Upload
datahero
View
251
Download
1
Embed Size (px)
DESCRIPTION
Chris Neumann provides insight on his experience of starting DataHero and offers guidance on how others can start their own tech company.
Citation preview
BUILDING A SUCCESSFULDATA STARTUP
© 2014 Datahero, Inc.
Chris Neumann | @ckneumann
BUILDINGDATA COMPANIES
IS HARD© 2014 Datahero, Inc.
© 2014 Datahero, Inc.
STARTUPS BEGIN LIKE THIS
DATA STARTUPS ARE ABOUTTIMING
© 2014 Datahero, Inc.
IT’S ABOUT TIMESuccessful data startups recognize
fundamental shifts in technology or business processes
and create solutions that address new pain points
© 2014 Datahero, Inc.
IT’S ABOUT TIMING• The Data Warehouse market resulted from a need for
businesses to have a “single source of truth”
• The Big Data market resulted from the dramatic increase in data volumes from machine-generated data
• The Cloud BI market is emerging as a result of the decentralization of enterprise data as services move to the cloud
© 2014 Datahero, Inc.
BIG DATA
© 2014 Datahero, Inc.
BEFORE• Companies generally analyzed relatively
small volumes of (transactional) data
• Larger volumes of data could be stored in data warehouses, but calculations were relatively simple (aggregates, basic business metrics, etc.
© 2014 Datahero, Inc.
THE SHIFT• Machine-generated data, such as from web server logs,
provided far more granular information about customers
• Companies wanted to perform complex analysis on these larger volumes of data in order to better understand customer behavior
– There was also the potential for entirely new businesses built around data
© 2014 Datahero, Inc.
THE CHALLENGE• The volume of structured data companies wanted to
analyze was becoming larger than what could fit in a single server
The rate of growth of data now exceeded the rate of increase of storage density
• This shifted the performance bottleneck to the network
© 2014 Datahero, Inc.
THE SOLUTIONS• First Generation: Fix it with hardware!
– Make servers bigger (Teradata)
– Make the network faster (Netezza)
• Second Generation: Fix it with software!
(Aster Data, Greenplum, Vertica)
– Be smarter about where we store the data
– Be smarter about when we move the data
– Be smarter about how we move the data
© 2014 Datahero, Inc.
THE SOLUTIONSOf the five original “Big Data” startups:
4 were acquired in a span of lessthan a year for more than $2.5B total
The fifth was one of the acquirers
© 2014 Datahero, Inc.
CLOUD DATA
© 2014 Datahero, Inc.
BEFORE• The data business users wanted to analyze was
generated by on-premises software
• Centralized data stores (EDWs) were used to aggregate data from a small number of strategic sources
• BI teams would create reports for business users to access
© 2014 Datahero, Inc.
THE SHIFT• Over the past five years, business software is being
replaced with cloud services, the vast majority of which are departmental
• Company data is no longer stored primarily in on-premises systems, but is increasingly found in the cloud
• For the first time in more than 20 years, company data is becoming decentralized
© 2014 Datahero, Inc.
THE CHALLENGE• Companies now have a large number of remote data
sources each used by a small number of users
For the first time ever, business users have direct access to their data
• Users now have access to the data they want to work with, but don’t have the tools to take advantage of it
© 2014 Datahero, Inc.
THE SOLUTIONS• First Generation: Pull the data back!
– Custom integrations to pull cloud data down into on-premises data warehouses
– Put the data warehouse in the cloud…and then pull the rest of the data in (GoodData, Birst, RJMetrics)
• Second Generation: Leave the data where it is!
(DataHero, SumAll)
– Treat cloud business services as the systems of record
– Take advantage of existing security and permission models
– Eliminate the process bottleneck by empowering users to connect directly to the services they need
© 2014 Datahero, Inc.
DATA DECODER
• Advanced classification algorithms identify and normalize data types across services and files
• Semantic types such as URLs, Email Addresses and Lists extend traditional data types to provide added metadata
• Confidence intervals drive an intuitive feedback interface with users
CUSTOM CONNECTORS
• High-speed connectors built in collaboration with partners for optimal performance
• Robust, extensible framework supports rapid development of new connectors
• Secure integrations leverage partner security models for consistent data visibility
DATA DECODER
EXTENSIBLE CONNECTION FRAMEWORK
CONNECTOR
CONNECTOR
CONNECTOR
THE SOLUTIONS
THE SOLUTIONSINTUITIVE HTML5 INTERFACE
• Intuitive drag-and-drop interface created through user-centric design process involving thousands of hours of user testing and hundreds of users
“NO CODE” DATA COMBINATIONS
• Intuitive interface enables business users to combine (join) data across services and spreadsheets without coding or SQL
• Recommendation engine suggests common keys based on metadata derived by the Data Decoder
IDENTIFY THE SHIFT, THEN FIND THE PAIN
POINT
© 2014 Datahero, Inc.
BE THE PAINKILLER• Every major shift in technology and/or
business process results in new opportunities and new pain points
– Many of those pain points will have easy solutions
– A few will require fundamentally new approaches
© 2014 Datahero, Inc.
SO WHEREARE WE?
© 2014 Datahero, Inc.
© 2014 Datahero, Inc.
© 2014 Datahero, Inc.
© 2014 Datahero, Inc.
© 2014 Datahero, Inc.
@ckneumann