Funding Breakthrough Research


  • ARPA launched in 1958
  • DARPA budget ~$3B
  • ~100 program managers in darpa
  • ARPA case studies
    • Missile defense
    • Stealth technology
    • GPS
    • The internet
    • Autnomous navigation
    • Personal computer
    • The laser
    • Microsoft Windows
  • Other ARPA-modeled agencies are HSAPRA, IARPA, and ARPA-E
  • Hard to evaluate ARPA because it specifically shoots for high-risk high return outcomes
    • Patents and even commercialization don’t work
      While DARPA has been lauded for significant technological successes ex post (e.g. the internet, GPS, etc.), ARPA-E, a relatively young agency, has yet to be associated with similar outsized wins, making the challenge of intermediate evaluation more difficult.
      The ARPA model has four characteristics
  • The flexible nature of the organization
  • The identification of “Technological whitespace” and designing programs to fill the void
  • The freedom of program managers to choose their projects
  • The active management of each project using specific milestones and timelines
  • Darpa has changed focus over time in terms of timescale and domains
  • Sputnik inspired mission was “prevent technological surprise”
  • Focus on the best people since the 60’s
  • DARPA does take unsolicited proposals
  • Vietnam war shifted DARPA focus towards more near term and military applications - “direct and apparent relationship to a specific military function”
  • 1987-1992 DARPA has a focus on industrial competitiveness and dual-use technology
  • Over time different directors changed focus and tightness
  • Presidents affected who was director
  • Roland and Shiman empirically studied the Strategic Computing Initiative
  • Roland and Shiman point out that DARPA’s role was to get research into the limelight that would otherwise be neglected
  • Roland and Shiman point out that DARPA played a role as “connectors” in a research ecosystem
  • Colatat implies that DARPA pushes people to have novel collaborations
  • DARPA is isolated from normal operations
  • Independent program managers are key to DARPA success
  • Famous program managers
    • Kradeep Khosla
    • Zacharay Lemnos
    • Arati Prabhakar
    • Robert Taylor
  • Program managers are responsible for 2-4 of the four attributes of the program (see P2)
  • Fewer layers of beaurocracy (ie fewer people with veto power) enable DARPA to be able to flexibly respond
  • DARPA has 3 layers, ARPA-E has 2 layers
  • Focus on external research enables fewer layers
  • Flexibility on hiring both who and how much they’re paid - don’t have to go through normal federal hiring channels and restrictions
  • Fixed term for program managers pulls new ideas into the org and seeds the world with alumni
  • Flexibility on how they fund and pay for things
  • There is a level of pitching from the program
  • Coming up with the program can take up to 18 months of a 36 month contract
  • Programs are typically 3 years
  • DARPA program managers talk to both military stakeholders and scientists
  • They fly groups of scientists in to brainstorm together
  • DARPA program managers attempt to address bottlenecks in the innovation systems when designing programs
  • Program approval often requires specific technical targets which create concrete benchmarks
    It enables program managers to deliberately connect
    “You get communities together that don’t naturally talk and you give them some latitude and some life, and you push them forward and see what comes out of it”
  • Program managers build a goal focused community
  • Peer review has a bias against novelty
  • APRA model avoids the peer review bias against novelty by giving program directors discretion on who to fund
  • The gateway in the DARPA program is on the program idea itself and then it’s blackboxes
  • ARPA-E funding is under more scrutiny than DARPA because of politics - they have a slightly more rule-based system
  • Fund competing ways of doing the same thing
  • Program directors connect researchers who would never have met in the hopes that they will submit good proposals
  • The discretion of program managers to build teams and fund unconventional proposals is illustrated by the case study of Robert Taylor and ARPANET
  • Active project management is “best practice” in industry R&D
  • Active project management is created by having “real options.” These include:
    • Delay
    • Abandon
    • Contract
    • Expand
    • Switch
      (Huchzermeir and Loch)
  • Most grant-giving agencies give researchers independence once they have funds
    • counter Example: HHMI gives tons of independence which in fact gets better results than NIH researchers with tighter restrictions
  • ARPA model moves the creativity and independence from the researchers to the Program director/managers
  • 1960’s and 1990’s had less tight control and 1970’s and 2010’s had more tight control
  • Killing projects that are performing poorly statistically enables arpa to avoid throwing good money after bad and therefore take on riskier projects
  • You can’t just take one of the practices apart from the rest
  • The bundling of different practices is called coherence
  • The coherence of the ARPA model revolves around the empowered talented independent program managers.
  • ARPA models require
    • Clear mission
    • friction in the markets for ideas and technology
    • Nascent S curve research areas
  • Program design process needs to be able to start with the mission
  • The mission needs to be associated with quantifiable goals
  • Mowery argues that the single customer is important to DARPA
  • Defines applied research as performance improvement of existing technologies
  • Existing technologies = existing technological paradigm
  • ARPA-E awards have higher rates of publishing and patenting
  • Foster* introduced the concept of technology S-curves in 1986 where performance follows a sigmoidal function of research effort
  • ARPA model works best at the beginning of the S curve: the curve exists, but hasn’t hit stride
  • ARPA model does best in part of Edison’s Quadrant where the use is nascent
  • Claims private sector underinvests in R&D because firms cannot capture all the benefits of an innovation
  • Contrasts pharmeceutical and energy market in terms of existing pathways and not despite similar costs
  • It’s hard to measure the impact of ARPA model programs because the timescales are so long to actually transform an industry
  • Paying different research orgs is hard
  • Independence means that outcomes are strongly tied to talent
  • Program managers don’t have as much skin in the game
  • Program design requires researchers to share - need to build trust with them
  • It is hard to maintain culture with a high turnover rate
    • One culture mechanism is alumni community helping review proposals
  • Two axes of innovation management strategies
    • Idea Generation
    • Project Execution
  • Idea generation can either be investigator initiated or a mission-inspired solicitation
  • Project execution can be driven by Investigator freedom or empowered program staff
  • Examples of investigator initiation+investigator freedom
  • Examples of mission-inspired solicitation + investigator freedom
  • Examples of empowered program staff + Investigator initiation


  • Quadrants of Innovation management strategies seem to very much align with sources of funding - is there a fundamental reason why that is?
  • Are


  • Anna Goldstein
  • Michael Kearney
  • Pierre Azoulay
  • Erica Fuchs


    Dropbox - Funding Breakthrough Research.pdf - Simplify your life

Web URL for this note

Comment on this note