granmetrics exp
TRANSCRIPT
-
8/9/2019 Granmetrics Exp
1/24
Granular Metrics,Granular Metrics,
Feedback and ExperimentationFeedback and Experimentation
Arnoldo HaxAlfred P. Sloan Professor of Management
-
8/9/2019 Granmetrics Exp
2/24
Contributions of the Delta ModelContributions of the Delta Model
Opening theOpening the The BestThe Best Three Strategic Options:Three Strategic Options:Product doesProduct doesmindset to newmindset to new
zz Best ProductBest ProductThe TriangleThe Triangle not always winnot always winstrategicstrategiczz Total Customer SolutionsTotal Customer Solutionspositionpositionzz SySystem Lockstem Lock--inin
LinkingLinkingStrategy withStrategy withAdaptiveAdaptive ExecutionExecution
ProcessesProcesses
Execution isExecution isnot thenot theproblem,problem,linking tolinking tostrategy isstrategy is
Execution is captured through ThreeExecution is captured through ThreeAdaptive Processes:Adaptive Processes:
zz Operational EffectivenessOperational Effectiveness
zz Customer TargetingCustomer Targeting
zz IInnnovnovationation
whose roles need to change to achievewhose roles need to change to achieve
different Strategic Positionsdifferent Strategic Positions
MeasuringMeasuring GoodGoodsuccesssuccess financials dofinancials do
AggregateAggregate not alwaysnot alwaysMetricsMetrics lead to goodlead to good
resultsresults
Aggregate performance metrics need toAggregate performance metrics need toreflect each of the Adaptive Processesreflect each of the Adaptive Processesand their role based upon the strategicand their role based upon the strategicpositionposition
zz Product performanceProduct performance
zz Customer performanceCustomer performance
DiscoveringDiscoveringGranularGranularperformanceperformanceMetrics andMetrics anddriversdrivers
FeedbackFeedback
Managing byManaging byaveragesaveragesleads to belowleads to belowaverageaverage
performanceperformance
zz Complementor performanceComplementor performanceBusiness is nonBusiness is non--linear. Performance,linear. Performance,particularly bonding, is concentrated.particularly bonding, is concentrated.Granular metrics allow us to focus onGranular metrics allow us to focus onunderlying performance drivers, tounderlying performance drivers, to
detect variability, to explain, to learn, anddetect variability, to explain, to learn, andto act.to act.
-
8/9/2019 Granmetrics Exp
3/24
The Delta Model and Granular MetricsThe Delta Model and Granular MetricsSelect Performance Indicators
Explain Variability
Identify performancedrivers
Correlate drivers withvariability
Detect Variability
Drill down to detailedsegmentation Isolate variability
Define program Define structured tests Rollout
Hypothesize cause andeffect
Evaluation
Act from Variability
Learn from Variability
-
8/9/2019 Granmetrics Exp
4/24
Drivers of variability for selected performance indicatorsDrivers of variability for selected performance indicators
Drivers of variabilityPerformance indicatorsPerformance indicators Drivers of variability
Product cost and qualityProduct cost and quality zz ScaleScalezz DensityDensity--e.g. concentrae.g. concentrattion of serion of servvicicee
zz LLoocatiocationn
zz Labor productivityLabor productivity-- practices andpractices and trainintrainingg
zz EquipmeEquipmennt product producttivivityity-- desigdesignn
zz Process designProcess design
zz
AnAnd sod so oonnCustomer profitCustomer profit zz Customer sizeCustomer size
zz Customer revenueCustomer revenue
zz TenureTenure
zz Acquisition costAcquisition cost
zz ChanChannel mixnel mix
zz
Customer care supCustomer care supppoortrtzz Customer investments inCustomer investments in relatiorelationnshshiippss
zz AnAnd sod so oonn
zz Size of relevSize of relevaant complementor productsnt complementor productsComplementor contributionComplementor contribution
zz Complementor investment in bComplementor investment in buusinesssiness
zz Relative size in customer value chainRelative size in customer value chain
n t customer economicszz
Complementor contributiComplementor contributioon too customer economicszz ExclusivityExclusivity of relaof relationshiptionship
zz AnAnd sod so oonn
Business segment economicBusiness segment economic zz ROI (return oROI (return onn investment)investment) profitabilityprofitabilityvaluevalue zz RiskRisk-- volatilityvolatility and covariancand covariancee
zz Option valueOption value
zz
Investment baseInvestment basezz CashflowCashflowss
-
8/9/2019 Granmetrics Exp
5/24
Experimentation as the basis for effective changeExperimentation as the basis for effective change
LargeLarge
Size ofSize of
ChangeChange
SmallSmall
The middle road:The middle road:lower returns, orlower returns, or
unacceptable whenunacceptable when
first moverfirst moveradvantage is highadvantage is high
Unacceptable risk,Unacceptable risk,as a starting point.as a starting point.Highly desirable asHighly desirable as
endpointendpointIneffective: lowerIneffective: lower
returns, orreturns, orunacceptable whenunacceptable when
first moverfirst moveradvantage is highadvantage is high
Testing: theTesting: the
relevant area forrelevant area forexperimentsexperiments
leading to largeleading to largechangechange
SlowSlow FastFastSpeed ofSpeed of
ChangeChange
-
8/9/2019 Granmetrics Exp
6/24
Cost DeCost De--Averaging:Averaging:
Using Granular Metrics ToUsing Granular Metrics ToDrive PerformanceDrive Performance
-
8/9/2019 Granmetrics Exp
7/24
Most
Expensive
$100
$1,000
Least
Expensive 0
20
%o
fcosts
% of orders
40
60
80
100
20 40 60 80 100Orders
2. Individual Order Cost
Cost/order
Switching Billing Marketing Etc.Order
fulfillment
1. The Value Chain of the Local Business Data Circuit
Average Cost = $395/order
3. The 80-20 Cost Effect
Transmission
THE BEHAVIOR OF COST - THE CASE OF BUSINESS DATA SERVICES
Figure by MIT OCW.
-
8/9/2019 Granmetrics Exp
8/24
The drivers of the cost of orderThe drivers of the cost of order--fulfillmentfulfillment
Cost of orderfulfillment
EquipmentLocation ofwork centers
Activities oforder fulfillment
Manpowerproductivity
-
8/9/2019 Granmetrics Exp
9/24
1
-
}
/
/
THE BEHAVIOR OF COST BY LOCATION
1,000
70
Headcount
requiredtoproce
ss
1,0
00o
rderspermonth
2,000 5,000
Chicago
Competitor 1
San Francisco
Kansas City
Competitor 2
San JoseCompetitor 2
Indianapolis
Competitor 3
Jacksonville
Competitor 1
Denver
New York
Tulsa Charlotte
Competitor 1 - Chicago
Competitor 1 - Houston
39%
gap
Orders Month
Or der Fulfillment
12,000 27,000 60,000
Client performance
curve
Company
Competitors
Scale = 72%
Log log graph
Best-in-class
performance
Figure by MIT OCW.
-
8/9/2019 Granmetrics Exp
10/24
(
)
(
)
THE BEHAVIOR OF COST BY FIVE ACTIVITIES IN THE VALUE CHAIN
Order
Facilitiesavailable
Access
available Pass test
Valid design
Accessavailable
Accessnotavailable
Accessnotavailable
Facilitiesnotavailable
Automaticswitch map
Fastfacilities test
Pass accesstest
Customerready
$100 per
access leg
10x cost
difference
1MILLIONPOSSIBLEORDERPA
THS
$2,000 per
access leg
Skin test
Invalid design
Manualswitch map
Failfacilitiestest
Fail accesstest
Customernot ready
'Defect-free Path'
'Defect-pr one Path '
Figure by MIT OCW.
-
8/9/2019 Granmetrics Exp
11/24
The behavior of cost by type of defectsThe behavior of cost by type of defectsClientClient
Business DataService Orders
Business Data
Service Orders
Business Data
Service Orders
Defect-free
Average
71%71% $100$100 5 Days5 Days 18%18%
5.0%5.0% $1,800$1,800 45 Days45 Days 22%22%
24.0%24.0% $900$900 25 Days25 Days 60%60%
1 $395 12.3 Days
ordersorderse cyclee cycle
timetime
Cost/Cost/
orderorder costcost
Focus ofFocus ofprogramprogram
developmendevelopmen
tt
AveragAverag% of% of TotalTotal
-
8/9/2019 Granmetrics Exp
12/24
0
00
A
(%)
(%)
/
B C D E F G H I J K L M N
2
4
6
8
( )
25 100
80
100
Craftsmen
Cum
ulativecost
Cumulative craftsmen
Craftsmen Efficiency
The Need for
Granular Metrics
Totalhoursclearedfault
S:1
10
Fault not fixed correctly
& repeats
Pair throw: reroute around
fault
Request assistance
Refer fault to someone else
specialist
Fix fault
THE BEHAVIOR OF COST BY INDIVIDUAL WORKERS
Figure by MIT OCW.
-
8/9/2019 Granmetrics Exp
13/24
00
00
j
THE BEHAVIOR OF COST BY TYPE OF EQUIPMENT
20
20
40
60
80
100
40
% of Junctions
Line Fault Concentr ation
%o
fLineFaults
60 80 100 20
20
40
60
80
100
40
% of Equipment
Records Errors Concentration
%o
fRe
cordsErrors
60 80 100
100% of demand pair
reclamations are
concentrated in 17%
of unctions
100% of records
errors are concentrated
in 19% of equipment
Figure by MIT OCW.
-
8/9/2019 Granmetrics Exp
14/24
/
/
---
-
a b
c
/
---
-
OPERATIONAL EFFECTIVENESS: ADAPTIVE FEEDBACK USING GRANULAR METRICS
Cost Drivers
Individual Performance
Select
Associates
Best Pr actice Sessions
Supervisor
Benchmarks
Supervisor 'Supervisor Coaching'
Individual Reviews
Standard
Standard
Supervisor Review District Center Reviews
Director Process Reviews
Ind.
Total
Actual Target
a b
CostUnit
Supervisor
Total
Actual Target
Figure by MIT OCW.
-
8/9/2019 Granmetrics Exp
15/24
Profit DeProfit De--Averaging:Averaging:
Using Granular Metrics ToUsing Granular Metrics ToDrive PerformanceDrive Performance
-
8/9/2019 Granmetrics Exp
16/24
0
PROFIT MARGIN CONTRIBUTION
THE NEED FOR GRANULAR METRICS
-7x
-6x
-5x
-4x
ProfitMargin
-3x
-2x
-x
Cumulative percentage of customers1x
2x
3x
85% of Profits
46% of Profits
40% of Profits
21% of Profits
10 20 30 40
35
50 60 70 80 90 100
Figure by MIT OCW.
-
8/9/2019 Granmetrics Exp
17/24
The drivers of customer profitability in the credit card businesThe drivers of customer profitability in the credit card businesss
Profit by customer
Months oftenure
Revenue percustomer
Number ofcustomers
Balance levelper month
-
8/9/2019 Granmetrics Exp
18/24
$( )
/
/
$( )
(
)
(
)
VARIABILITY OF PROFITS BY USAGE LEVELS
20MonthlyProfit
Customer
$10-$100 $100-$200
Revenue Customer
$200-$400 $400+
10$0
$10
$20
$30
$40
$50$60
$70
75th Percentile +1.5* Box Size 'Upper
limit of connected data'
75th Percentile
Median
25th Percentile
25th Percentile -1.5*
Box Size 'Lower limitof connected data'
Figure by MIT OCW.
-
8/9/2019 Granmetrics Exp
19/24
(
)
0
/
0
FURTHER EX AMINATION OF PROFIT DRI VERS
5MM
15%
22%18%
10MM 20MM
Number of Customers
Company Scale Customer Share Customer Tenure
ProfitMargin
ROS
30MM $1000-50
-40
-30
-20
-10
10
20
30
$2000 $3000 $4000
Balance Level $ Month
10 20 30
-50
-40-30
-20
-10
10
20
30
Months of Tenure
Figure by MIT OCW.
-
8/9/2019 Granmetrics Exp
20/24
T H E E F F E C T S O F C U S TO M E R TA R G E T I N G
150,000
0
x
3.5x
Tailor offer s to individual customers
(i.e., increase response & profit with differentialoffers tailored to customer's preferences)
Offer one product to customers, based on
both r esponse & pr ofitability
(i.e., acquire customers likely to have good volume/
margin/lifetime characteristics)
Offer one pr oduct to those who will respond
(i.e., maximize response for a given offer, assuming average economics)
Offer one product to
everyone
Untargeted solicitations
yield a negative profit
5x
CumulativeProfit
Cumulative Number of Customers Solicited
300,000
}
}}}
}}}}
Figure by MIT OCW.
-
8/9/2019 Granmetrics Exp
21/24
CustomersCustomers targeting behaviorstargeting behaviorsBehaviorsBehaviors Typical competitorsTypical competitors Capital OneCapital One
zz OfferOffer 5,000 products5,000 products75 products75 products
zz Test/yTest/yearear 3030 15,00015,000
zz Service tierService tier 44 1313
zz KeKeyy metricmetric ROE, DelinquencyROE, Delinquency NPVNPV
zz TargetsTargets High response likelihoodHigh response likelihood High NPV potentialHigh NPV potential
zz Solicitations/ySolicitations/yearear 50 million50 million 400 million400 million
zz Segmentation frequencSegmentation frequencyy BiannualBiannual OngoingOngoing
zz Risk orientationRisk orientation Risk averseRisk averse Price for riskPrice for riskzz Data sourcesData sources 8080 50,00050,000
zz Staff analyStaff analysststs 2020 150150
-
8/9/2019 Granmetrics Exp
22/24
Capital OneCapital One-- achieving a Total Customer Solutions throughachieving a Total Customer Solutions through
Customer TargetingCustomer Targeting
FurtherFurtherimprovementimprovement
Variation 2
Variation 1
Test cells
SelectOffers
OfferCreation
IntereInterest rast ratestes FeesFees PromotionsPromotions
MassMassmarketingmarketing
Screen forprofitability
Full lifeFull life
cycle customercycle customer
profitabilityprofitability
Variation 2
Variation 1
Core offer
Collectcustomerresponse
-
8/9/2019 Granmetrics Exp
23/24
Success metricsSuccess metricsTypical competitorTypical competitor Capital OneCapital One
zz Customer NPVCustomer NPV
zz
Acquisition costAcquisition costzz ActivationActivation
zz PercenPercent witht with revorevolvinglving bbaalanlancesces
zz Fee income per accountFee income per account
zz ChargeCharge--off per accountoff per account
zz TenureTenure
$62$62
$77$7758%58%
25%25%
$55$55
$71$71
4 years4 years
$427$427
$67$6779%79%
40%40%
$73$73
$54$54
6 years6 years
-
8/9/2019 Granmetrics Exp
24/24
0
1)
2)
3)
4)
5)
0
5
i)
)
i)
)
i)
)
i)
)
% of Equity Investment
100% ofvalue
created
6 BUs contribute 100% of value created
16 more BUs contribute another 42% of value created
Those 22 BUs are worth $ 15.5 billion, on an equity investment of only $4.2 billion
44% of new capital investment and 53% of R&D spending over the next three years is going to BUs
earning economic profits
56% of new capital investment and 47% of R&D spending are going to BUs that will be generating
economic losses
-40
-20
-5
10
15
20
60
%o
fTotalCompan
yValueCreated
25
-21% of value created -21% of value created42% more value
created
Value Created by Business Unit
24% of
new
capital
21% ofR&D
ii
20% of new
capital
32% of R&Dii
36% of new capital
25% of R&Dii
50 75 100
SOURCES OF VALUE CREATION
20% of new capital
22% of R&Dii
Figure by MIT OCW.