munich, lmu – big data conference – march 13, 2015 · sales person said, “it makes us try to...
TRANSCRIPT
Big Data strategy beyond technology
Munich, LMU – Big Data Conference – March 13, 2015
CONFIDENTIAL AND PROPRIETARYAny use of this material without specific permission of McKinsey & Company is strictly prohibited
Big Data
7.2 billion
140 million
48 hours
1 billion
Contact to 350 universities in 35 countries
to recruit data analysts
page views per day
active users
of videos uploaded per minute
tweets every 72 hours
35%
Facebook decreased degree of separation
of all pictures taken are posted on Facebook
from 6 to 4
2
Big Data is only as relevant for a company as the business problem that it can solve
Big Data is about
getting insights
to solve high impact
business problems
out of data
Big Data projectsare primarily business initiatives andonly secondary technology efforts
3
PredictionInsights from historic datateach us what is likely tohappen
Example: Telecom provider"next product to buy"
MonitoringWe see what is happeningnow and can deriveimplications/insights
Example: Sentiment analysis of Twitter datafor Box office results
Backward looking Present Forward looking
Data MiningDo we find unusual„outliers“?
Example:Misuse of drugs
Evaluation We know we do thingsdifferently, but: is therean effect?
Example:
Cancer Center quality
Big Data use cases can be categorized by their "backward" or "forward" looking nature
Data Data Data
Analytics Analytics
Process
Increasing level of automation
Automated decision makingEstablish a „no touch“ backoffice process
Example: Automated handlingof mortage applications
Example Retail: What can they do with this?
How?
1 Created an algorithm to identify about 25 products that, when analyzed together, assign each shopper a “pregnancy prediction” score
Estimate her due date to within a small window, so Target could send coupons timed to very specific stages of her pregnancy
2
Use coupons, advertising and direct mail to entice those women or their husbands to visit Target and buy baby-related products
3
Then apply cue-routine-reward calculators to start pushing them to buy groceries, bathing suits, toys and clothing, as well
4
Identify pregnant women early to get them into the store and change shopping behavior
Example Retail: What customer data does Target have access to?
▪ Individual purchases
▪ Basket size
▪ What you looked at online
▪ Ethnicity
▪ Job history
▪ Magazines you read
▪ Bankruptcy or divorces
▪ Year you bought (or lost) your house
▪ Where you went to college
▪ What kinds of topics you talk about online
▪ Whether you prefer certain brands of coffee, paper towels, cereal or applesauce
▪ Political leanings
▪ Reading habits
▪ Charitable giving
▪ Number of cars you own
Purchase information Target can buy data about your
Demographic information including
▪ Age
▪ Marital status
▪ Kids?
▪ Which part of town you live in
▪ How long it takes you to drive to the store
▪ Estimated salary
▪ Whether you’ve moved recently
▪ Credit cards you carry in your wallet
▪ Web sites you visit
SOURCE: Press reports
Example Telecoms: The NPTB recommendation engine ranks all relevant products down to the VAS level for each customer
Customer
Product
Basic Mobile plan
iPhone mobile plan
Music VAS1 AutoRoam
IPTV Sports Package
IPTV News Package
Broadband
12% Owns Owns 87% Owns39% Owns
Owns 89% 12% 64% 97%22% 10%
15% Owns 95% 21% 31% 32% Owns
Product Probability
IPTV News Package 87%
Broadband 97%
Music VAS 95%
Recommendation – Customer View
Customer Probability
95%
39%
22%
Recommendation – Music VAS
1 VAS = value added service
Example Telecoms: “Can it really be this good…help me figure out how to operationalize this” – CEO of telco
Overview Results
Retail shop
▪ Partnered with sales staff at store to recommend products based on NPTB engine
▪ Collected feedback and uptake rates
▪ 15.3% conversion rate
▪ Salesman said, “I will definitelymake full use of it.” [referring to the NPTB engine]
Inbound telesales
▪ Partnered with sales staff at call centre to recommend products based on NPTB algorithm
▪ Collected feedback and uptake rates
▪ 13% conversion rate for up/x-sell
▪ Sales person said, “It makes us try to cross sell and up sell more”[referring to the NPTB engine]
Out-bound telesales
▪ Recommended Chinese or Complete Pack to customers selected by NPTB algorithm
▪ Collected uptake rates
▪ 238% greater sales conversion than control sample for Complete Pack
▪ 100% greater sales conversion than control sample for Chinese Pack
E-mail DM
▪ E-mailed BBOM offer to customers selected by NPTB algorithm and client heuristics
▪ Collected uptake rates
▪ 154% greater sales conversion than client selection
▪ Uncovered white space by selling to stand alone mobile customers
Out-bound SMS
▪ SMSed VAS offers for VAS 1 and VAS 2 to customers selected by NPTB engine, client heuristics, and client logreg model
▪ Collected uptake rates
▪ 95% greater sales conversion than client selection for VAS 1
▪ 33% greater sales conversion than client selection for VAS 2
AIB reduced cost of mortgage application process by 70%through digitalization
70% reduced cost for mortgage application
99% reduction in average processing time for applications (down to 5 minutes)
75% service take-up rate regarding customer satisfaction
80% of loan applications are fully handled by program; decision quality greater than for manual processing
2001 - 11 CAGR
Peer group 5%
12%
Companies that use big data effectively are more successful than their competitors
Pilot with basketdata from 14 stores
First statement mailing
Revise price strategy to price sensitive customers
GBP 1 billion given back to customers
Promotions cut by 1/4 –customer perception
improves
Assortmenttool
"Value index" integratesprice and promotion indices
Macro space optimization
Promotions data
Clubcardlaunched
Tesco baby clublaunch, followed by wine club
Tesco personal finance launch
Tesco.com launched
Roll out to international markets
Price sensitivitysegmentation
100% data
10% dataLifestyles
segmentation
Identify "gaps" in baskets
Finest launched
Standard reporting of customer insight KPIs and analytics
Clubcardrelaunchedwith key fobs
Coupons@Till
All rangereviews use
substitutability analysis
All promotions post-evaluated
Shopper panel launch
Pioneers like Tesco …Net profit (indexed)
1991 1995 2000 2005 2010
600
200
300
400
500
700
100
800
McKinsey’s perspective on an insight-driven business
Insight value chain Key principles
Strategy
Decision backwards
Build the capability by starting with the business decisions you want to drive and working backwards
Step-by-step
Focus on specific topics and set each element in place – a chain is only as strong as its weakest link
Change journey
Implementing data driven processes is against the DNA of many successful companies. Mastering change – on multiple levels – is at least as important as technology
Stra
tegy
Each insight value chain can be defined across industry application areas ...
Specific use cases per application area to be defined to detail the insight journey
Application areasExamples
Retail
PricePro-
motionRanging MediaSpace
Purchas-ing
Avail-ability
InsightFoundation
Insurance
Product mgmt.
Marketing & Sales
Policy ad-ministration
Claims mgmt.
Support functions
Asset mgmt.
InsightFoundation
Insight value chain
Healthcare
InsightFoundation
Contracting.Revenue collection
Claims/ Fraud
Provider steering
EvaluationPatient
steering.
… and translated into an insight heat map ...2 3 4 51
Basic Insight-driven
ILLUSTRATIVE HEALTHCARE PAYOR EXAMPLE
ContractingRevenue Collection
Claims/ Fraud
InsightFoundation
Application areas
Churn MgmtProvider Steering
EvaluationPatient
Steering
4 5 2
4
4 3
1 4
2
5 3
4 2
3
2
4
1
2
2 1
1
2
2
3
3
4
4
3
3
2
3
2
2
3
2
2
1
2
1
1
2
2
4
2
2
1 2
4
Use case
Data
Analytics
Technology
People
Process
Decision
Insig
ht
… and translated into an insight heat map ...2 3 4 51
Basic Insight-driven
Data
Analytics
Technology
People
Process
Insig
ht
Use case
Decision
ContractingRevenue Collection
Claims/ Fraud
InsightFoundation
Application areas
Churn MgmtProvider Steering
EvaluationPatient
Steering
ILLUSTRATIVE HEALTHCARE PAYOR EXAMPLE
4 34 2 4 2 2 2
4 5 2
4
4 3
1 4
2
5 3
4 2
3
2
4
3
2
2 1
1
2
2
3
3
4
4
3
3
2
3
2
2
3
2
2
1
2
1
1
2
2
4
2
2
1 2
4
1
TechnologyOverall, created a good foundation with common ERP and single DWH; can be expanded to other areas, where not yet leveraged
4 34 2 4 2 2 2
4 5 2
4
4 3
1 4
2
5 3
4 2
3
2
4
3
2
2 1
1
2
2
3
3
4
4
3
3
2
3
2
2
3
2
2
1
2
1
1
2
2
4
2
2
1 2
4
1
… and translated into an insight heat map ...2 3 4 51
Basic Insight-driven
Data
Analytics
Technology
People
Process
Insig
ht
Use case
Decision
ContractingRevenue Collection
Claims/ Fraud
InsightFoundation
Application areas
Churn MgmtProvider Steering
EvaluationPatient
Steering
ILLUSTRATIVE HEALTHCARE PAYOR EXAMPLE
AnalyticsOverall rated weak, since with rare exceptions application of “rule based expert systems” instead of advanced analytics
4 34 2 4 2 2 2
4 5 2
4
4 3
1 4
2
5 3
4 2
3
2
4
3
2
2 1
1
2
2
3
3
4
4
3
3
2
3
2
2
3
2
2
1
2
1
1
2
2
4
2
2
1 2
4
1
… and translated into an insight heat map ...2 3 4 51
Basic Insight-driven
Data
Analytics
Technology
People
Process
Insig
ht
Use case
Decision
ContractingRevenue Collection
Claims/ Fraud
InsightFoundation
Application areas
Churn MgmtProvider Steering
EvaluationPatient
Steering
ILLUSTRATIVE HEALTHCARE PAYOR EXAMPLE
Churn-ManagementProven process in place, executed with high diligence. However: focus not on value-adding segments (e.g., preventing high margin individuals and their families from churning)
4 34 2 4 2 2 2
4 5 2
4
4 3
1 4
2
5 3
4 2
3
2
4
3
2
2 1
1
2
2
3
3
4
4
3
3
2
3
2
2
3
2
2
1
2
1
1
2
2
4
2
2
1 2
4
1
… and translated into an insight heat map ...2 3 4 51
Basic Insight-driven
Data
Analytics
Technology
People
Process
Insig
ht
Use case
Decision
ContractingRevenue Collection
Claims/ Fraud
InsightFoundation
Application areas
Churn MgmtProvider Steering
EvaluationPatient
Steering
ILLUSTRATIVE HEALTHCARE PAYOR EXAMPLE
Contracting and EvaluationWeakest overall rating, due to lack of analytics and people skill in applying it, e.g., in hospital price negotiations; starting in “Contracting” more promising, as established process already in place
4 5 2
4
4 3
1 4
2
5 3
4 2
3
2
4
3
2
2 1
1
2
2
3
3
4
4
3
3
2
3
2
2
3
2
2
1
2
1
1
2
2
4
2
2
1 2
4
… and translated into an insight heat map ...2 3 4 51
Basic Insight-driven
Data
Analytics
Technology
People
Process
Insig
ht
1
Use case
Decision
ContractingRevenue Collection
Claims/ Fraud
InsightFoundation
Application areas
Churn MgmtProvider Steering
EvaluationPatient
Steering
ILLUSTRATIVE HEALTHCARE PAYOR EXAMPLE
PeopleVery large variation in people skills. In the focus areas of last 18 months good skills were built; biggest lever for sustainability
4 5 2
4
4 3
1 4
2
5 3
4 2
3
2
4
3
2
2 1
1
2
2
3
3
4
4
3
3
2
3
2
2
3
2
2
1
2
1
1
2
2
4
2
2
1 2
4
1
Data
Analytics
Technology
People
Process
… and translated into an insight heat map ...2 3 4 51
Basic Insight-drivenIn
sig
ht
Expected impact:
Improving “Contracting”
Analytics for Churn Management
Additional potential in claims
EUR 500 million
-40% Churn
EUR 150 million
Use case
Decision
ContractingRevenue Collection
Claims/ Fraud
InsightFoundation
Application areas
Churn MgmtProvider Steering
EvaluationPatient
Steering
ILLUSTRATIVE HEALTHCARE PAYOR EXAMPLE
ContractingRevenue Collection
Claims/ Fraud
InsightFoundation
Application areas
Churn MgmtProvider Steering
EvaluationPatient
Steering
4 5 2
4
4 3
1 4
2
5 3
4 2
3
2
4
1
2
2 1
1
2
2
Use case
Data
Analytics
Technology
People
Process
Decision
Insig
ht
3
3
4
4
3
3
2
3
2
2
3
2
2
1
2
1
1
2
2
4
2
2
1 2
4
… which is the starting point for the Insight Journey
Sequencing of initiatives is
driven by ease of implemen-
tation and financial impact
1
1
2
2
3
4
4
5
5
3
2013 2014 2015 20172016
2013 2014 2015 20172016
Most common challenge is the interface between people and software
Put in place a Build insightIntegrate insight
dedicated insight capabilities and trainteam into business
team with strong cross-functionallyunits through
analytical talent throughout insightformal processes
journey
The category accelerator: build capabilities to leverage insights through "navigated self-discovery" in a dedicated facility
Plan integration
▪ Category strategy
▪ Customer decision trees
▪ Private label strategy
▪ Promotion optimizer
▪ Visual merchandising
▪ Optimal store clustering
▪ Inventory management
▪ COGS playbook
▪ Merchandising strategy
▪ Supply Chain Management
▪ Negotiation playbook
SKU health check
Node
NumberNode Name SKU Ref # SKU Description
# Name # Description
1 7¼ General Purpose (Value)1983280 SAW CIRC.7 1/4"13AM SK548001
1 7¼ General Purpose (Value)1197016 SAW CIRCULAR 7 1/4" 15AM
1 7¼ General Purpose (Value)00275540 SAW CIRCULAR SAW 7 1/4" 15AM
2 7¼ General Purpose (Premium)00275497 SAW CIRCULAR 7 1/4" 15AM
1 7¼ General Purpose (Value)1983224 SAW CIRC. 7 1/4" 15AM CS10
1 7¼ General Purpose (Value)00275552 SAW CIRCULAR 6 1/2" 18VT(BARE)
2 7¼ General Purpose (Premium)1197142 SAW CIRC. 7 1/4" 15AM 5007FAK
2 7¼ General Purpose (Premium)1197280 SAW CIRC. 71/4+LAMP 5007MG
2 7¼ General Purpose (Premium)8400048 SAW CIRC. 7 1/4" 14AM HCS002
1 7¼ General Purpose (Value)19835530 SAW CIRC.LASER 7 1/4" 15AM
1 7¼ General Purpose (Value)1983247 SAW CIRC. 7 1/4" 15AM CS5
5 6½ General Purpose (Value)19835163 SAW CIRC. 7 1/4" 15AM SHD77
Total Node
5.6 6.0 5 .8 5.5 5.7 6.0 6.0 Prox. Plus
4.8 4.3 4 .5 4.7 4.5 4.5 4.4 B-Box Plus
Is my range profitable? Should I include this product in my core or core plus assorment?Should I include this product in my core or core plus assorment?
Competitor
price tie r
3Net
margin $
m illion
3Net
margin $
rank
3Net
margin %
Growth in
1st margin
%
CO
RE
Clu
ste
r 1
Clu
ste
r 2
Clu
ste
r 3
Re
gio
n 1
Re
gio
n 2
Re
gio
n 3
CORE Clus ter 1 Clus ter 2 Clus ter 3 Region 1 Region 2 Region 3
$0.5 28% 3%
Better $0.06 1 24% 1% 2 1 3 1 3 2 2 Pot. Delete
Best $0.06 2 26% 3% 5 5 6 6 5 2 5 B-Box Core B-Box Plus Prox. Plus Prox. Plus B-Box Plus B-Box Plus B-Box Plus
Best $0.02 9 20% 5% 5 6 5 5 6 6 5 B-Box Core B-Box Plus B-Box Plus B-Box Plus Prox. Plus Prox. Plus B-Box Plus
Better $0.04 5 31% 3% 6 4 7 6 5 8 7 Prox. Core B-Box Plus Prox. Plus Prox. Plus B-Box Plus Prox. Plus Prox. Plus
Best $0.01 11 24% 4% 4 4 5 5 5 4 3 Pot. Delete
Best $0.01 13 23% 5% 5 6 3 3 5 6 6 B-Box Core B-Box Plus B-Box Plus B-Box Plus B-Box Plus Prox. Plus Prox. Plus
Better $0.05 4 33% 3% 5 4 6 4 6 7 6 B-Box Core B-Box Plus B-Box Plus B-Box Plus Prox. Plus Prox. Plus B-Box Plus
Best $0.05 3 36% 5% 6 6 8 6 7 7 7 Prox. Core B-Box Plus Prox. Plus Prox. Plus Prox. Plus Prox. Plus Prox. Plus
Better $0.03 7 39% 3% 5 3 5 6 6 6 5 B-Box Core B-Box Plus Prox. Plus Prox. Plus B-Box Plus
Better $0.01 12 29% 5 5 5 6 5 5 3 Pot. Delete B-Box Plus Prox. Plus
Best $0.02 8 29% 5 6 6 5 5 5 7 B-Box Core B-Box Plus B-Box Plus B-Box Plus Prox. Plus
Best $0.01 14 23% 5 4 6 4 5 4 6 Pot. Delete B-Box Plus B-Box Plus B-Box Plus
Is it priced well? Is it driving profit and/or profit growth? What is the SKU score? Should I include this in CORE PLUS?
Score
Average o f decile performanceRecommendationProfit
Profit
growth
GMROI tree
From-To category strategy
Standardized category plansCoffee - Executive Summary Category Role: Core
From To
Customer
* Under- indexing in Budget customers compared to the Supermarket as a whole
* Two clusters (Café Culture and Instant), with no range differences and minimal planogram flexing
Customer
*Winning in all customer segments including Budget Customers, ENGAGING customers!
* Moving to 3 clusters (Café Culture, Everyday, Value) with tailored range and space allocation
* Targeted customer offers developed e.g.. 20 Greek Stores focusing on "Turkish & Greek Coffee"
Own Brand
* Own Brand Penetration 2.40% (Aztec 03/07/2011)
* Homebrand 1.10% (-8.1% on LY)
* Select 1.30% (+86.8% on LY)
* Macro n/a
Own Brand
* Own Brand penetration 4% by end of FY12 focussing on Macro and Select growth
* Homebrand: target 1.25%
* Select: target 2.50%
* Macro: target 0.25% FY12 (Launch 3 skus Q3)
Financials
* Market share of 43.50% by end of FY12
* F2012 - Sales +7.19% on LY ($398,597,284) Budget, inflation anticipated at 3-4% due to price increases
* Developing a plan to invest in key pack groups to drive sales and make budget
* F2012 - Profit +12.47% on LY, Rate 24.35%
Value
* Reduce Churners and remove Losers, reinvest funds in more effective promotions
* Reduce number of promotions by 20% and move towards FCP in 3 nodes and additional 'loser' lines
* Only invest in All Star and Traffic Builder activity
* Tailored promotional program for Premium and Turkish/Greek stores
Value
* 2% All Stars, 0.43% Traffic Builder, 45.72% GP Builders, 45.2% Churners, 16.5% Losers
* Knock-down has been limited to Nescafe Blend 43 150g (coming off in September)
* 55% sales on promotion, $51.2 Deal subsidy, $17.4 unfunded
*Promotional program consistent across all stores
Financials
* TY MAT market share is 43.08% (Aztec 03/07/2011)
* F2011 - Sales +1.42% on LY ($371,843,686)
* F2011 - Units +0.76% on LY
* F2011 - Profit +22.40% on LY, Rate 24.95%
* F2011 - Swell Investment $747K
COGS
* $18.6m eCOGS opportunity identified
* Nestle Represent 70% of the total eCOGS opportunity
* High shrinkage in Roast and Ground due to damage at the store level
COGS
* 3% ($8.6m) COGS reduction by end of FY12
* Work with Vendors to adjust packaging to reduce shrinkage to 0.7% and re-coup losses in Roast and
Ground
* Negotiate waste and markdown agreements with Vendors to capture Shrink losses
Merchandising
* Range tailored to 2 Clusters: Cafe Culture and Instant
* 684 stores have base (7 bays) or above layouts
* SFP not currently used in the category
- 145/229 in SFP's in base layout 63%
- 63/229 being used as 2 ctn fill 27%
Merchandising
* Range is tailored to 3 Clusters, Cafe Culture, Everyday and Value
* Specific Turkish/Greek stores offered a tailored range
* Specific small range products remain to tailor to communities (Byron Bay)
* Category dominated by SFP
Range
* Recent significant range reduction in 2010/11
* Trial Range Reduction of 20% in Roast and Ground Café Cluster: sales -5.30%, volume -4.50% and profit -6.4%
on control stores (breadth and depth of range important to these customers)
Competition
* Coles: Maintain competitiveness particularly through promotions (enforce clash policy)
* Aldi: HB used to match everyday, with Select Knockdown to provide improved perception
* Costco: Value Cluster to focus on Bulk packs and Value
Range
* Medium Term sku count to remain stable, due to recent reductions in range, however greater range
variation between clusters, maintain overall SKU count, with reductions by cluster.
* Clearly defined price tier strategy in each segment (particularly Roast and Ground)
Competition
* Coles: category is being used as a traffic driver, OB offer growing (but penetration similar to WOW)
* Aldi: maintain a small range which includes Nescafe Blend 43 100g & Gold 80g (local)
* Costco: Moccona 2x400g, Blend43 500g, Illy 250g and range of Kirkland Roast & Ground
Pricing
Competitive ranging Brand-price ladder
…
8 - 10 tailored modules, e.g.,
Loyalty card data Click-stream dataInternal transaction data Sales and profit data Nielsen/IRI/GfK, etc data
MODULES
Analysts act as
navigators in the category accelerator
10Ranging1 2
Key points and summary
Always start with the use case for Big Data from a business perspective
Implement Big Data "business back", not "technology forward"
Big Data is a business initiative
More data is better
Start with internal data – value is in linking different sources
Learn to accept messiness
Big Data@scale requires automation
Transformational impact from Big Data by enabling data-driven processes
However, not all use cases need automation across all alyers
Don’t be afraid of analytics
Nobel prize level analytics is not required
There are very innovative ways of sourcing good solutions
From ToSmall,i.e., averages andsamples are"good enough"
Comprehensive,i.e.,all available data is analyzed on the most granular level
Data volumes
Exact,i.e., zero-fault tolerance of input data
Messy,i.e., incorrect single data points are acceptable, since overall number of data points is huge
Data quality
Causality,i.e., cause-effectconfirmend by data
Correlation,i.e., significant correlation between input-output data good enough, even if causality is not understood
Management
conclusion
From ToSmall ComprehensiveData volumes
Exact MessyData quality
Causality CorrelationManagement
conclusion
Why?