R [T - Hit(wit,Vit) - l it, Vit + wit Hit(w it, Vit) + pit] + Eit................(4)

Treating lit as a marginal loss of leisure and pit as a marginal gain in income, the net utility
gain, or loss from participation in the survey is given by

¨* = -*Llit + *Ypit + Eit

= (-witl it + pit)*Y + Eit. «««««««««««««««««..(5)

where

*L is the marginal utility of leisure,

*Y is the marginal utility of income
w it = *L/*Y is the shadow price of time in nonmarket uses which is equal to the market
wage rate if the individual is working in the labor force.

The individual will participate in wave t of the survey if the rewards from doing so
outweigh the costs according to the decision rule

Participate if E it/*Y > witlit - pit; otherwise, refuse ««««««.(6)

Where Eit/*Y is the monetary value of the psychic costs and rewards of the survey
experience ± the problem here being, that a person who is leaving an organization wants
neither psychic utility nor rewards, and thereby his perceived-utility is low, therefore he is
under no obligation to respond correctly/accurately to attrition surveys.
)  !c "$
  

One of the most recent trends in HR is treating the employees as internal customers, and
the next step is to consider attrition as a customer-switching problem ± and once we can
do that, attrition rate prediction may be dealt with similarly as in customer switching
problem in case of marketing.

The solution proposed here is the application of Markov analysis to customer switching
problems ± clearly stated, a Markov analysis to find out the attrition rate and prediction of
its stability within time period t, which would give HR people a relevant input in terms of
their manpower planning and recruitments.

A Markov chain is a random process for which the future depends only on the present
state; it has no memory of how the present state was reached. This simplifying
assumption leads to a family of systems having a mathematical theory, as well as many
applications to modeling in more applied science. A central property of `nice' Markov
chains is that they settle down into a (stochastic) equilibrium.

The basic method for solving this is to construct the transition probability matrix, which
takes in attrition probability data by using instruments as mentioned in the TABLE2. The
validity of the output would depend on the validity of this probability, which is a problem
area, because of the inaccuracy of responses according to the Social Exchange Theory
as mentioned above.

Here I propose to exemplify the construction of the transition-probability matrix as under:

In analyzing switching between companies, the reason for attrition, the organization
needs to have data that is needed to form the transition probability matrix.

As an example laid down below, say the probability that the employee stays in the
organization is 0.95. The corresponding probabilities of his/her switching to competitor
companies 2, 3, and 4 are say 0.02, 0.02 and 0.01 respectively. The other figures put in
the example are self-explanatory.

Thus we construct our probability matrix as follows:

To company
1 2 3 4
From comp 1 | 0.95 0.02 0.02 0.01
2 | 0.05 0.90 0.02 0.03
3 | 0.10 0.05 0.83 0.02
4 | 0.13 0.13 0.02 0.72

Say for the present time, say this month, the probability of switching to companies 2, 3, 4
are 23%, 20% and 12%, and for staying in the company itself is 45%.

[The probability is calculated on various parameters that evoke switching, for example,
competitors¶ pay, work environment, perks, etc]

 
 

) º -/+

÷
 

1.??????? While exemplifying through the matrix, it has been assumed that the strategic
sourcing group of the organization aims to have a 75% target of the probability of
employees wanting to remain, that is, around 25% attrition rate.
2.??????? The basic assumption of Markov analysis is also applied here, that the process is a
stochastic one, whereby any event would only depend on the preceding event, and
nothing else.

We have the initial system state s1 given by s1 = [0.45, 0.23, 0.20, 0.12] and the transition
matrix P given by

P = | 0.95 0.02 0.02 0.01 |

| 0.05 0.90 0.02 0.03 |
| 0.10 0.05 0.83 0.02 |
| 0.13 0.13 0.02 0.72 |


Hence after one month has elapsed the state of the system s2 = s1P = [0.4746, 0.2416,
0.1820, 0.1018] and so after two months have elapsed the state of the system = s3 = s2P
= [0.494384, 0.249266, 0.16742, 0.08893] and ofcourse the elements of s2 and s3 add to
one (as required).

[     r  
    
  
r  r  
 rrr 
    

r! 
"#   .]

Hence the employee demand after two months have elapsed are 49.44%, 24.93%,
16.74% and 8.89% for companies 1, 2, 3 and 4 respectively.

Assuming that in the long-run the system reaches an equilibrium [x1, x2, x3, x4] where

[x1, x2, x3 , x4] = [x1, x2, x3, x4]P and x1 + x2 + x3 + x4 = 1

we have that

x1 = 0.95x1 + 0.05x2 + 0.10x3 + 0.13x4

x2 = 0.02x1 + 0.90x2 + 0.05x3 + 0.13x4
x3 = 0.02x1 + 0.02x2 + 0.83x3 + 0.02x4
x4 = 0.01x1 + 0.03x2 + 0.02x3 + 0.72x4
x1 + x2 + x3 + x4 = 1
Rearranging we get
0.05x1 = 0.05x2 + 0.10x3 + 0.13x4 (1)
0.10x2 = 0.02x1 + 0.05x3 + 0.13x4 (2)
0.17x3 = 0.02x1 + 0.02x2 + 0.02x4 (3)
0.28x4 = 0.01x1 + 0.03x2 + 0.02x3 (4)
x1 + x2 + x3 + x4 = 1 (5)

Now from equation (3) we have

0.17x3 = 0.02(x1 + x2 + x4)

and from equation (5) we have

x1 + x2 + x4 = 1 - x3

Hence

0.17x3 = 0.02(1-x3)
i.e. 0.19x3 = 0.02
i.e. x3 = (0.2/0.19) = 0.10526

Now subtracting equation (2) from equation (1) we get

0.05x1 - 0.10x2 = 0.05x2 + 0.10x3 - 0.02x1 - 0.05x3
i.e. 0.07x1 - 0.15x2 = 0.05x3 (6)

Also substituting for x4 from equation (5) in equation (4) we have

0.28(1 - x1 - x2 - x3) = 0.01x1 + 0.03x2 + 0.02x3
i.e. 0.28 = 0.29x1 + 0.31x2 + 0.30x3
i.e. 0.29x1 + 0.31x2 = 0.28 - 0.30x3 (7)

Multiplying equation (6) by 0.31 and equation (7) by 0.15 and adding we get

(0.31)(0.07)x1 + (0.15)(0.29)x1 = (0.31)(0.05)x3 + (0.15)(0.28) - (0.15)(0.30)x3

and since we know x3 = 0.10526 we have x1 = 0.59655

Hence from equation (6) we find that x2 = 0.24330 and from equation (5) that x4 =
0.05489

As a check we have that these values for x1, x2, x3 and x4 satisfy equations (1) - (5) (to
within rounding errors). Hence the long-run employee demands for the companies are
59.66%, 24.33%, 10.53% and 5.49% for companies 1, 2, 3 and 4 respectively.

We need a long-run system state of [0.75, x2, x3, x4] where x2, x3 and x4 are unknown (but
sum to 0.25) and we have a transition matrix given by

P=
| p 1 p2 p3 p4 |
| 0.05 0.90 0.02 0.03 |
| 0.10 0.05 0.83 0.02 |
| 0.13 0.13 0.02 0.72 |

where p1, p2, p3 and p4 are unknown (but sum to one).

 Hence using the equation
[0.75, x2, x3, x4] = [0.75, x2, x3, x4]P

we have the equations

0.75 = 0.75p1 + 0.05x2 + 0.10x3 + 0.13x4
x2 = 0.75p2 + 0.90x2 + 0.05x3 + 0.13x4
x3 = 0.75p3 + 0.02x2 + 0.83x3 + 0.02x4
x4 = 0.75p4 + 0.03x2 + 0.02x3 + 0.72x4

Together with x2 + x3 + x4 = 0.25 ; p1 + p2 + p3 + p4 = 1

Here we have six equations in seven unknowns and so to solve we need an appropriate
objective. In order to avoid having to change the transition probabilities too much a
suitable objective would be

c9,4-0º

I.e. find the largest value for the transition probability from company 1 to itself such that
the recruiter achieves the /+.#+-,0/---,+6'&<

'  
 


The above approach through a Markov analysis is a proposed model. This model may be
followed and can be mapped to a much more complex data through the construction and
the solving of the probability matrix through a mathematical tool. The objective of the
paper was to propose a quantitative way to predict attrition rate in any industry and
therefore take the necessary steps to prevent it, or plan the manpower inventory
accordingly.

Companies should project retirements and attrition over the next five years. List the
internal and external forces that can contribute to the problem. Then take the worst-case
scenario. The main approach to preventing attrition should be grooming leaders, rather
than just treating employees the way it is normally done.

In fact, the companies with leading-edge retention programs address all the areas
ix[9]
mentioned below. According to International Data Corp.'s guru on resourcing
strategies, Michael Boyd, program elements can include the following:

R Ongoing education and training.

R A mix of job assignments.
R The organization of small groups and teams.
R Peer group and mentoring programs.
R Organized career counseling.
R Flextime and other lifestyle benefits, including on-site day care, fitness clubs and
sponsored charity work.
R Internal marketing and communication with employees.

But in case nothing works, the best way is to predict it and act accordingly. Thus
prediction becomes vital.

******************end******************
*   !  ! 

Suvro Raychaudhuri is working as an HR Process Consultant in one of the leading IT
Solutions Firm, in the e-HR practice area in the capacity of a domain consultant. He holds
a Degree in Mechanical Engineering and is a Post-Graduate in Personnel Management
and Industrial Relations from one of the premier Business Schools in India.
 º  

?
?
?
?
?

? ???????????????????????????????????????? ???????????????
?

?The German Word for ³Propaganda´ or ³Silent Warfare´.
?

2
³Why attrition is a chance to prove the value of KM´, KM Review Briefings, Vol6, Issue1,
March/April 2003?

3
Drew Robb, CustomerInterface March 2002 Issue, P-34,35

4
Purdue *niversity Centre for customer driven quality

5
³Proactive strategies to combat attrition´, Rowan Wilson and Jennifer Wilson, KM Review,
Vol 4, Issue 6, Jan/Feb2002

6
³Why attrition is a chance to prove the value of KM´ KM Review Briefings, Volume 6 Issue 1
March/April 2003, P-10.

7
³Hay Group Study Identifies Training as One of Top 7 Employee Attrition Fighters´ IOMA¶s
report on managing training & development, April 2002 issue, P-13

8
pa    ÷÷p ÷
 , By: Hill, Daniel H., Willis, Robert J., Journal of Human
Resources, 0022166X, Summer2001, Vol. 36, Issue 3.

9
÷


p
÷ ÷
 
÷ p   , By: Gantz, John,
Computerworld, 00104841, 7/3/2000, Vol. 34, Issue 27