April 11 2018 3 KB Introduction Javier Anta Callersten Partner at BCG in London Leading BCG Gamma for Western Europe Core member of global retai l leadership team leading Advanced Analytics in Retail ID: 928400
Download Presentation The PPT/PDF document "Transforming retail through advanced ana..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Transforming retail through advanced analytics &AI
April 11, 2018
3 KB
Slide2Introduction
Javier Anta
Callersten
Partner at BCG in London
Leading
BCG Gamma for Western EuropeCore member of global retail leadership team, leading Advanced Analytics in RetailHas led various advanced analytics programs in retailers across 14 different countries
Markus Hepp
Partner at BCG in Cologne
Leading BCG's Consumer Practice in GermanyCore member of global retail leadership team, leading the Retail Sector in EuropeSpecialized in retail transformations across major retail sectors
Slide3Contents
Advanced Analytics & AI as an opportunity for retailers
Main challenges that need to be overcome
Our beliefs on how to successfully transform through
AA & AI
and typical journeys1.2.3.
2 KB
Slide4The world is changing rapidly, driven by digital
and data…
Unprecedented visibility on customers, business activities and
market trends
Omnichannel
,
sensors, always connected
Processing power, storage and robotics ready for AI and automation
Ready to engage with brands anytime, anywhere
Disintermediation, sharing economy, crowdsourcing, etc.
Data
Explosion
Pervasive
Digitalization
Enabling
Technologies
New
Consumer
New Market
Forces
Ability to build and monetize data assets drives competitive advantage
1,964 KB
Slide5… leading to high impact opportunities
Segment-of-1 content creation and recommendations
Real time fraud detection with lower
risk of false positives
Predictive asset maintenance across industries
Personalized health services
$1B
Annual
Value-50%Fraud False Alerts$1BCostSavings-50%Antibioticuse
Hyper-personalized offers and continuous test and learn
Optimize resource utilization through multiple neural nets
Telematics to
optimize routing, personalized service
Predict next customer contact channel and product for servicing request
3x
Net
Incremental
Revenue
-40%
IT Cooling Costs
100M
Miles Reduction
88%
Accuracy
122 KB
Slide6Large part driven by a huge step-up in analytics: self-learning AI systems can now be leveraged by businesses
Financial reports
Geo analysis
Heat maps
....
Prescriptive
Predictive
Descriptive
Business IntelligenceTraditional AnalyticsDeep LearningCampaign ManagementSentiment analysis....
Efficient personalization
Context-aware (e.g. mobile) recommendations
....
Structured
Low Volume
Batch Load
Type of Insights
Unstructured
High Volume
Real Time
Type of Data
Artificial Intelligence
Machine Learning
Next best action
Recommendation engine
Churn prediction
....
Slide780%
60%
20%
0%
40%
20%
80%
0%
40%
60%
Large effect of AI on
Offerings
Manufact
.
Transp. / Travel
Public Sector
Insurance
Utilities
Prof.
Serv
.
Logistics
IT & Tech.
Logistics
Energy
Manufacturing
Constr
.
Energy
Automotive
Consumer
Public Sector
Insurance
IT & Tech.
Ent. / Media
Cap. Markets
HC
Equip
. /
Serv
.
Chemicals
Transp
. /
Travel
HC
Equip
. /
Serv
.
Consumer
Construction
Telco
Banking
Automotive
Pharma
/
Biotech
Agriculture
Cap.
Markets
Telco
Retail
Utilities
Prof.
Serv
.
Chemicals
Pharma
/
Biotech
Retail
Ent. / Media
Agriculture
Banking
Large effect of AI on
Processes
Across industries, AI is expected to have a strong impact in coming years
In 5 years
Today
% of
respondents
% of
respondents
Average
Source: Joint BCG-SMR research, AI@BCG
Slide8Retail: Use cases across the entire value chain
Non-exhaustive
Production & logistics
CRM /loyalty / marketing
Selling
Store layout & build-upCategory management
Inventory optimization across the logistic network
Utilization optimization of transport capacity
Routing optimization Origin transparency through blockchainOptimized joint planning through shared data with suppliersSupplier risk management through shared data & analyticsReal-time inventory tracking through RFIDAccelerating picking through augmented reality
Trend detection + real time customer feedback
Optimized & localized assortment and pricing structure
Identification of savings potential with price elasticity analysis of parts
Predictive modeling of new product launch uptake based on non-traditional inputs (e.g. social buzz)
Assortment optimization (e.g. modeling impact of assortment change)
Customer centric store lay-out (online & offline)
Optimize store locations
Seamless
Omnichannel
Check-out free shopping
Perfect Store 2.0 and
on-premise
customer activation
Dynamic assortment based on real-time conditions
Store workforce optimization
Personalized 1:1 promotions and targeting
Optimized mass-market promotions
Automated targeted buying process for online ads
Loyalty program optimization through user behavior and incentive response analytics
Recommendation engines for app and website optimization
Promotions optimization through automated post-event analytics
Customer churn reduction
Automation /
robotization
Regional segment detection and assortment optimization
Real-time in-store personalized promotions
Smart markdowns
Dynamic (online) pricing
Monitor and improve performance with real-time end-to-end dashboards
Predictive demand forecasting based on non-traditional inputs (e.g. social buzz)
Cross-format selling
Digitization-based
Analytics-based
7 KB
Slide9Hyper- personalization
Selected AA & AI use case examples in Retail
1
2
4
5Massive re-allocation of investment to most effective+250m$ margin
Value >35x
programme cost
150% increased marketing engagement300% increase NIR100m$ additional revenue year 1Supply chain optimizationLeaner footprintBetter utilizationBetter fulfillment
10% reduction in warehousing/ transport costs
Locali
-
zation
More relevant to customer, more profitable E2E
350-450m$ unlocked sales
Promotion optimization
Smart markdowns
44% savings in markdown costs
~€20M in single market
3
Slide10AI
algos
can deliver unprecedented levels of accuracy for promo analysis and forecasting
Base sales
+ uplift
Halo:
Foot Fall
Halo:
Complementarity
Pull Forward
Cannibalization
Discount
Vendor funding
Machine Learning Elastic Net
algo
accounting for 20+ dimensions
Category-SKU regression
Pre & post customer composition
Conditional probability
Machine Learning Time series
with frequency modulation
Entirely bespoke solution
Battle of the
algorithms
1
Slide11Hyper personalization at Starbucks: Each customer's experience personalised "just for the individual"
Personalised offers and experience
Anticipate customer behaviour
Drive transaction and ticket
~3x improvement in campaign results run rate
3x+ incremental revenue per redeemer per year
2
Slide12Example: Email offer optimization
12 million
30
400,000
Segmentation
Individualisation12 million380,00032No. People :
No. Variants :
People/variant:
Multiple machine learning models individualise the message for content as well as the point on the economic efficiency curve. The sole reason for a cluster of >1 is that some people simply like the same thingDynamic construction of the message Real-time tracking of the progress for the experience2
Slide13The analytics engine was built based on three key dimensions
Customer DNA
Pathways
Habits/Preferences
Social graphs
HeadroomPropensitiesOffer DNATypeProduct (s)Sequence, TimingReward levelContext & locationLocationTime/day
ProximityWeather
What is the right
Offer forthis customer ?What is the right curriculum? What behavior(s) do we want for this customer? When is the right context (location, time and format)to place the offerthis customer ?2
Slide14An illustrative view on the analytics engine
ID
1
2
3
4
5
6
789
1011
12
13
14
15
16
17
18
19
20
Lo
cation
0000001-ABCDEFG-00001
Locn Type A
0000001-ABCDEFG-00002
Locn Type B
0000001-ABCDEFG-00003
Locn Type C
0000001-ABCDEFG-00004
Locn Type D
0000001-ABCDEFG-00005
Locn Type E
0000001-ABCDEFG-00006
Locn Type F
0000001-ABCDEFG-00007
Locn Type G
0000001-ABCDEFG-00008
Locn Type H
0000001-ABCDEFG-00009
Locn Type I
0000001-ABCDEFG-00010
Locn Type J
0000001-ABCDEFG-00011
Locn Type K
0000001-ABCDEFG-00012
Locn Type L
0000001-ABCDEFG-00013
Locn Type M
0000001-ABCDEFG-00014
Locn Type N
0000001-ABCDEFG-00015
Locn Type O
0000001-ABCDEFG-00016
Locn Type P
0000001-ABCDEFG-00017
Locn Type Q
0000001-ABCDEFG-00018
Locn Type R
0000001-ABCDEFG-00019
Locn Type S
0000001-ABCDEFG-00020
Locn Type T
0000001-ABCDEFG-00021
Locn Type U
0000001-ABCDEFG-00022
Locn Type V
0000001-ABCDEFG-00023
Locn Type W
0000001-ABCDEFG-00024
Locn Type X
0000001-ABCDEFG-00025
Locn Type Y
0000001-ABCDEFG-00026
Locn Type Z
0000001-ABCDEFG-00027
Locn Type AA
0000001-ABCDEFG-00028
Locn Type AB
0000001-ABCDEFG-00029
Locn Type AC
Day time preference
-1.000
1.000
Avg. headroom
0.0000
0.2500
Risk variations
0
60
Avg. game score
0.0
100.0
|1.000
Engagement score
Risk score
Headroom
Engagement
score
Risk of attrition
and fading
Headroom
to grow share of wallet
Each row is one customer
Product preferences captured
Location
preference
Time
of day preference
43
54
2
59
11
90
27
36
0
27
100
4
77
69
86
16
63
2
28
53
33
46
42
44
100
18
1
23
25
1
4
25
73
89
40
10
94
60
53
5
11
38
46
12
3
58
38
5%
3%
6%
8%
1%
12%
2%
12%
5%
15%
Product
Customer-Product propensity score
2
Illustrative
Slide15Significant impact was
achieved
150%
increase in marketing engagement
We are building a true, real-time, personalization capability which will begin to power personalized experiences and communications within our app... Our digital flywheel momentum accelerated ... with the launch of
true one-to-one personalization ... Starbucks hyper-personalized e-mail reward offerings – with more than 400,000 variations – have more than doubled customer response rates over previous segmented email campaigns, translating into increased customer engagement and, importantly, accelerated spend. Starbucks has delivered personalized offers to customers directly on the front screen of the mobile app. By early 2017, the company expects to complete the rollout of suggested selling and recommendations (suggesting items for pairing or additions to a customer’s order) during Mobile Order and Pay checkout, which the company believes will further fuel engagement and growth.
Kevin Johnson, President & COO
Matt Ryan, Chief Strategy Officer
Howard Schultz, Chairman and CEO
Our new one-to-one personalized marketing capability ... will prove to be a
retail industry game changer
.
300%
increase in net incremental revenue
=
+$100M
in year 1
2
Slide16When delisting this chair how many should we send to each store and how should we price it?
Slide17Goal is to improve management of discontinued goods
Decrease manual /
ad hoc efforts in management of discontinued goods
Improve predictability of demand, assist in stock management
Decrease markdown costs
Increase overall profitability
Slide18We built a robust optimization engine...
Base demand model
Uplift model
Substitution & Complementation model
Optimization engine
Predicted sales volumes by store over time without discounts
Predicted incremental demand by store over time due to discounts
Predicted impact on demand by store due to presence of other products
Allocation & markdown by store over time to maximize net profit
Time-series forecasting (Prophet – additive regression models), Bayesian models
Hierarchical model selection, exponential regression, Decay effects
Association rules, basket analysis
(Mixed integer) linear programming, stochastic optimization
Slide19...to be integrated into business processes
Allow for read and react during markdown period
Integrate into ways of working
Incorporate business rules
Slide20Model found
44% savings in markdown costs,
~
€20
M in single market
Slide21However, most players struggle to capture value
Only 15% of companies with big data investments have put solutions into production
Art of the possible not well understood
Talent supply limited
Pressure on profitability impedes deep investment
New ways of working requiredLegacy technology and trapped dataProcesses and operating model designed for weekly vs. real time
Innovation culture difficult to institute – stifling of new ideas
117 KB
Slide22´To take a use case from idea to production is
10%
algorithms,
20%
technology,
70% about changing how people work'BCG Gamma2 KB
Slide23Three observed approaches companies take
Analytics led
"Lets hire a bunch of data scientists and find problems to solve"
Data/tech led
"Lets collect and clean all the data and then find problems to solve"
Business led"I have a problem, how can analytics help me solve it"
12 KB
Recommended
Slide24Recommended approach is to think big, start small, grow fast
Start with the business opportunity
Build, test, iterate
Scale to solution
Transform organization
Business first
Value focus
Lean technology
Right design
Practical application of AI and Big data
Well defined use cases
Iterative technology scale up
Purpose fit tools from existing technologies
New ways of working
Analytics and business strategy in lock-step
Right organization and processes
Advanced analytics as BAU
15 KB
Slide25An integrated approach is required to actually change
Analytics Transformation
Technology & Deployment
Strategic design
Data & Analytics
Extensive use of real-time data; deep learning and AI analyticsAnalytics and digital as an integrated part of the overall strategy, approach and governanceScalable technology, real-time access, secure platform
People &
Capabilities
Ways of working;
agile approach; rapid test & learn;
developing capabilities, acquiring and developing talent
21 KB
Slide26Use cases: rapid testing and scaling is essential
Pressure test one use case
Set ambition
Define & evaluate specifications
Assess data quality & accessibility
Make go / no go decision
Launch MVP in market and improve through
test and learn
Run agile sprints to test solution “in-market” and learn how to improveDesign and test new ways of working
Run technology in controlled environmentCommit to scale-up
Build customized Proof of Concept to validate business case and feasibility
Backtest
on historical data
Confirm value
Put first brick of technology in place
Agree on plan to incubate
Value creation
Scale up solution, transform organization, increase value impact
Run technology and business process at scale
Analytics resources/ governance in place
Teams trained
Client capability to own full solution in place
P&L
neutral in first 12 months, with exponential growth beyond
Articulated case for value capture
Tangible prototype with business case and plan to execute
MVP with impact assessment and scaling plan
Full scale solution integrated into environment
New ways of working instilled in your team
Outcomes
6–12+ weeks
2–4 weeks
3–6 months
6+ months
Prototype
Proof of Concept
Incubate
Scale
26 KB
Slide271 KB