How to move hundreds or thousands of entities, each comprised of several shared sub-entities, to another environment

1. A method for prioritization
of subjects with multiple
components
KIRAN ANNAVARAPU
CLEVELAND, OHIO
FEBRUARY 2016
Kiran ANNAVARAPU, Cleveland, OH
© KIRAN ANNAVARAPU 1

2. What are we solving for?
Consider a situation where you need to move hundreds or thousands of entities, each
comprised of several shared sub-entities, to another environment. You want to do this in a way
that (i) is efficient when working with entities, (ii) considers the complexity of sub-entities, (iii)
and considers other factors important to you. Some situations where this could apply:
• A private wealth management firm wants to move its hundreds of customers, each with
several products/funds, to a new set of comparable products with a different financial services
company
• A company wants to migrate its thousands of customers, each with possibly hundreds of
versions of products over the past several years, to a new system
How do you quickly figure out which entities and sub-entities to work with first to get the biggest
bang for the buck, especially when you are constrained by analytics or IT resources? Here’s a
quick method using just our trusty good ol’ Excel.
Kiran ANNAVARAPU, Cleveland, OH
© KIRAN ANNAVARAPU 2

3. Summary of method
1. Lay out all the factors impacting your entities or customers (e.g. value to the business, cost,
public-relations risk, strategic importance). Segment the customers by factor and divide into
buckets (e.g. high, medium, low). You could also create a composite factor by weighting each
factor, and then dividing the composite factor into those buckets. Or you could evaluate each
factor separately
2. Sort the sub-entities or products by complexity (high-to-low, for our illustration) and assign a
number to each sub-entity in order (more on that later)
3. For each customer, compute sums of numbers assigned to products that comprise the entity
4. Sort these sums in descending order and bucket the distribution at natural points of inflexion
(e.g. high, medium, low)
5. Map out the entities on one axis and the distribution of sums on the other. Pick the block of
low/medium/high you want to tackle first
6. Split out the sub-entities that the first block is composed of, and execute on those specific sub-
entities first. Proceed likewise with the rest of the blocks according to your timeline. Sub-entities
already migrated in prior blocks should reduce the time for subsequent blocks
Kiran ANNAVARAPU, Cleveland, OH
© KIRAN ANNAVARAPU 3

4. Hypothetical situation to illustrate the
method
Let’s say our company wants to migrate its thousands of customers, each with several products,
and scores of versions of products over the past several years, to a new system.
We decide to migrate the customers that contribute the most value first, since our efforts have
been well-publicized and customers are looking forward to it. It is important that we work
quickly to retain our high-value customers.
Each customer uses several products, and each product has a new version each year. Some of
the customers have been with us for 20+ years. The products can get incredibly complex. We
decide to migrate the most complex products first because given the budget, current state of
enthusiasm in the organization, vendor relationships and learning curve, that approach makes
most sense (you could just as easily decide to do the opposite in an appropriate situation).
Kiran ANNAVARAPU, Cleveland, OH
© KIRAN ANNAVARAPU 4

5. Segment customers by priority
Step 1 Step 2 Step 3 Step 4 Step 5 Step 6
H
M
L
Lay out all factors impacting your
entities or customers (e.g. value to the
business, cost, public-relations risk,
strategic importance). Segment the
customers by factor into buckets (e.g.
high, medium, low). You could also
create a composite factor by weighting
each factor, and then dividing the
composite factor into those buckets. Or
you could evaluate each factor
separately
Customer
Value to us
($M) Products
C4 $1,500 P4, P5, P7
C1 $1,100 P1, P2, P3
C8 $950 P1, P10
C2 $900 P2, P4
C12 $850 P5, P6, P7
C3 $800 P1, P3, P5
C7 $750 P9
C9 $675 P1, P4, P5
C11 $570 P6, P8
C10 $425 P3, P7
C5 $300 P4
C6 $35 P3, P9
Kiran ANNAVARAPU, Cleveland, OH
© KIRAN ANNAVARAPU 5

6. Sort the sub-entities
Step 1 Step 2 Step 3 Step 4 Step 5 Step 6
Sort the sub-entities or products by
complexity (high-to-low, for our
illustration) and assign a number to
each sub-entity in order
• ‘Computed Label’ is simply a power of
2 of the complexity rank
• We use the power of 2 because:
– These computed labels can be
summed up for analysis, knowing
that each sum is possible only by one
and only combination of computed
labels
– For example, a customer with
products P2 and P8 will have a sum
of 6 (i.e. 2 + 4)
– The sum 6 is possible only from a
combination of 2 and 4
– Here’s why: any power of 2 in binary
form has only one digit as “1” and
the rest are always “0”. In binary
form, 2 is 0010 and 4 is 0100. So the
sum 6 (0110 in binary) can only be
formed by adding 0010 and 0100.
Product
Complexity
Rank
Computed
Label
P10 9 512
P7 8 256
P4 7 128
P6 6 64
P1 5 32
P8 4 16
P9 3 8
P2 2 4
P5 1 2
P3 0 1
Kiran ANNAVARAPU, Cleveland, OH
© KIRAN ANNAVARAPU 6

7. Compute sums of numbers assigned to
products
Step 1 Step 2 Step 3 Step 4 Step 5 Step 6
For each customer, compute sums of
numbers assigned to products that
comprise the entity
Customer
Value to us
($M)
Products
Sum of numbers
assigned to products
C4 $1,500 P4, P5, P7 386 (=128+2+256)
C1 $1,100 P1, P2, P3 37 (=32+4+1)
C8 $950 P1, P10 544 (=32+512)
C2 $900 P2, P4 132 (=4+128)
C12 $850 P5, P6, P7 322 (=2+64+256)
C3 $800 P1, P3, P5 35 (=32+1+2)
C7 $750 P9 8
C9 $675 P1, P4, P5 162 (=32+128+2)
C11 $570 P6, P8 80 (=64+16)
C10 $425 P3, P7 257 (=1+256)
C5 $300 P4 128
C6 $35 P3, P9 9 (=1+8)
Kiran ANNAVARAPU, Cleveland, OH
© KIRAN ANNAVARAPU 7

8. Sort these sums in descending order and
bucket
Step 1 Step 2 Step 3 Step 4 Step 5 Step 6
Sort these sums in descending order
and bucket the distribution at natural
points of inflexion (e.g. high, medium,
low)
Customer
Value to us
($M)
Products
Sum of numbers
assigned to products
C8 $950 P1, P10 544 (=32+512)
C4 $1,500 P4, P5, P7 386 (=128+2+256)
C12 $850 P5, P6, P7 322 (=2+64+256)
C10 $425 P3, P7 257 (=1+256)
C9 $675 P1, P4, P5 162 (=32+128+2)
C2 $900 P2, P4 132 (=4+128)
C5 $300 P4 128
C11 $570 P6, P8 80 (=64+16)
C1 $1,100 P1, P2, P3 37 (=32+4+1)
C3 $800 P1, P3, P5 35 (=32+1+2)
C6 $35 P3, P9 9 (=1+8)
C7 $750 P9 8
H
M
L
Kiran ANNAVARAPU, Cleveland, OH
© KIRAN ANNAVARAPU 8

9. Map customers and select block(s)
Step 1 Step 2 Step 3 Step 4 Step 5 Step 6
Map out the entities by value on one
axis and the distribution of sums on the
other. Pick the block of
low/medium/high you want to tackle
first
C1 C4
C3, C7 C2, C12 C8
C11, C6 C9, C10, C5
H
M
L
HML
Complexity of customer’s products
Customer’sValueProp
Kiran ANNAVARAPU, Cleveland, OH
© KIRAN ANNAVARAPU 9

10. Act on products for selected customer(s)
Step 1 Step 2 Step 3 Step 4 Step 5 Step 6
Split out the sub-entities that the first
block is composed of, and execute on
those specific sub-entities first. Proceed
likewise with the rest of the blocks
according to your timeline. Sub-entities
already migrated in prior blocks should
reduce the time for subsequent blocks
Customer
Value to us
($M)
Products
Sum of numbers
assigned to products
C8 $950 P1, P10 544 (=32+512)
C4 $1,500 P4, P5, P7 386 (=128+2+256)
C12 $850 P5, P6, P7 322 (=2+64+256)
C10 $425 P3, P7 257 (=1+256)
C9 $675 P1, P4, P5 162 (=32+128+2)
C2 $900 P2, P4 132 (=4+128)
C5 $300 P4 128
C11 $570 P6, P8 80 (=64+16)
C1 $1,100 P1, P2, P3 37 (=32+4+1)
C3 $800 P1, P3, P5 35 (=32+1+2)
C6 $35 P3, P9 9 (=1+8)
C7 $750 P9 8
Tip: You can use Excel to systematically figure out products for the selected customer easily by
successively using the ‘BITAND()’ operator on the sum. So in our example, 386 BITAND 8 is 0, so
P9 is not one of C4’s products. But 386 BITAND 128 is 128, so P4 is one of C4’s products.
Kiran ANNAVARAPU, Cleveland, OH
© KIRAN ANNAVARAPU 10

11. Conclusion, Suggestions
• So now, we have decided to migrate (or act in some other fashion) on customer C4 and
products P4, P5 and P7. You could have chosen C1 if that customer was somehow a higher
priority for you, or C8 if those products were more important to migrate from an execution
standpoint
• When presenting to an audience like the Board of Directors, you’d of course want to avoid
going through the ‘math-iness’ of this method (you don’t want to be thaaat guy, right?). Focus
on the “so what?” and a high-level “what (you did)”. Show the “how” to those interested in
knowing
• If you are aware of methods to solve other business problems, I’d love it if you would share it
with me
• Contact me if you have any questions, feedback or suggestions
© KIRAN ANNAVARAPU 11