Tips for Authoring Grant Proposals
Mark D. Hill
Computer Sciences Department
University of Wisconsin-Madison
Writing grant proposals is hard. They are harder than papers because
they are part fiction. Fiction tends to come unnaturally to scientists and
engineers who may be more comfortable dealing with known facts.
Ph.D. students encounter a similar challenge when proposing their Ph.D. work.
Below I present:
Both are based on over twenty years of writing and evaluating grant proposals.
They offer a starting point for developing your thoughts.
Most successful grant proposals do not answer the questions explicitly,
but rather weave a convincing story.
- a checklist of questions good research proposals tend to answer and
- a generic 1+15-page grant outline.
Take-away point: Focus first on the problem
you seek to solve and why it is interesting.
Only when properly prepared, will readers and evaluators
be motivated to care about your potential solutions.
Are you tackling an important problem? If you can make progress
on it, will anyone care?
Why now? If this problem is so important, why has it not been addressed
Do you have concrete ideas for starting an attack on the problem and
a vision for proceeding further?
Is initial progress likely and subsequent progress possible?
Do you have some preliminary results? Do you demonstrate a good
understanding of the problem and the methods needed attack it further?
Do you have sensible plans and methods
(e.g., concrete steps and ways of decoupling risks)?
Why you? Why are your qualifications and infrastructure appropriate?
Have you followed the rules of the solicitation (e.g.,
compelling broader impacts for NSF)?
I find the above criteria valuable for both writing and evaluating grant
Most NSF grants ask for the research to be described in a 1-page summary,
followed by a 15-page description, not counting references.
A rough outline
is as follows, but feel free to expand and contract sections to
fit your circumstances.
- Summary (1 page):
A one-page abstract that touches on all seven criteria.
- Introduction (2 pages):
Discuss the problem, addressing criteria CARE and NOW
and forecasting answers to other criteria.
- Preliminary Work (3 pages):
Show some progress on understanding and addressing the problem,
e.g., half of a good paper, to address the criterion
- Research Directions (4 pages):
Show vision with concrete steps for middles parts of the grant (after the
preliminary work) and
promising directions for the later years, addressing criterion IDEAS.
- Research Methods/Plan (2 pages):
Show you know what methods you need, steps you may follow, and why you
can do it, addressing criteria PLAN and CAN-DO.
- Related Work (2 pages):
Discuss the related work of others if this is not already integrated
- Own Prior Work (1 page):
Discuss your prior work, addressing the criterion CAN-DO.
- Broader/Educational Impacts (1 page):
Discuss broader and education impacts, etc.,
addressing the criterion LEGAL.
Read and consider citing articles on these matters.
- Conclusion (0 pages):
Finish with a short, one-paragraph conclusion,
as this is too late to say something important.
- References (extra pages):
Show that you are familiar with the state-of-the-art,
addressing the criterion CAN-DO.
Especially when authoring your first grant proposals, ask colleagues to
privately share with you examples of successful proposals.
Read and initially emulate their successful style;
later develop your own approach. Also, liberally use figures, diagrams,
and graphs to promote understanding
even from those who only skim your proposal.
It is valuable have your proposal read by researchers who are
representative of the review panel
that will evaluate your proposal. (This step has the added
benefit of forcing you to do more work early.) One can classify
- Someone within the field who is very knowledgeable of the area
of your proposal (e.g., an expert in microarchitecture reviews a
proposal on future branch predictors).
- Someone within your field whose research area is somewhat
distant from yours (e.g., an expert in microarchitecture reviews
a proposal on memory consistency models).
- Someone who does research in another area of computer engineering
or computer science (e.g., CAD, VLSI, computer networks).
In my experience, getting comments from the last type of panelist
is most important, because we naturally tend to bias too far toward our
narrow specialties. One might consider the last group
infinitely wise, yet perhaps not too familiar with your area.
See also the earlier and more famous
Acknowledges and Disclaimer
Thank you to
and Milo Martin
for feedback and suggestions regarding this page.
Thanks to José Martínez for suggesting
and contributing to the "Get Feedback" section.
Thanks to David Wood with whom I have written many proposals.
The views represented above are mine alone and do not necessarily represent
the views of those who have funded me, principally the US National Science
Please let me know what you think of this advice, as it is sometime
hard to know what is and is not obvious.