Hartmann Data Driven Business models presentation

Capturing Value from Big Data through
Data-Driven Business Models
Patterns from the Start-up world
Philipp Hartman,
Dr Mohamed Zaki and Prof Duncan McFarlane
Cambridge Service Alliance
University of Cambridge
“Data is the new oil”1
Data volume per year (Exabytes)2 56% 45000 40000 35000 Top Priority: “How to get va
lue from big data” 3
30000 25000 20000 3 Gartner “Big D
ata Study” 201
3 15000 10000 5000 0 2005 2010 2 IDC's Digital Universe Study, December 2012 1 various authors, e.g. Clive Humby 2015 2020 Two general areas can be identified where
big data creates value
How to get value from Big Data? OpLmizaLon of exisLng business1 New Business Models1 Chesbrough, Rosenbloom (2002): Business model to capture value from an innovaLon Crisculo (2012): New technologies oQen first commercialized through start-­‐up companies 1 cf. McKinsey 2011, IBM 2012, Davenport 2006, AT Kearney 2013, EMC 2012 3 Sub quesLons Research QuesLon Based on this motivation the research
question was developed
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? How to idenLfy pa^erns? Data-­‐driven business model framework Clustering 4 The research was done in five steps
Literature Review Build the framework How to analyse data-­‐
driven business models? Data collecLon & coding Finding Pa^erns Case studies How to idenLfy pa^erns? 5 The first step was a literature review with
three different topics
Build the framework Literature Review Data collecLon & coding Finding Pa^erns Case studies DefiniLon Big Data Literature Review Value CreaLon DefiniLon Business Model Related Work Business Model Frameworks Data driven business Models Cloud business models 6 The first step was a literature review with
three different topics
Build the framework Literature Review Data collecLon & coding Finding Pa^erns Case studies DefiniLon Big Data Literature Review Value CreaLon DefiniLon Business Model Related Work Business Model Frameworks Data driven business Models Cloud business models 7 Literature review: Business Model
Build the framework Literature Review Data collecLon & coding Finding Pa^erns Case studies DefiniLon Big Data Literature Review Value CreaLon DefiniLon Business Model Related Work Business Model Frameworks Data driven business Models Cloud business models 8 Business model key components were
synthesized from existing frameworks
-­‐  No universally accepted definiLon of the concept (Weill, Malone et al. 2011) -­‐  Most definiLons refer to value creaFon & value capturing Business Model Key Components Value CreaFon
Business Model DefiniLon Key Resources
Key AcLviLes
Value ProposiLon
-­‐ 
-­‐ 
-­‐ 
-­‐ 
-­‐ 
-­‐ 
-­‐ 
Chesbrough & Rosenbloom 2002 Hedman & Kaling 2003 Osterwalder 2004 Morris 2005 Johnson, Christensen et. al. 2008 Al-­‐Debei 2010 Burkhart 2012 Value Capturing
ExisLng Business Model Frameworks Customer Segment
Revenue Model
Cost structure
Only a few papers are available in this field
Build the framework Literature Review Data collecLon & coding Finding Pa^erns Case studies DefiniLon Big Data Literature Review Value CreaLon DefiniLon Business Model Related Work Business Model Frameworks Data driven business Models Cloud business models 10 The review was extend to cloud business
models
Build the framework Literature Review Data collecLon & coding Finding Pa^erns Case studies DefiniLon Big Data Literature Review Value CreaLon DefiniLon Business Model Related Work Business Model Frameworks Data driven business Models Cloud business models 11 The literature review identified several gaps
Build the framework Literature Review Data collecLon & coding Finding Pa^erns Case studies • 
Li^le academic research on big data and value creaLon – mostly whitepapers • 
Gap in literature: data-­‐driven business models • 
O^o, Aier (2013) interesLng paper but limited to specific industry > no generalizaLon possible • 
Similar research for cloud business models (cf. Labes, Erek et. Al. 2013) 12 The framework was build from literature
starting from the key components
Literature Review Build the framework Data collecLon & coding Case studies Data-­‐Driven Business Model Framework Business Model Key Components (Dimensions) Data Sources Finding Pa^erns Internal Features for data sources Data Sources External Data GeneraLon Data AcquisiLon Key AcLvity exisLng data Self-­‐
generated Data Acquired Data Customer provided Free available Crowdsourci
ng Tracking & Other Processing Key AcLvity Data-­‐Driven Business Model Offering Target Customer Features for each dimension Data-­‐Driven-­‐
Business Model predicLve VisualizaLon prescripLve Data Offering Revenue Model Revenue Model Specific cost advantage descripLve AnalyLcs DistribuLon Target Customer Specific cost advantage AggregaLon InformaLon/
Knowledge Non-­‐Data Product/
Service B2B B2C Asset Sale Lending/
RenLng/
Leasing Licensing Usage fee SubscripLon fee AdverLsing Open Data Social Media data Web Crawled Data Synthesizing the different sources leads to
the taxonomy
exisLng data Internal Data Sources Self-­‐generated Data Acquired Data External Customer provided Open Data Free available Social Media data Web Crawled Data 14 Dimension: Activities
Crowdsourcing Data GeneraLon Tracking & Other Data AcquisiLon Processing Key AcLvity AggregaLon descripLve AnalyLcs predicLve VisualizaLon prescripLve DistribuLon 15 Dimension: Offering
Data Offering InformaLon/
Knowledge Non-­‐Data Product/Service 16 Dimension: Revenue Model
Asset Sale Lending/RenLng/
Leasing Licensing Revenue Model Usage fee SubscripLon fee AdverLsing 17 Dimension: Target Customer
B2B Target Customer B2C 18 The final framework
Literature Review Build the framework Data collecLon & coding Finding Pa^erns Case studies exisLng data Internal Data Sources Self-­‐generated Data Acquired Data External Customer provided Open Data Free available Social Media data Crowdsourcing Web Crawled Data Data GeneraLon Tracking & Other Data AcquisiLon Processing Key AcLvity Data-­‐Driven-­‐
Business Model AggregaLon descripLve AnalyLcs predicLve VisualizaLon prescripLve DistribuLon Data Offering InformaLon/
Knowledge Non-­‐Data Product/Service B2B Target Customer B2C Asset Sale Lending/RenLng/
Leasing Licensing Revenue Model Specific cost advantage Usage fee SubscripLon fee AdverLsing 19 Data collection and coding
Literature Review Sampling Build the framework Data collecLon & coding Data collecLon Finding Pa^erns Case studies Data analysis 20 The data was generated using public
available sources
Sampling Tag: “big data” “big data analyLcs” 1329 companies Data collecLon Company informaLon •  Company websites •  Press releases Public sources Data analysis •  Coding of sources using data driven business model framework •  Nvivo Random sample cleaning 100 Companies 299 Sources ~3 sources/comp 100 binary feature vectors 21 Overall Analysis: Data Source
0% 10% 20% 30% 40% 50% 60% Acquired Data Customer&Partner-­‐provided Data Free available •  >50% of companies rely on free available data •  >50% of companies use data provided by customers/partners Crowd Sourced Tracked & Other Note: Sum > 100% as companies might use mulLple data sources 22 Overall Analysis: Key Activities
0% 10% 20% 30% 40% 50% 60% 70% 80% AggregaLon AnalyLcs DescripLve AnalyLcs PredicLve AnalyLcs PrescripLve AnalyLcs Data acquisLon Data generaLon •  >70% of companies use analyLcs -­‐ mostly descripLve Data processing DistribuLon VisualizaLon Note: Sum > 100% as some companies rely on mulLple revenue models 23 Overall Analysis: Revenue Model
0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% AdverLsing Asset Sales Brokerage Fees Lending RenLng Leasing Licensing SubscripLon fee Usage Fee •  Majority of revenue models based on subscripLon and/
or usage fee •  No informaLon about the revenue model as many companies are in an early stage No informaLon Note: Sum > 100% as some companies rely on mulLple revenue models 24 Overall Analysis: Target Customer
13% •  There seems to be a noteworthy predominance of B2B business models 17% 70% B2B B2C •  But no reference data found both 25 BM patterns were identified using a
clustering approach
Literature Review 1. Clustering Variables Build the framework Data collecLon & coding 2. Clustering method Finding Pa^erns 3. Number of Clusters Case studies 4. Validate & Interpret C. Ketchen, David J.; Shook, Christopher L. (1996): The ApplicaLon of Cluster Analysis in Strategic Managment Reserach: An Analysis and CriLque. 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 ValidaLon: Results and ImplicaLons for Applied Analyses. In Phipps Arabie, Lawrence J. Hubert, Geert de Soete (Eds.): Clustering and classificaLon. pp. 341–376. 26 7 Business Model Cluster were identified
1 2 3 4 5 6 7 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 AggregaLon 1 0 0 0 0 1 1 AnalyLcs 0 1 1 1 1 0 1 Data acquisLon 0 0 1 0 0 0 0 Data generaLon 0 0 0 1 0 0 1 Number of companies 17 28 5 16 14 6 14 Type A B -­‐ C D E F Key AcLvity Data Source Cluster 27 6 significant Business Model types were
identified
Type A: “Free Data Collector & Aggregator” Type B: “AnalyLcs-­‐as-­‐a-­‐Service” Type C: “Data generaLon & AnalyLcs” Type D: “Free Data Knowledge Discovery” Type E: “Data AggregaLon-­‐as-­‐a-­‐Service” Type F: “MulL-­‐Source data mashup and analysis” 28 Customer provided Type C Type E Type B Type F Free available Key Data Source Tracked & generated The 6 BM types are characterised by the key
activities and key data sources
Type A Type D AggregaLon AnalyLcs Key acFvity Data generaLon 29 Type D: “Free Data Knowledge Discovery”
Companies 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.  Trendspo^r Key Data Source Key AcLviLes -­‐  Free available -­‐  AnalyLcs 0 -­‐  Social Media -­‐  Open Data -­‐  Web Crawled 5 10 15 DescripLve PredicLve PrescripLve Revenue Model 0 2 4 Target Customer 6 8 SubscripLon Usage Fee AdverLsing Brokearge Fees No InformaLon B2B B2C Type D: Examples
“Using patent-­‐pending technology, Gild evaluates the work of millions of developers so companies using Gild’s talent acquisiLon tools know who’s good and can target the right candidates.” •  Key Data: Free available websites (GitHub, Google Codes) “ Our goal is to provide the most accurate and honest reviews possible by using the data consumers create. We listen to the conversaLons, analyze them and visualize them for consumers.” •  Key Data: Twi^er •  Key AcLviLes: AnalyLcs •  Key AcLviLes: AnalyLcs •  Revenue Model: Monthly subscripLon •  Target Customer: B2B •  Revenue Model: AdverLsing •  Target Customer: B2B (B2C) 31 The cases studies will be validated the
framework and the clustering
Literature Review Build the framework Data collecLon & coding Finding Pa^erns Case studies Purpose: 1.  Validate framework & clusters 4 case studies with companies from the sample such as 2.  Illustrate business model types through examples 3.  IdenLfy specific challenges 32 Summary
-­‐  Gap in literature idenLfied -­‐  RQ: 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? -­‐  5 (7) different business model pa^erns idenLfied -­‐  Data-­‐driven business model framework created to enable analysis -­‐  Pa^ern idenLficaLon through clustering -­‐  ValidaLon through Case-­‐Studies 33 Limitations & Outlook
LimitaLons •  Only 100 samples •  Only start up companies •  Bias of data source (AngelList) •  StaLsLcal significance of clustering result •  Only public available sources used •  No statement about success of a parLcular 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 organizaFons 3.  Develop methodology to judge (financial) performance of different business models 34 Appendix
35 Literature
Al-­‐Debei, Mutaz M.; Avison, David (2010): Developing a unified framework of the business model concept. In Eur J Inf Syst 19 (3), pp. 359–376. Burkhart, Thomas; Wolter, Stephan; Schief, Markus; Krumeich, Julian; Di ValenLn, ChrisLna; Werth, Dirk et al. (2012): A comprehensive approach towards the structural descripLon of business models. In : Proceedings of the InternaLonal Conference on Management of Emergent Digital EcoSystems. New York, NY, USA: ACM (MEDES ’12), pp. 88–102. Chesbrough, H.; Rosenbloom, R. (2002): The role of the business model in capturing value from innovaLon: evidence from Xerox CorporaLon's technology spin-­‐off companies. In Industrial and Corporate Change 11 (3), pp. 529–555. Criscuolo, Paola; Nicolaou, Nicos; Salter, Ammon (2012): The elixir (or burden) of youth? Exploring differences in innovaLon between start-­‐ups and established firms. In Research Policy 41 (2), pp. 319–333. 36 Literature
Davenport, Thomas H.; Harris, Jeanne G. (2007): CompeLng on analyLcs. The new science of winning. Boston, Mass: Harvard Business School Press. Everi^, Brian; Landau, Sabine; Leese, Morven (2011): Cluster analysis. 5th ed. London, New York: Arnold; Oxford University Press. Hagen, ChrisLan; Khan, Khalid; Ciobo, Marco; Miller, Jason; Wall, Dan; Evans, Hugo; Yadava, Ajay (2013): Big Data and the CreaLve DestrucLon of Today's Business Models. ATKearney. Han, Jiawei; Kamber, Micheline; Pei, Jian (2011): Data Mining. Concepts and Techniques. 3rd ed. Burlington: Elsevier Science. Hedman, Jonas; Kalling, Thomas (2003): The business model concept: theoreLcal underpinnings and empirical illustraLons. In European Journal of Informa=on Systems 12 (1), pp. 49–59. 37 Literature
Johnson, Mark W.; Christensen, Clayton M.; Kagermann, Henning (2008): ReinvenLng your business model. In Harvard Business Review 86 (12), pp. 57–68. Labes, SLne; Erek, Koray; Zarnekow, Ruediger (2013): Common Pa^erns of Cloud Business Models. In : AMCIS 2013 Proceedings. Manyika, James; Chui, Michael; Brown, Brad; Bughin, Jacques; Dobbs, Richard; Roxburgh, Charles; Hung Byres, Angela (2011): Big data: The next fronLer for innovaLon, compeLLon, and producLvity. McKinsey Global InsLtute. Morris, Michael; Schindehu^e, Minet; Allen, Jeffrey (2005): The entrepreneur's business model: toward a unified perspecLve. In Special Sec=on: The Nonprofit Marke=ng Landscape 58 (6), pp. 726–735. Negash, Solomon (2004): Business Intelligence. In Communica=ons of the Associa=on for Informa=on Systems 13, pp. 177–195. 38 Literature
Osterwalder, Alexander (2004): The Business Model Ontology. A ProposiLon in Design Science Research. These. Ecole des Hautes Etudes Commerciales de l’Université de Lausanne, Lausanne. O^o, Boris; Aier, Stephan (2013): Business Models in the Data Economy: A Case Study from the Business Partner Data Domain. With assistance of Rainer Alt, Bogdan Franczyk. In : Proceedings of the 11th InternaLonal Conference on WirtschaQsinformaLk (WI 2013) : The 11th InternaLonal Conference on WirtschaQsinformaLk (WI 2013), vol. 1. Leipzig, pp. 475–489. Pham, D. T.; Dimov, S. S.; Nguyen, C. D. (2005): SelecLon of K in K-­‐means clustering. In Proceedings of the Ins=tu=on of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science 219 (1), pp. 103–119. Pham, D. T.; Afify, A. A. (2007): Clustering techniques and their applicaLons in engineering. In Proceedings of the Ins=tu=on of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science 221 (11), pp. 1445–1459. 39 Literature
Rousseeuw, Peter J. (1987): Silhoue^es: A graphical aid to the interpretaLon and validaLon of cluster analysis. In Journal of Computa=onal and Applied Mathema=cs 20 (0), pp. 53–65. Schroeck, Michael; Shockley, Rebecca; Smart, Janet; Romero-­‐Morales, Dolores; Tufano, Peter (2012): AnalyLcs: The real-­‐world use of big data. How innovaLve enterprises extract value from uncertain data. IBM InsLtute for Business Value; Saïd Business School at the University of Oxford. Singh, Ranjit; Singh, Kawaljeet (2010): A DescripLve ClassificaLon of Causes of Data Quality Problems in Data Warehousing. In IJCSI Interna=onal Journal of Computer Science I 7 (3), pp. 41–50. Weill, Peter; Malone, Thomas W.; Apel, Thomas G. (2011): The business models investors prefer. In MIT Sloan Management Review 52 (4), p. 17. 40 The Clustering Process
1. Clustering Variables Variables relevant to determine clustering (Miligan 1996) 2. Clustering method 3. Number of Clusters 4. Validate & Interpret C. 2 Dimensions: “Data source” & “Key AcLvity” 9 variables #Variables has to match #samples (Mooi 2011) Avoid high correlaLon between variables (<0.9) (Mooi 2011) ~ 2m samples for m variables: 6-­‐7 variables max. correlaLon: 0,5 41 The Clustering Process
1. Clustering Variables 2. Clustering method ParLLoning Clustering Method (Han 2011) 3. Number of Clusters 4. Validate & Interpret C. K-­‐Medoids Hierarchical Density-­‐based Grid-­‐based Proximity Measure Include neg. match Euclidean Distance Exclude neg. match 42 There is no “one right solution” for the
number of clusters
1. Clustering Variables large to reflect specific differences 2. Clustering method 3. Number of Clusters k? 4. Validate & Interpret C. k << n Several different approaches (Pham 2005, Mooi 2011, Han 2011, Everi^ et. al. 2011): 1.  Use a-­‐priori knowledge to determine number of clusters 2.  Visual approaches 𝑘 ~√⁠​𝑛/2 →𝑘 ~ 7 3.  Rule of thumb (Han 2011): 4.  “Elbow” method 5.  StaLsLcal methods 43 “Elbow” method
1. Clustering Variables 2. Clustering method 3. Number of Clusters 4. Validate & Interpret C. “Elbow Method” (cf. Ketchen 1993, Mooi 2011): 1.  Hierarchical clustering first 2.  Plot agglomeraLon coefficient against number of clusters 3.  Search for “elbows” 44 “Elbow” method
1. Clustering Variables 2. Clustering method 3. Number of Clusters 4. Validate & Interpret C. Clustering Coefficient (distance) 2.500 2.000 1.500 1.000 0.500 0.000 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 7 16 <29 Number of cluster k 45 Statistical Measure: Silhouette
1. Clustering Variables Silhoue^e Coefficient s(i) For datum i: Compares distance within its cluster to distance to nearest neigbouring cluster −1≤𝑠(𝑖)≤1 2. Clustering method 3. Number of Clusters 4. Validate & Interpret C. Silhoue^e Coefficient 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Number of cluster k Rousseeuw, Peter J. (1987): Silhoue^es: A graphical aid to the interpretaLon and validaLon of cluster analysis. In Journal of Computa=onal and Applied Mathema=cs 20 (0). 46 The Clustering Process
2. Clustering method 1. Clustering Variables 3. Number of Clusters 4. Validate & Interpret C. Silhoue^e -­‐0.40 -­‐0.20 -­‐ 1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96 0.20 0.40 0.60 0.80 1.00 Silhoue^e Value no cluster good 0.335 -­‐1 -­‐0.5 0 0.5 1 47