close this bookVolume 1: No. 08
View the documentNews -- extended IJCAI workshop deadline
View the documentDiscussion -- changing the system, living in the system
View the documentQueries -- startup funding
View the documentClarifications -- Robert Siminoff; NSF deadlines
View the documentNews -- SBIR awards
View the documentOther software-related awards are
View the documentDiscussion -- competing for SBIR awards
View the documentDiscussion -- SBIR, the rest of the story

Eleven Federal agencies have SBIR programs. NSF's was the first, and the model for others. These programs serve two main purposes: to promote national interests by stimulating small business innovation, and to keep the applied proposals separate from regular research competitions -- for the benefit of both. (I speak for myself here, not for NSF or the Government.) NSF's program also helps program officers stay in contact with relevant industrial researchers. In theory, it keeps officers aware of the needs of small businesses and the importance of technology transfer.

At issue is whether industrial research is less worthy of support than academic research. It could be the same professor or researcher doing the work in each case, and the same rules for publication could be made to apply. Scientists argue that industry already has adequate sources of research capital. Business leaders argue that academic-only funding works against technology transfer and the creation of new industries. Congress sees it both ways, but is particularly sympathetic to the needs of small businesses. SBIR programs are a political solution, and NSF will be happy to run one as long as Congress provides the funds.

There is a politically correct way to discuss SBIR programs. The trick is to state only benefits of the program and to ignore compensating drawbacks. The press release for last year's winners (NSF PR 91-15), for instance, says that cumulative private investment and product sales attributable to NSF's SBIR program now exceed $1,000,000. That sounds good, but doesn't address the investment and sales that would have been achieved by these people without NSF support, nor NSF's costs and the good that could have been done at universities -- and hence for U.S. industry -- with the same money. Nor the good that would have been done by the same money in taxpayer's pockets.

I'm not saying that it's a bad program. Maybe the good news is that there's no bad news. I'm just saying that you won't ever find out by talking to a bureaucrat.

In practice, the proposals that NSF sees are not of the same research quality as academic proposals. Perhaps the winning proposals are of equivalent quality, although you can't tell because the proposal format and program goals are different. In terms of the advancement of knowledge, however, I would guess that the average SBIR proposal has little scientific value. These are applied or engineering proposals, not scientific investigations. In direct competition with university proposals, the SBIRs would not have a 1 in 7 chance.

All right, so what? My point is that we're fooling ourselves. Ask what NSF is doing for tech transfer and you will be given statistics from the SBIR program. Ask what NSF is doing for scientific instrumentation and you will be told that it was a special SBIR program emphasis in 1991. (Never mind that NSF's Engineering Directorate shut down its instrumentation program in 1990.) Ask whether these are really meritorious proposals worthy of NSF support and you will be assured that research quality was the number one criterion, with twice the weight of other factors.

The truth is -- correction, another viewpoint is -- that SBIR is a minor program evolved to deflect criticism so that NSF can get on with business as usual the rest of the year. The result may be that less attention is given to tech transfer than would be the case without SBIR.

The SBIR budget is a few million per year, about the size of any single-topic program. The overhead involves a half-dozen permanent staff -- e.g., Senich, Schoen, Smith, Coryell, Tibbetts, Gorman -- plus the [unbudgeted?] cooperation of program staff throughout NSF for several weeks each year. A relatively small number of outside panelists is also brought in.

Each year the program announcement goes out and the proposals come in. Identical research at identical (or greater!) funding levels could be proposed in any of the regular research programs, but program officers will try to steer non-academic proposals to the SBIR program instead. Ostensibly, this is because commercial proposals wouldn't stand a chance with academic reviewers, and because commercial potential is not a considered factor for regular research proposals. (As a taxpayer, I have to wonder WHY it is not a factor -- and why appropriate reviewers are not chosen from industry, and why NSF does not provide enough guidance to help PIs write competitive proposals. As a former program officer, I know that it's too much to expect one person to supervise both science and industry; industry should be nurtured by the Commerce Department. I also know that budgets are limited, and that training people to write good proposals would just increase the number of good proposals that must be declined.)

The SBIR solicitation is probably the most detailed, explicit, and legalistic of any that NSF prints. Only proposals directly addressing the solicited topics are accepted. (Not all NSF programs are included.) Literature surveys, market research, demonstration projects, or product development proposals will be rejected, regardless of any good that they might do. "For reasons of competitive fairness, contact with NSF regarding this solicitation is restricted during the proposal preparation period." Ordinarily, NSF sees no conflict between a helpful attitude and the public interest, so lack of budget and personnel -- and lack of domain expertise -- may be the real reasons for denying the help that academic applicants would get.

Proposals deviating in any way from the specified format and content are rejected. This is designed to reduce the time needed to scan, evaluate, and rank the proposals, and to avoid any need for special consideration or judgment. The process is made as mechanical as possible. (If the number of competing proposals is thereby reduced, few program officers would mind.) NSF is proud of the judgment exercised by program officers in other matters, but these proposals are to be dispatched as quickly as possible.

Proposals that survive format screening are passed to the programs in the specified topic areas. Program officers verify that research content and commercial potential are present. Reassignments are negotiated so that each proposal reaches an appropriate program, regardless of the topic the PI specified. Proposals are then grouped into panels, and program officers recruit panelists. NSF looks for a balance as to race, sex, occupation, and geographic region. The number of panelists is severely limited, so experienced generalists are sought. Budget is limited, so nearby people are often preferred. Industrial researchers are preferred, although few are in the NSF database or are known to most program officers. Panelists used in past years, or on other recent panels, are best avoided. Competitors of the proposing companies must be avoided. And the panelists have to be available and willing, which greatly narrows the selection. Still, NSF does makes sure that every proposal will be reviewed by at least one "advocate" panelist with some knowledge of the general problem domain. (One. Have I made my point?)

Panelists are brought in for a limited time to evaluate and rank a large number of proposals. The workload differs with the topic, but an average of ten or fifteen minutes per proposal for group consideration and writeup would be considered adequate. Program officers later add up the points, summarize the comments, tweak the rank orderings, add their own comments, and recommend that some proposals be further considered. Program officers are not spending their own money, so it's usually easy to recommend forward whatever percentage of proposals is suggested by the SBIR office -- even if scientific quality is lacking.

From there, my knowledge is hazy. The SBIR office convenes a panel of experienced business people to consider commercial potential of the recommended proposals. Then SBIR program officers have the final say. Political issues such as sex, race, and geographic diversity may have an influence. Balance among the various NSF divisions is considered. The enthusiasm of a program officer's recommendation is considered. Available funds are distributed as the program officers think best. 184 made it through last year, and the announcement for this year suggests about 200.

Winners celebrate, losers wonder whether to try again, and NSF turns its attention back to Science. To many, the SBIR process is an interruption, just one more darn thing that delays proposal processing and interferes with reading, writing, and recommending. The brief SBIR experience each year is not enough to focus officer's attention on industry or its research needs.

The bottom line? If you need startup money, apply. Don't be scared off if your business plan doesn't call for theoretical research of mainstream quality. Apply to other agencies as well. Make your best pitch in each proposal, and don't be subtle. If you are declined, accept the fact that the process is random and superficial; try again. If your research really is of mainstream quality, you should also apply in the regular research programs. And if you're a crusader looking to help U.S. competitiveness, give some thought to improving or replacing SBIR.