This document focuses on writing proposals to NSF, but the following general advice can be applied to writing any proposal.
All proposals should answer the following questions in one form or another.
Read the program announcements before you talk to the program director so that your questions will be direct and specific. You can either call or send e-mail to the program director for your program area to discuss the ideas in your proposal. You can the list of telephone numbers and e-mail addresses from the NSF web site (http://www.nsf.gov/). Some program directors prefer e-mail; some prefer phone calls. Find out which program supports your research area (it's not always obvious); ask if there are other people you should talk to and what special initiatives might apply to you. Listen to what the program director says. Remember to say "thank you." (Don't be discouraged if they are rough on you. They spend all day on the phone and the rest of the time they're traveling and staying in government-rate hotels.) Treat the program directors as if they are intelligent people (even if you doubt it). The program director will assign the reviewers and will make the final decision. You don't have to be a sycophant, just be polite. (This advice comes from a former NSF program director.)
Most of your correspondance with NSF will be through email, but if you call, you will probably talk to a secretary several times before you get through to the program director. Be polite to the secretaries on the phone too -- They take a lot of grief from unhappy PIs (Principal Investigators). The instructions to proposers get more specific every year and FastLane gets better at rejecting proposals that don't meet the formatting requirements. However, You are still responsible for ensuring that your proposal meets all the particular program requirements. Follow the directions! (The secretaries are often heard muttering things like: "If they're so smart, why can't they read?")
. The NSF home page usually has a link to the latest version.
Clearly state the question you will address:
You can build your credentials in this section by summarizing other people's work clearly and concisely and by stating how your work uses their ideas and how it differs from theirs.
This should be equivalent to a PhD thesis proposal for the big leagues. Write to convince the best person in your field that your idea deserves funding. Simultaneously, you must convince someone who is very smart but has no background in your sub-area. The goal of your proposal is to persuade the reviewers that your ideas are so important that they will take money out of the taxpayers' pockets and hand it to you.
This the part that counts. WHAT will you do? Why is your strategy an appropriate one to pursue? HOW will you achieve your goals? Concisely, coherently, this section should complete the arguments developed earlier and present your initial pass on how to solve the problems posed. Avoid repetitions and digressions.
In general, NSF is more interested in ideas than in deliverables. The question is: What will we know when you're done that we don't know now? The question is not: What will we have that we don't have now? That is, rather than saying that you will develop a system that will do X, Y and Z, instead say why it is important to be able to do X, Y and Z; why X, Y and Z can't be done now; how you are going to go about making Z, Y and Z possible; and, by the way, you will demonstrate X, Y and Z in a system.
Right now, NSF is more open to application-oriented research. They need to show Congress that the money spent on research benefits the US economy. Some years ago, the word "applied" was a bad word at NSF. Now it's a good word. The pendulum between focussing on basic or applied research has about a 20 year periodicity. You always need to check to find out where it is at the moment. Check with the program director and knowledgeable colleagues.
In the last few years, NSF has started to take educational goals much more seriously. This section used to be boilerplate; it can't be any more. You need to think about what impact your research will have on education. Be specific but don't overstate.
Discuss expected results and your plan for evaluating the results. How will you measure progress?
Include a discussion of milestones and expected dates of completion. (Six months is the about the smallest time chunk you should include in an NSF proposal.) You are not committed to following this plan - but you must present a FEASIBLE plan to convince the reviewers that you know how to go about getting research results.
For new PIs, this is often the hardest section to write. You don't have to write the plan that you will follow no matter what. Think of it instead as presenting a possible path from where you are now to where you want to be at the end of the research. Give as much detail as you can. (You will always have at least one reviewer who is a stickler for details.)
Program directors often look in the bibliography for potential reviewers, and reviewers often look in the bibliography to see if their work is cited. If your bibliography has a lot of peripheral references, your proposal may be sent to reviewers whose work is not directly related to yours and who may not understand your proposal. On the other hand, if you do not cite the relevant literature, your proposal may be sent to reviewers who are not cited and who will criticize you for not knowing the literature. Most of the references in the bibliography will be cited in the Related Work section.
List up to five relevant publications, patents, copyrights, or software systems, plus up to five other significant publications.
Graduate students advised and postdocs sponsored in the past five years and total numbers advised & sponsored.
List long-term associates with whom you have collaborated in the past two years plus your graduate and postdoc advisors. This is for conflict-of-interest: NSF will not send your proposal to your close colleagues, your thesis advisor, nor to anyone at your current institution. You may list such people explicitly, if you wish.
Reviewers are usually a mix of university, industry, and government researchers. Almost always, the majority are academics.
Be sure to include all the support costs that you will need including computer services, travel, supplies. etc. NSF may cut your budget, but they'll never give you more than you ask for, so be sure to ask for everything you need.
Describe, justify, and estimate cost of equipment items $1000 or more. If your equipment needs change between the time you submit the proposal and the time it is granted, you can still buy what you need -- But be sure to talk to the university grants office BEFORE you buy the new equipment. There are special rules about equipment money because it is usually exempt from overhead charges.
Describe travel needs. If you are requesting funds for foreign travel, identify the country. All foreign travel on US government funds requires PRIOR APPROVAL. If a US airline flies your route, you must take a US airline. You cannot be reimbursed for foreign travel if the paper work is not processed before you leave. There is NO escape clause on this rule.
The business manager in your department or grants office will usually help you fill out the budget form once you have identified the direct costs.
If you have submitted the same proposal to more than one agency, be sure that you declare it on the cover page and in the current and pending support section. If you don't and the same reviewer is picked by both agencies, you won't get funded and your reputation will be damaged. Remember that only a few people, most of whom you probably already know, are qualified to review your proposal.
OPTIONAL: Special considerations if some work will occur off-campus
All proposals arrive at the NSF mail room. (99% arrive by express mail.) They are unpacked and given a preliminary sort based on the program announcement number or the NSF division given by the PI. (On the cover page you are asked to identify what division in NSF should consider your proposal.) A program director takes a look at the proposals that don't have an obvious place to go. Before the proposals go to the divisions, the information from the demographic sheet is entered into a database. Then the information from the cover sheet is entered into the main database and the proposal is assigned a number. (The two databases are not associated. No traceable data is kept with the demographic data.)
Next, the proposals are distributed to the divisions. Someone in each division checks to be sure that the proposals belong in that division and decides which program to assign them to. Sometimes, a proposal does not fit neatly into a single program or division. In that case, several program directors may be responsible for a single proposal. The original of the proposal is put in a jacket (a dark brown file folder) and data forms are filed with it.
Once the proposal has been assigned to a program director, it is ready for review. There are two basic review mechanisms used at NSF: mail review and panel review. Both are single blind peer review mechanisms: that is, the reviewers (who are the PI's peers) know who the PI is, but the PI does not know who the reviewers are.
Mail review: The program director chooses about six reviewers who are knowledgeable in the area of the proposed work. These reviewers are a mix of academics, industry and government reviewers, with academics being the majority. NSF has a database of reviewers and the proposals are assigned to the reviewers in the database. (The PI database, proposal database and reviewer database are all separate databases.) The proposal is mailed to the reviewers with a cover letter, a review form, and background information about the division to which the proposal was submitted. The reviewer is given about two weeks to a month to review the proposal. (The reminder letters for late reviews are generated automatically along with the labels, original cover letters, review forms, etc.) When four or five reviews have been received, the program director reviews the proposal and the reviews, then makes a decision to fund or decline the proposal. The program directors are supposed to exercise judgement. For example, a reviewer might appear to be a perfect match for a proposal -- but when the review comes in, it may be obvious that the PI's work conflicts with the reviewers work, and the reviewer is biased.
Panel review: If a program has a large volume of proposals, mail review is not possible. Here's the math: Most reviewers will not review more than 5 proposals a year without revolting (reviewing a proposal is a lot of work). If 150 proposals are submitted to a program, that means 900 review request must be sent out. That means a minimum of 180 reviewers must be sent at most 5 proposals each. Three reviews per person per year is more realistic - so that means the program director must have access to 300 of the proposal writers' peers in order to get the peer review system to work. And that's just for one program. All the other program directors are working with the same numbers -- and the expertise of many reviewers overlaps several programs. So instead of using mail review, program directors form panels. A program director will convene a panel of 10 to 15 experts in a field and have them review a set of related proposals. The panelists receive the proposals ahead of time and then come together to discuss which proposals should get funded. Most reviewers find it easier to rank a set of proposals than to write a detailed review of each proposal. The reviews from a panel review are often not as detailed as the ones from a mail review -- but they usually are more directed. If one reviewer completely misses the point of a proposal (which they sometimes do), this will come out during the panel discussion so you get fewer out-in-left-field reviews from panels than from mail review. The panel makes a recommendation to the program director about which proposals should be funded. Again, the program director makes the funding decision.
Often the decision to fund involves deciding whether to fund the proposal at the full or reduced amount. The program director makes the decision based on the program budget, the proposals that have been funded, and the pending proposals. The program director completes a form to support the decision. The proposal goes to the division director who must concur with the decision for it to be official. The decision is entered in the computer. If the proposal is declined, a letter is sent to the PI (cc:d to the grants office), and the file copy of the proposal with its associated reviews and forms is sent to storage. If the proposal is funded, the jacket is sent to the NSF grants office. The grants office deals with all the paper work required to make a grant. (A grant from NSF goes to the institution, not to the PI. But if you change institutions, it is easy to take an NSF grant with you.) After the paper work has been completed, the jacket is returned to the program director. The jacket stays with the program director until the grant is completed, and you have filled out your 98A (final report). Once the final report is received, your jacket is sent off to storage.
Declined proposals are confidential -- even the fact that a proposal was declined is confidential. For grants, the titles, abstracts, PIs, funding amounts, .. are public information, but the proposal itself is confidential.
The FastLane system http://www.fastlane.nsf.gov is an interactive real-time system used to conduct NSF business over the Internet. Almost all programs now require that proposals be submitted through FastLane. The grants office at your insitution can set up an account for you so that you can submit proposals and check their status through FastLane. If you are asked to write a review or be on a panel, the program officer will give you an id and password to give you access to the proposals.