Minimum Viable Bureaucracy

Why Agencies Should Look for the Least Restrictive Way of Doing Something

At the Good Science Project, we’ve published a number of pieces on reducing the administrative burden on researchers–most recently estimated at a staggering 44% of their research time. See “The Burden of Administrative Compliance,” “The Efforts to Reform Research Bureaucracy To Date,” “Here’s How We Could Get An Extra $2 Billion In Biomedical Research Per Year — For Free,” and “How To Actually Reduce The Administrative Burden On Research.”

There has been some effort in the past to ask federal agencies to use consistent processes, such as proposal formats. While a good idea in theory, I suspect this is mostly a waste of time so far.

For example, in 2021, the NIH announced that “in an effort to support strong collaboration between Federal research agencies, NIH has made every effort to align the Biographical Sketch (Biosketch), Other Support format page and Application Form Instructions with the guidance issued by the Office of Science and Technology Policy Joint Committee on the Research Environment.”

What did that mean? One main requirement was that NIH-funded researchers should list their positions and appointments in reverse chronological order rather than the former requirement (chronological order).

This change was not well received by researchers who (correctly) saw this as increasing burden, not reducing it.

Rather than wasting everyone’s time like this, we could look for a much more fruitful way to coordinate across the federal government:

“Minimum Viable Bureaucracy.”

That is, across the federal government, we should look for agencies or divisions that accomplish the same goal with the smallest number of procedural requirements and bureaucracy. Everyone else should then try (or be required) to take the same minimal approach. 

Where did this idea come from?

To take a step back, the idea of “Minimum Viable Bureaucracy” is obviously drawn from the “Minimum Viable Product” idea that became so popular in the startup world over a decade ago primarily due to Eric Ries’ book “The Lean Startup.”

To paint with a broad brush, the general idea is that rather spending years to build the “perfect” software product before releasing it (at which point it might be outdated, expensive, and far removed from customers’ needs), startups should try to build the minimum level of a product that would actually be viable in the market. Then, they should release it, get customer feedback, see what customers actually need and don’t need, and iterate. 

From there, folks started writing essays about how startups should also aim for the “minimum viable bureaucracy,” i.e., the leanest and most streamlined way of accomplishing goals like legal compliance, procurement, budgeting, meeting schedules, internal coordination, etc.

Applying this Silicon Valley concept to federal agencies . . .

Across federal science funding agencies, we should adopt the same idea: minimum viable bureaucracy. After all, we inevitably need some level of bureaucratic oversight and reporting (it’s not as if we’re going to hand out tens of billions in science funding and then just walk away). But we should actively look for who is doing something in the most streamlined way, and force everyone else to use the same format/procedure. 

Not just funding agencies, either. For example, IRBs (which are a particular source of time-wasting for many people) should have to do the same: figure out which IRB at which university has the most streamlined procedures that do a good job of protecting human subjects, and then use those procedures more widely.

Is there a good example?

Oh yes.  

Here’s a long post by a graduate student describing what it took to apply for graduate research fellowships from both NIH and NSF. As he points out, the NIH fellowship (called an F31) is “one of the smallest NIH grants, intended to support a single PHD student.” After he reviewed the 139-page instruction manual [!!], here’s what he had to produce. Apologies for the lengthy quotation, but it helps emphasize the point here: 

• Form F.200 - SF 424 (R&R), a general form about the type of grant requested and applicant information. Instructions for this form are on pages 18–30.

• Cover Letter for Form F.200 (1 page). Instructions are on pages 31–32.

• Form F.220, “Other Project Information”. Basically this is to list anything that may require additional review. For example, research:

  • on humans or vertebrate animals,

  • on biohazards or potentially dangerous substances,

  • with potential environmental impact, or

  • with foreign collaborators

This form also includes a 30-line project summary/abstract attachment, a list of bibliographic references, and a separate attachment describing the facilities and equipment used for the research. Instructions are on pages 33–42.

• Form F.230, “Project/Performance Site Location(s)”. This form is very similar to the attachment to F.220, but goes into more detail on the site location. Instructions are on pages 43–47.

• Form F.240, “R&R Senior/Key Person Profile (Expanded)”. Biographical information about the people involved. This will involve submitting Biosketches for me and my advisor. A Biosketch is basically a CV, except in a format that’s standardized by the NIH. It includes:

  • Name, title, eRA Commons username

  • Educational history

  • Personal statement

  • List of positions and honors

  • Description of five “Contributions to Science”

  • Research Support (basically, a list of previous grants)

  • Scholastic performance (basically, a list of courses and grades)

There is a 5-page limit on each Biosketch, so I’ll need to prioritize what I write about. Instructions for this form are on pages 48–59.

• Form F.430, “PHS Fellowship Supplemental Form”. This is the big one, where I’ll actually write about my proposed research. Instructions are on pages 60–83. It includes:

  • Applicant’s Background and Goals. 6 pages total, including:

    • Doctoral Dissertation and Research Experience

    • Training Goals and Objectives

    • Activities Planned Under this Award

  • Specific Aims (1 page). Basically, what are the goals of my research and why are they important?

  • Research Strategy (6 pages, not including references). What’s my plan for my research? This will address both the “Significance” and the “Approach”. I can also include the preliminary data I have from my experiments so far.

  • Respective Contributions (1 page). Basically, who is doing what.

  • Selection of sponsor and institution (1 page). Why did I choose the university and advisor that I have?

  • Training in the responsible conduct of research (1 page). Basically, I just need to document that I’ve taken a course in research ethics.

  • Sponsor and Co-Sponsor statements (6 pages). My advisor will (hopefully) write this section. It includes:

    • Research Support Available (a list of other grants that our lab has)

    • List of outcomes for previous PhDs and postdocs

    • The plan for my research training

    • Number of other PhD students currently in the lab

    • A recommendation letter about my qualifications

  • Description of Institutional Environment and Commitment to Training (2 pages). My department will write this section, describing why it’s a good place for my research.

  • Diversity F31 only — Description of Candidate’s Contribution to Program Goals. This form is for the university to explain why the applicant would “promote diversity”. It’s interesting that they’re very particular about this form being submitted on the university’s letterhead and signed by an institutional official. Anyway, I can ignore this one.

  • More sections on use of vertebrate animals or biohazards. I can skip these.

  • Resource Sharing Plan. A 1-paragraph description of how I plan to share the data and materials I generate.

  • Authentication of Key Biological and/or Chemical Resources (1 page). A description of how I plan to avoid wasting my time due to bad starting materials. (Apparently this is a common issue.)

  • Human Embryonic Stem Cells. If I’m using hESCs I’ll need to describe my plan. I’m using hiPSCs so I don’t need to do this.

  • Other data, including about degree sought, field of training, other support, and citizenship status.

  • Budget requested, including salary, “tuition and fees”, and other supplementary funding.

• If I were doing research on human subjects, there would be many other forms required. Thankfully, I’m not.

• Form F.600, “PHS Assignment Request Form”. This form is optional, but recommended. It is for requesting reviewers with a particular expertise. I could also request the exclusion of certain reviewers that might have conflicts of interest (although fortunately I don’t know any of these). Instructions are on pages 121–123.

This sounds like a lot. Mind-numbing, even. Is it all necessary? 

Not likely. 

As the graduate student points out, “Overall, there are 6 forms, and 37 pages of attachments. And this is for one of the smallest NIH grants. In contrast, my NSF fellowship application was a single 6-page form, plus a 3-page personal statement and a 2-page research proposal.”

Is there any evidence that NIH needs 6 forms and 37 pages of attachments where NSF gets by with a single main form and 5 pages of attachments? 

No. There is no evidence that NIH is better at selecting graduate research fellows than NSF, or that NIH graduate fellows produce better research.¹

So, with the “minimum viable bureaucracy” idea in hand, we would say the following: 

• Here are two agencies that are literally doing the same thing: handing out graduate research fellowships.

• NIH requires about 5 times more paperwork than NSF.

• Until and unless we see compelling evidence that 5x paperwork is better, NIH should be forced to adopt NSF’s application process. 

Then we can repeat the same inquiry across more types of grants and more federal funding agencies. This makes much more sense than trying to make all agencies do the same thing (even when it’s a bad idea that takes more of everyone’s time, not less).

If we can create more pressure to find (and move to) the minimum viable level of bureaucracy, we could save countless time and effort for federally-funded researchers. 

***

By the way, we can get a rough estimate for the possible efficiency gain as to graduate research fellowships.

First, the NIH publishes its success rates for all types of grants here. For the F31 research fellowship, we can see that in 2022, there were a total of 3,600 applications out of which 959 were granted. 

Second, the graduate student above reported spending around 65 hours on the F31 application package and said that anecdotally, “most people spend between 50 – 100 hours per F31 application.” Note that this doesn’t account for the advisor’s time in writing 6 pages on other research grants, a list of outcomes for previous students, a training plan, and a recommendation letter. Let’s say another 5 hours for that.

Third, if the NIH could cut the total burden to, say, 20 hours by using the NSF’s format instead, that would save a collective 180,000 hours–the same as 90 people working full-time for an entire year!

That’s what we could potentially save just for one tiny grant mechanism that is a small fraction of what NIH does.

1. I could imagine evidence as to bits and pieces of the application process. For example, the NIH requires one page on “Authentication of Key Biological and/or Chemical Resources,” which is an important issue related to the reproducibility of scientific research. It’s possible that due to having to write this statement, NIH graduate fellows do a somewhat better job of ensuring that their research materials are valid (compared to NSF graduate fellows). But that is by no means certain. One would need an empirical inquiry into the rate of irreproducible graduate research projects across NIH and NSF, the amount of time/effort at issue, the amount of time spent on the one-pager for several thousand applications per year, and so forth. I’ve never heard of any such empirical study.