This presentation demonstrates a study which provides a series of implications that may be particularly helpful to companies already leveraging ‘big data’ for their businesses or planning to do so. The Data Driven Business Model (DDBM) framework represents a basis for the analysis and clustering of business models. For practitioners the dimensions and various features may provide guidance on possibilities to form a business model for their specific venture. The framework allows identification and assessment of available potential data sources that can be used in a new DDBM. It also provides comprehensive sets of potential key activities as well as revenue models.The identified business model types can serve as both inspiration and blueprint for companies considering creating new data-driven business models. Although the focus of this paper was on business models in the start-up world, the key findings presumably also apply to established organisations to a large extent. The DDBM can potentially be used and tested by established organisations across different sectors in future research.

1. 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

2. “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

3. How to get value from Big Data?
3
OpKmizaKon
of
exisKng
service
Data
Driven
Business
Models1

4. 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
quesKons
Data-­‐driven
business
model
framework
How
to
idenKfy
paVerns?
Research
QuesKon
Clustering

5. 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?

6. 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

7. 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
CapturingValue
Crea@on
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

8. 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

9. 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
Crowdsourci
ng
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

10. 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

11. Dimension: Activities
11
Key
AcKvity
Data
GeneraKon
Crowdsourcing
Tracking
&
Other
Data
AcquisiKon
Processing
AggregaKon
AnalyKcs
descripKve
predicKve
prescripKve
VisualizaKon
DistribuKon

12. Dimension: Offering
12
Offering
Data
InformaKon/
Knowledge
Non-­‐Data
Product/Service

13. Dimension: Revenue Model
13
Revenue
Model
Asset
Sale
Lending/RenKng/
Leasing
Licensing
Usage
fee
SubscripKon
fee
AdverKsing

14. Dimension: Target Customer
14
Target
Customer
B2B
B2C

15. 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

16. Data collection and coding
16
Case
studies
Finding
PaVerns
Build
the
framework
Literature
Review
Data
collecKon
&
coding
Data
collecKon
Data
analysis
Sampling

17. 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

18. 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

19. 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

20. 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

21. 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

22. 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.

23. 7 Business Model Cluster were identified
23
Cluster
1
2
3
4
5
6
7
Data
Source
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

24. 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”

25. 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
Customer
provided
Tracked
&
generated
Key
ac@vity
Key
Data
Source

26. 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

27. 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)

28. 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

29. 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.

30. 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

31. Further Reading
31
hVp://www.cambridgeservicealliance.org/uploads/downloadfiles/
2014_March_Data%20Driven%20Business%20Models.pdf

32. 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

33. Appendix
33

34. 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

35. 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

36. 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):

37. “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

38. “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

39. 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).

40. 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