Problem:

Manzana Insurance
z Loss of business due to late renewals, long
lead times on new policies and quotes
z Competitive pressure (long lead times have
created opportunity for competitor to take
market share)
z Overemphasis on new policies causing loss of
profitability (insurance losses)

Using Little’s Law: Turnaround Time

z Lead time = WIP/TP z Worst case scenario quoted for each operation
z WIP = 82 policies (95% SCT)
z E.g. distribution of RUNs avg 41, 95% SCT
z Throughput = 39 policies/day 128.1 minutes
z => Leadtime approximately 2 days z Final “turnaround time” 2-3 times actual lead
time
z => loss of business

Priorities Queuing Inputs

z Should be first come, first served z Processing time to a station: p

z Changed by underwriters to RUN, RAP, RAIN, z Time between arrivals to a station: a
RERUN z Utilization: p/a
z => RERUNs done at end of day (rather like z Squared coefficient of variation: CVp, CVa
producing a batch size = 1 day’s demand) z Process flow diagram
z RERUNs released one day before due – why? z Number of workers at each station

1
Note that there are actually 5
products: Processing Rates (Exhibit 4)

z RUNs RUN RAP/RUN RAP RAIN RERUN Avg.

z RAPs that are accepted (15%) – RAP/RUN DC 68.5 50 50 43.5 28 41
z RAPs not accepted UW 43.6 38 38 22.6 18.7 28.4
z RAINs R 75.5 64.7 64.7 65.5 75.5 70.4
z RERUNs PW 71 71 0 54 50.1 54.8
Labor input per policy 258.6 223.7 152.7 185.6 172.3 194.6
Hours 4.3 3.7 2.5 3.1 2.9 3.2

Calculating Utilization for

Distribution Utilization for all operations
DC 0.89
z Arrival rate = 39 policies per day UW 0.82
z Distribution service rate per day:
R 0.76
– 41 min/policy on average
– 7.5 hours * 60 minutes * 4 DC clerks = 1800 PW 0.64
min/day available
– => Distribution capacity 1800 min/day / 41
min/policy = 44 policies/day
z Utilization = arrival rate/service rate = 89%

Adding in extra DC work (e.g. 30

But policy creation is not all that minutes/clerk/day)
the distribution clerks do!!

“Distribution was also responsible for

analyzing and disseminating data DC 0.96
published each month by the state
insurance commissioner, researching
UW 0.82
and verifying insurance rates quoted by R 0.76
local competitors, and overseeing the PW 0.64
rating operations…”

2
Standard Deviation of Processing
Times (from Exhibit 4)
Distribution of Demand

%
DC UW R PW RUNs 0.07
RUNs 30.7 32 20.5 10.3
RAP/RUN 24.9 24.5 13.6 10.3 RAP/RUN 0.06
RAPs 24.9 24.5 13.6 0
RAINs 9.2 11.7 15.9 8.6
RAPs 0.33
RERUNs 6.2 19.8 9.7 9.5 RAINs 0.10
RERUNs 0.44

Estimating the CVa Queuing Inputs

z Utilizations fairly high => CVp at previous

station decent predictor for CVa at current
DC UW Rate PW
station SCVa 1.00 0.05 0.12 0.01
z Assumption: CVa to DC = 1 SCVp 0.05 0.12 0.01 0.01
Utilization 0.96 0.82 0.76 0.64
serv rate 41 28.4 70.4 54.8
m 4.00 3 8 5

Wq for the M/M/m queue For the DCs:

z m=4
z 1/(m µ) = 1/44
z ρ = 0.96
1 ⎛ ρ 2 ( m +1) −1 ⎞ Hence, expected waiting time estimated at 0.52 days,
⋅⎜ ⎟ z
Wq = or 235 minutes at the first operation
mµ ⎜ 1− ρ ⎟
⎝ ⎠ z Adjusting for the low CV of processing times (SCVa =
1, SCVp = .05) => 123 minutes
z Lq = 123 min/pol * 39 pol/day / (7.5*60 min/day)
z ~ 11 policies
z => Ls ~ about 12 policies (case: 16 policies at DC)

3
Queueing Results

z Only DC results are consistent with WIP

numbers in Exhibit 3
DC UW Rate UW z Note that UW teams are dedicated (single line,
W q - min 123.15 2.90 1.30 0.11 single server queues)
W s - min 164.15 31.30 71.70 54.91
z => redo calculations for individual territories
Lq 10.67 0.25 0.11 0.01
Ls 11.63 1.07 0.87 0.65

Policies/Day/Territory (Ex 7) Queueing Results, UW

Terr 1 Terr 2 Terr 3

Terr 1 Terr 2 Terr 3 Total
arr/day 14.6 13.2 11.2
1.4 0.8 0.7 2.9
util, UW 0.92 0.83 0.71
0.9 0.7 0.7 2.3
5.4 3.6 3.7 12.7 Wq,UW 28.17 11.99 5.79
1.6 1.0 1.1 3.8 Ws, UW 56.57 40.39 34.19
5.3 7.0 5.0 17.3 Lq, UW 0.91 0.35 0.14
14.6 13.2 11.2 39.0 Ls, UW 1.84 1.18 0.85

Revised Results: What is going on?

z Total Ws, Territory 1: 5.79 hours z Only RERUNs are late

z Total WIP: z Time given to process: 1 day
– DC: 12 z Put to back of queue: given a 6 hour
– UW: 4 (across all territories) processing time for average policies amazing
– R: 1 that any RERUNs are on time!
– PW: 1 z Note that Exhibit 3 from Monday morning:
z Still less than observed by casewriter weekend arrivals?

4
Take-aways Next step:

z Competitive role of lead time reduction Construction of an MPX model of Manzana’s

z “Mental models” operations to provide a higher level of decision
z Role of “invisible work” support.
z Dedication of teams => excessive utilization for
1 team => higher losses
z Priority rules, cost of expediting (“lot sizing”) -
who makes the decision?