Managing Innovationat Microsoft Research
Roy LevinApril 8, 2010
Outline
• The “what” and “why” of computing science research
• Microsoft Research: “why” and “how”
• Some research successes
http://research.microsoft.com2
What is (Computing) Research?
• “Basic” vs. “Applied”?• “Relevant” vs. “Blue-sky”?• “Short-term” vs. “Long-term”?• “Practical” vs. “Theoretical”?
There’s no simple definition!
http://research.microsoft.com3
Research: Reward/Risk
• Researchers (and their management) must answer these questions:– How likely is it to succeed? [Risk] – If it does, will it have value for my organization?
[Reward] • How?• How much?• When?
http://research.microsoft.com4
University Research
http://research.microsoft.com
Focus: Broad, government-supported, public-domain Determined by faculty/funding agency interest Education vehicle for students (perpetuate system)
Success metric (reward): Publications Faculty reputation (tenure track decision)
Needs in order to succeed: Funding agency approval
5
Small company research
http://research.microsoft.com
Focus: Short-term; bounded risk. Advanced development
Success metric (reward): Artifacts transferred to product organizations
Needs in order to succeed: Medium-term management support Close co-operation with receiving organizations
6
Big company research
http://research.microsoft.com
Focus: Long-term; varying breadth. Riskier than small company research. Costlier than university research.
Success metric (reward): Enhance existing businesses; create new ones.
Needs in order to succeed: Highly creative people Long-term management support Organizational stability
7
Challenges for Research (big company)
• Focus– long-term but relevant
• Payoff– big gains come infrequently and unpredictably
• IP: a two-edged sword– protective but can induce isolation
• Management commitment in hard times
http://research.microsoft.com8
Challenges Managing Research
• Staying ahead (keeping enough risk)– the comfort zone– the competition
• Technology transfer (getting reward)– hazards are well known (Christensen; Moore)– eternal vigilance and creativity
• Metrics– Patents? Publications? Profit?
http://research.microsoft.com9
Outline
• The “what” and “why” of computing science research
• Microsoft Research: “why” and “how”
• Some research successes
http://research.microsoft.com10
MSR Labs at a Glance
Lab Location Founded Researchers
Redmond 1991 250
Cambridge (UK) 1998 125
Asia (Beijing) 1999 220
Silicon Valley 2001 70
India (Bangalore) 2005 50
New England 2008 10
http://research.microsoft.com
Omits other research-related groups totaling about 400 people and over 1000 interns.
11
Where We Sit
http://research.microsoft.com
Steve BallmerCEO
Steven SinofskyWindows and Windows Live
Division
Bob MugliaServer and Tools
DIvision
Stephen ElopMS Business Division
Robbie BachEntertainment
and Devices Division
Qi LuOnline Services
Division
Craig MundieChief Research
and Strategy Officer
Rick RashidSVP MSR
Sales, marketing, and corporate functions (HR, Finance, Legal, etc.) are omitted.
12
Research Areas
• Broad spectrum, 50+ areas (see web site)– speech recognition, user interface research, programming tools
and methodologies, distributed systems and networking, graphics, natural language processing, robotics, machine learning, databases, search , web interaction, search intent, security, information retrieval, …
• Driven by technology, not specific business needs– long-term and uncertain relevance, e.g., sensor nets, quantum
computing, computing theory
http://research.microsoft.com13
Research Areas
• Broad spectrum, 50+ areas (see web site)– speech recognition, user interface research, programming tools and
methodologies, distributed systems and networking, graphics, natural language processing, robotics, machine learning, databases, search, web interaction, search intent, information retrieval, …
• Driven by technology, not specific business needs– short-term targets of opportunity, e.g., file systems for flash memory– long-term and uncertain relevance, e.g., sensor nets, quantum
computing, computing theory
Our Mission (unchanged since 1991)
• Advance the state of the art.
• Bring advances quickly to Microsoft products and services.
• Ensure Microsoft products and services have a future.
http://research.microsoft.com15
Why World-Wide?
• Talent availability
• University connections
• Geographically flavored work– natural language processing (Asia, Redmond)– networking (Asia, India)
• The next billion users
http://research.microsoft.com16
Microsoft Research Norms
• Bottom-up– researchers create projects, not management
• Collaborative– within and across groups and labs, and externally
• Flat management structure– as much as possible, given lab sizes
• Open– most work presented publicly
• IP-based– patent protection routinely sought
• Publish “at the right time”
http://research.microsoft.com17
Relationship to MS Businesses
• Historically, technology transfer is the research’s toughest problem.
• MSR-PM (program management)– The “connector-facilitators”
• A contact sport– geography can pose challenges– development in Redmond, SVC, Beijing, Hyderabad
• Tech Fest• Building on past success
– Most MS products affected• Incubation• IP Licensing
http://research.microsoft.com18
More on Research Management
http://research.microsoft.com
Paper:
A Perspective on Computing Research Management
Available at http://research.microsoft.com/users/roylevin
19
Outline
• The “what” and “why” of computing science research
• Microsoft Research: “why” and “how”
• Some research successes
http://research.microsoft.com20
Selected Technology Transfers
• Natural language processing– Office help system– Knowledge base automated
translation• Graphics
– Windows Media– DirectX/Direct3D– Numerous effect technologies
(Xbox)• Web search
– MSN core engine– Relevance ranking– Spam reduction– Intention/Tasks– Maps (routing)
• Large-scale spatial databases– Virtual Earth (Bing Maps)
• Machine learning– Drivatar (Forza Motorsport)– Filters in Outlook/Exchange (spam
reduction)• Software development tools
– PREfix/PREfast (find security holes)– Static driver verifier
• New user interface paradigms– Microsoft Surface
• Data centers & cloud computing– Dryad/DryadLINQ massively data-parallel
application paradigm– Storage infrastructure for Hotmail
http://research.microsoft.com21
http://research.microsoft.com
Focus Areas for MSR Silicon Valley
• Distributed Systems – e.g., Dryad
• Security and Privacy – e.g., differential privacy
• Web Search – e.g., Web spam
• Computer System Architecture – e.g., Beehive
• Computing Theory – e.g., shortest path routes
22
http://research.microsoft.com