capturing value from big data through data driven business models prensetation
Post on 10-Jan-2017
555 Views
Preview:
TRANSCRIPT
Capturing Value from Big Data through Data-Driven Business Models
Patterns from the Start-up world
Philipp Hartman, Dr Mohamed Zaki and Prof Andy Neely
Cambridge Service Alliance University of Cambridge
“Data is the new oil”1
1 various authors, e.g. Clive Humby
0
5000
10000
15000
20000
25000
30000
35000
40000
45000
2005 2010 2015 2020
Data volume per year (Exabytes)2
2 IDC's Digital Universe Study, December 2012
56%
Top Priority: “How to get value from big data” 3
3 Gartner “Big Data Study” 2013
How to get value from Big Data?
3
OpKmizaKon of exisKng service
Data Driven Business Models1
Based on this motivation the research question was developed
4
What types of business models that rely on data as a key resource (i.e. data-‐driven business models) can be found in start up companies?
How to analyse data-‐driven business
models?
Sub
quesKo
ns
Data-‐driven business model framework
How to idenKfy paVerns?
Research
Que
sKon
Clustering
The research was done in five steps
5
Case studies Finding PaVerns
Data collecKon & coding
Build the framework Literature Review
How to analyse data-‐driven business
models?
How to idenKfy paVerns?
The first step was a literature review with three different topics
6
Literature Review
Big Data
DefiniKon
Value CreaKon
Business Model
DefiniKon
Business Model Frameworks
Related Work
Data driven business Models
Cloud business models
Case studies Finding PaVerns
Data collecKon & coding
Build the framework Literature Review
Business model key components were synthesized from existing frameworks
ExisKng Business Model Frameworks
-‐ Chesbrough & Rosenbloom 2002 -‐ Hedman & Kaling 2003 -‐ Osterwalder 2004 -‐ Morris 2005 -‐ Johnson, Christensen et. al. 2008 -‐ Al-‐Debei 2010 -‐ Burkhart 2012
Value Ca
pturing
Value Crea@o
n Key Resources
Key AcKviKes
Cost structure
Revenue Model
Customer Segment
Value ProposiKon
Business Model DefiniKon Business Model Key Components
-‐ No universally accepted definiKon of the concept (Weill, Malone et al. 2011)
-‐ Most definiKons refer to value crea@on & value capturing
The literature review identified several gaps
8
• LiVle academic research on big data and value creaKon – mostly whitepapers
• Gap in literature: data-‐driven business models
• OVo, Aier (2013) interesKng paper but limited to specific industry > no generalizaKon possible
• Similar research for cloud business models (cf. Labes, Erek et. Al. 2013)
Case studies Finding PaVerns
Data collecKon & coding
Build the framework Literature Review
The framework was build from literature starting from the key components
Data-‐Driven-‐Business Model
Data Sources
Internal exisKng data
Self-‐generated
Data
External
Acquired Data
Customer provided Free
available
Open Data
Social Media data
Web Crawled Data
Key AcKvity
Data GeneraKon
Crowdsourcing
Tracking & Other Data
AcquisiKon
Processing
AggregaKon
AnalyKcs
descripKve
predicKve
prescripKve VisualizaKon
DistribuKon
Offering
Data
InformaKon/Knowledge Non-‐Data Product/Service
Target Customer
B2B
B2C
Revenue Model
Asset Sale Lending/RenKng/Leasing Licensing
Usage fee
SubscripKon fee
AdverKsing
Specific cost advantage
Data-‐Driven Business Model
Data Sources
Key AcKvity
Offering
Target Customer
Revenue Model
Specific cost advantage
Data collecKon & coding Case studies Finding
PaVerns Literature Review Build the framework
Features for each dimension
Data-‐Driven Business Model Framework
Business Model Key Components (Dimensions)
Data Sources Features for data sources
Synthesizing the different sources leads to the taxonomy
10
Data Sources
Internal
exisKng data
Self-‐generated Data
External
Acquired Data
Customer provided
Free available
Open Data
Social Media data
Web Crawled Data
Dimension: Activities
11
Key AcKvity
Data GeneraKon
Crowdsourcing
Tracking & Other
Data AcquisiKon
Processing
AggregaKon
AnalyKcs
descripKve
predicKve
prescripKve VisualizaKon
DistribuKon
Dimension: Offering
12
Offering
Data
InformaKon/Knowledge
Non-‐Data Product/Service
Dimension: Revenue Model
13
Revenue Model
Asset Sale
Lending/RenKng/Leasing
Licensing
Usage fee
SubscripKon fee
AdverKsing
Dimension: Target Customer
14
Target Customer B2B
B2C
Data collecKon & coding
The final framework
15
Case studies Finding PaVerns Literature Review Build the
framework
Data-‐Driven-‐Business Model
Data Sources
Internal exisKng data
Self-‐generated Data
External
Acquired Data
Customer provided
Free available
Open Data
Social Media data
Web Crawled Data
Key AcKvity
Data GeneraKon Crowdsourcing
Tracking & Other Data AcquisiKon
Processing
AggregaKon
AnalyKcs
descripKve
predicKve
prescripKve VisualizaKon
DistribuKon
Offering
Data
InformaKon/Knowledge
Non-‐Data Product/Service
Target Customer B2B
B2C
Revenue Model
Asset Sale
Lending/RenKng/Leasing
Licensing
Usage fee
SubscripKon fee
AdverKsing
Specific cost advantage
Data collection and coding
16
Case studies Finding PaVerns
Build the framework Literature Review Data collecKon
& coding
Data collecKon Data analysis Sampling
The data was generated using public available sources
17
Tag: “big data” “big data analyKcs” 1329 companies
Data collecKon
Company informaKon • Company websites • Press releases
Public sources
• Coding of sources using data driven business model framework
• Nvivo
Data analysis
299 Sources ~3 sources/comp
Sampling
100 Companies
cleaning
Random sample
100 binary feature vectors
Overall Analysis: Data Source
18
0% 10% 20% 30% 40% 50% 60%
Acquired Data
Customer&Partner-‐provided Data
Free available
Crowd Sourced
Tracked & Other
Note: Sum > 100% as companies might use mulKple data sources
• >50% of companies rely on free available data
• >50% of companies use data provided by customers/partners
Overall Analysis: Key Activities
19
0% 10% 20% 30% 40% 50% 60% 70% 80%
AggregaKon
AnalyKcs
DescripKve AnalyKcs
PredicKve AnalyKcs
PrescripKve AnalyKcs
Data acquisKon
Data generaKon
Data processing
DistribuKon
VisualizaKon
• >70% of companies use analyKcs -‐ mostly descripKve
Note: Sum > 100% as some companies rely on mulKple revenue models
Overall Analysis: Revenue Model
20
0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50%
AdverKsing
Asset Sales
Brokerage Fees
Lending RenKng Leasing
Licensing
SubscripKon fee
Usage Fee
No informaKon
• Majority of revenue models based on subscripKon and/or usage fee
• No informaKon about the revenue model as many companies are in an early stage
Note: Sum > 100% as some companies rely on mulKple revenue models
Overall Analysis: Target Customer
21
70%
17%
13%
B2B B2C both
• There seems to be a noteworthy predominance of B2B business models
• But no reference data found
BM patterns were identified using a clustering approach
22
Ketchen, David J.; Shook, Christopher L. (1996): The ApplicaKon of Cluster Analysis in Strategic Managment Reserach: An Analysis and CriKque. In: Strat. Mgmt. J. 17 (6). Han, Jiawei; Kamber, Micheline (2011): Data mining. Concepts and techniques. Mooi, Erik; Sarstedt, Marko (2011): Cluster Analysis. In: A Concise Guide to Market Research. S. 237-‐284. Miligan, Glenn W. (1996): Clustering ValidaKon: Results and ImplicaKons for Applied Analyses. In Phipps Arabie, Lawrence J. Hubert, Geert de Soete (Eds.): Clustering and classificaKon. pp. 341–376.
Case studies Data collecKon & coding
Build the framework Literature Review Finding
PaVerns
2. Clustering method
1. Clustering Variables
3. Number of Clusters
4. Validate & Interpret C.
7 Business Model Cluster were identified
23
Cluster 1 2 3 4 5 6 7
Data Sou
rce
Acquired Data 0 0 1 0 0 0 0
Customer-‐provided Data 0 1 1 0 0 1 1
Free available 1 0 1 0 1 0 1 CrowdSourced 0 0 0 0 0 0 0
Tracked, Generated & other 0 0 0 1 0 0 0
Key AcKvity AggregaKon 1 0 0 0 0 1 1
AnalyKcs 0 1 1 1 1 0 1 Data acquisKon 0 0 1 0 0 0 0 Data generaKon 0 0 0 1 0 0 1
Number of companies 17 28 5 16 14 6 14 Type A B -‐ C D E F
6 significant Business Model types were identified
24
Type B: “AnalyKcs-‐as-‐a-‐Service”
Type C: “Data generaKon & AnalyKcs”
Type D: “Free Data Knowledge Discovery”
Type A: “Free Data Collector & Aggregator”
Type E: “Data AggregaKon-‐as-‐a-‐Service”
Type F: “MulK-‐Source data mashup and analysis”
The 6 BM types are characterised by the key activities and key data sources
25
Type F
Type A Type D
Type E Type B
Type C
AggregaKon AnalyKcs Data generaKon
Free
available
Custom
er
provided
Tracked &
gene
rated
Key ac@vity
Key Da
ta Sou
rce
Type D: “Free Data Knowledge Discovery”
1. DealAngel 2. Gild 3. Insightpool 4. Juristat 5. Market Prophit 6. MixRank 7. Numberfire 8. Olery 9. PeerIndex 10. PolyGraph 11. Review Signal 12. Tellagence 13. traackr 14. TrendspoVr
-‐ Free available
-‐ Social Media -‐ Open Data -‐ Web Crawled
B2B B2C
Key AcKviKes
Revenue Model
Key Data Source
-‐ AnalyKcs
Target Customer
0 5 10 15
DescripKve
PredicKve
PrescripKve
0 2 4 6 8
SubscripKon Usage Fee AdverKsing
Brokearge Fees No InformaKon
Companies
Type D: Examples
27
“Using patent-‐pending technology, Gild evaluates the work of millions of developers so companies using Gild’s talent acquisiKon tools know who’s good and can target the right candidates.” • Key Data: Free available websites
(GitHub, Google Codes)
• Key AcKviKes: AnalyKcs
• Revenue Model: Monthly subscripKon
• Target Customer: B2B
“ Our goal is to provide the most accurate and honest reviews possible by using the data consumers create. We listen to the conversaKons, analyze them and visualize them for consumers.” • Key Data: TwiVer
• Key AcKviKes: AnalyKcs
• Revenue Model: AdverKsing
• Target Customer: B2B (B2C)
Finding PaVerns
The cases studies will be validated the framework and the clustering
28
Data collecKon & coding
Build the framework Literature Review Case studies
4 case studies with companies from the sample such as
Purpose:
1. Validate framework & clusters
2. Illustrate business model types through examples
3. IdenKfy specific challenges
Summary
29
-‐ Findings: -‐ This study explores how start-‐up business models capture value from
big data. -‐ The study also introduces the DDBM framework with which the
business models can be studied and analysed -‐ A proposed taxonomy consisKng of six types of start-‐up business
model is developed. -‐ These types are characterised by a subset of six of nine clustering
variables from the DDBM framework. -‐ Prac@cal implica@ons:
-‐ The study helps not only future researchers to structure their work around data-‐driven business models but also companies to build new DDBMs.
-‐ The proposed taxonomy will help companies to posiKon their acKviKes in the current landscape.
Limitations & Outlook
30
LimitaKons
• Only 100 samples
• Only start up companies • Bias of data source (AngelList)
• StaKsKcal significance of clustering result
• Only public available sources used
• No statement about success of a parKcular business model
Outlook/Next Steps
1. Improve validity of findings 1. Increase sample size to test
clusters 2. More Case-‐studies to
illustrate/validate clusters
2. Include established organiza@ons
3. Develop methodology to judge (financial) performance of different business models
Further Reading
31
hVp://www.cambridgeservicealliance.org/uploads/downloadfiles/2014_March_Data%20Driven%20Business%20Models.pdf
Forthcoming Webinars
32
0ct. 13th 2014 Industry transformaKon towards a service logic: a business model approach. Speaker: Anna Vijakainen Nov.10th 2014 The B2C lock-‐in effect. Speaker: Marcus Eurich
Appendix
33
The Clustering Process
34
Variables relevant to determine clustering
(Miligan 1996)
#Variables has to match #samples (Mooi 2011)
~ 2m samples for m variables:
6-‐7 variables
Avoid high correlaKon between variables (<0.9) (Mooi 2011)
2 Dimensions: “Data source” & “Key AcKvity”
9 variables
max. correlaKon: 0,5
2. Clustering method
3. Number of Clusters
4. Validate & Interpret C.
1. Clustering Variables
The Clustering Process
35
ParKKoning
Hierarchical
Density-‐based
Grid-‐based
Clustering Method
(Han 2011)
Proximity Measure
4. Validate & Interpret C.
1. Clustering Variables
3. Number of Clusters
2. Clustering method
K-‐Medoids
Include neg. match
Exclude neg. match
Euclidean Distance
There is no “one right solution” for the number of clusters
36
large to reflect specific differences k << n
1. Use a-‐priori knowledge to determine number of clusters
2. Visual approaches 3. Rule of thumb (Han 2011):
4. “Elbow” method
5. StaKsKcal methods
𝑘 ~√𝑛/2 →𝑘 ~ 7
k?
2. Clustering method
4. Validate & Interpret C.
1. Clustering Variables
3. Number of Clusters
Several different approaches (Pham 2005, Mooi 2011, Han 2011, EveriV et. al. 2011):
“Elbow” method
37
“Elbow Method” (cf. Ketchen 1993, Mooi 2011): 1. Hierarchical clustering first 2. Plot agglomeraKon coefficient against number of clusters 3. Search for “elbows”
2. Clustering method
4. Validate & Interpret C.
1. Clustering Variables
3. Number of Clusters
“Elbow” method
38
0.000
0.500
1.000
1.500
2.000
2.500
2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54 56 58 60 62 64 66 68 70 72 74 76 78 80 82 84 86 88 90 92 94 96 98
Clustering Coefficient (distance)
<29 7 16
2. Clustering method
4. Validate & Interpret C.
1. Clustering Variables
3. Number of Clusters
Number of cluster k
Statistical Measure: Silhouette
39
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
SilhoueVe Coefficient
2. Clustering method
4. Validate & Interpret C.
1. Clustering Variables
3. Number of Clusters
For datum i: Compares distance within its cluster to distance to nearest neigbouring cluster −1≤𝑠(𝑖)≤1
SilhoueVe Coefficient s(i)
Number of cluster k Rousseeuw, Peter J. (1987): SilhoueVes: A graphical aid to the interpretaKon and validaKon of cluster analysis. In Journal of Computa2onal and Applied Mathema2cs 20 (0).
The Clustering Process
40
0.335
-‐1 -‐0.5 0 0.5 1
SilhoueVe Value -‐0.40 -‐0.20 -‐ 0.20 0.40 0.60 0.80 1.00
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96
SilhoueVe
2. Clustering method
1. Clustering Variables
3. Number of Clusters
4. Validate & Interpret C.
good no cluster
top related