NIH Funding Success Rate by Topic: How to Compare Research Areas Honestly

Why "Success Rate by Topic" Is Usually the Wrong Question

A success rate is a fraction: applications funded divided by applications reviewed. To compute it for a topic — say, "cancer immunotherapy" or "gut microbiome" — you would need both numbers for that topic. Public data gives you only one of them. NIH RePORTER, the public database behind this site and every other grant-analytics tool, contains funded awards only. The applications that were reviewed and not funded — the denominator — are never published at the project level. They exist inside NIH's internal systems and in aggregate statistics, but you cannot count them by keyword.

This is not a limitation of any particular tool. It is a structural property of the data. Any website, consultant, or colleague who quotes you a precise success rate for a keyword-defined topic is either citing an internal NIH analysis (rare, and usually labeled as such) or making the number up. When you see a table that says "neuroscience: 19.8%, cancer: 22.3%," ask where the denominator came from. There usually isn't a good answer.

The good news: the question behind the question — "is my research area a reasonable place to build an NIH funding strategy?" — can be answered, just not with a single percentage. The rest of this guide covers what NIH actually publishes, what you can responsibly infer from award-side data, and how to combine the two.

Where Official Success Rates Actually Live

NIH does publish success rates — just not by topic. The official numbers are organized along two axes that NIH can actually count cleanly:

By institute or center (IC): Each of NIH's 27 institutes and centers reports how many competing applications it received and how many it funded in each fiscal year. NCI, NIAID, NIGMS, NIMH, and the rest each have their own published rates, and they differ meaningfully because each institute has its own budget and application volume.
By activity code (mechanism): Success rates are also broken out by grant type — R01-equivalents, R21, F32, K99, and so on. The spread across mechanisms within a single institute is often larger than the spread across institutes for the same mechanism.

These tables live in the NIH Data Book on RePORT (report.nih.gov), updated after each fiscal year closes. If you want a real number to anchor your expectations, that is the source — not a blog table, including this one. As a broad orientation, NIH-wide R01-equivalent success rates have hovered around one in five applications in recent fiscal years, but the institute-by-mechanism cell that matches your situation is the figure worth looking up.

Note what this structure implies: your effective odds depend far more on which institute reviews your application and which mechanism you choose than on the topical keywords in your title. A K99 application about tau aggregation and an R21 application about tau aggregation face different competitive pools, different review dynamics, and different funding decisions — same topic, different math.

The Denominator Problem

It is worth being precise about why topic-level rates resist computation, because the same logic protects you from several common analytical traps.

Three reasons the denominator is unknowable from public data

• Unfunded applications are confidential. NIH never publishes the titles, abstracts, or keywords of applications that were not funded. You can't count what you can't see.
• Topics don't map to administrative units. "Aging research" is funded by NIA, but also by NINDS, NHLBI, NIMH, and a dozen other institutes. Even NIH's own RCDC spending categories (which tag funded projects by topic) report dollars spent, not applications received.
• Keyword matching is fuzzy. A project can mention "machine learning" in one sentence of its abstract without being a machine-learning grant. Funded-side keyword counts are already noisy; an application-side count would be noisier still, if it existed.

The practical consequence: every topic-level analysis you build from RePORTER data — including everything on this site — is an analysis of what got funded, full stop. That is genuinely useful information. It just answers a different question than "what fraction of people like me succeeded?"

What You Can Infer from Award Data

Funded-side data supports three honest inferences, and all three are directly relevant to planning an application or a research program.

1. Award volume and trajectory

How many new awards mention your topic each year, and is that count rising or falling? A topic with a growing award count is one where study sections are seeing — and funding — more of that science. It says nothing about your individual odds, but it tells you whether the field has active, expanding NIH investment or is contracting. Our Trends tool charts exactly this from live RePORTER data: award counts and funding totals by keyword and year.

2. Institute mix

Which institutes fund your topic? This is arguably the most actionable inference available, because official success rates are published per institute. If your keyword's awards are split between two institutes, you can look up each institute's published success rate in the NIH Data Book and know which assignment gives your application a friendlier environment. Institute assignment is partly under your control — through the cover letter, the funding announcement you respond to, and how you frame the health relevance of your aims.

3. Mechanism mix

Is your topic funded mostly through R01s? Are there meaningful numbers of R21s, K awards, or fellowships? A topic dominated by large center grants and program projects is structurally harder for a junior investigator to enter than one with a healthy spread of small and early-career mechanisms. Looking at the activity codes attached to recent awards in your area — visible in PI Finder results and in RePORTER records — tells you which doors are actually open at your career stage.

A Worked Example: Comparing Two Research Areas

Suppose you work at the intersection of two fields and could credibly frame your next application in either direction. Here is the comparison workflow we would actually run — no invented numbers required, because you generate the real ones yourself from live data:

Chart both keywords side by side

Run both terms through Compare Topics. You are looking for direction and scale: is one area's new-award count flat while the other's has doubled over five years? Volume differences of that magnitude are signal; differences of a few percent are noise.

Identify the funding institutes for each framing

Open a sample of recent awards for each keyword and note which institutes made them. Two different framings of the same project can land at institutes with meaningfully different published success rates and budget trajectories.

Look up the official rates for those institute-mechanism cells

Go to the NIH Data Book and pull the success rate for the specific institute and activity code you would apply under. This replaces a fabricated topic rate with two real, official rates that bracket your actual situation.

Sanity-check who is winning

Use PI Finder to scan recently funded investigators in each area. If every recent award in one framing went to established labs with multiple R01s, while the other framing shows new and early-stage investigators winning, that tells you something the aggregate numbers never will.

The output of this exercise is not a single percentage. It is a defensible judgment: "Framing B puts me at an institute with a stronger published success rate, in a growing award pool, through a mechanism that early-career people in this area actually win." That judgment is worth far more than a fake decimal.

Five Common Misreadings

Award data rewards careful reading. These are the mistakes we see most often when researchers try to turn RePORTER-style data into strategy:

Treating award counts as success rates

A topic with many awards may simply attract many applications. High volume can coexist with brutal competition; small volume can coexist with generous odds in a niche study section. Counts measure activity, not difficulty.

Ignoring reporting lag

Awards appear in public data after the Notice of Award is issued, and a fiscal year's totals keep accumulating for months. A "decline" in the most recent year is often just data that hasn't landed yet. Compare completed fiscal years, and treat the current year as provisional.

Keyword ambiguity

"Stress" matches psychology, materials science, and cellular biology. "CRISPR" matches projects that use the tool incidentally and projects that develop it. Before drawing conclusions from a keyword trend, open a dozen of the matched awards and confirm they are actually the science you mean.

Confusing dollars with opportunity

Funding totals are dominated by a small number of very large awards — centers, cooperative agreements, clinical trial networks. A topic can show enormous dollar growth driven by three U54s that no individual investigator could have applied for. For personal strategy, new-award counts in investigator-initiated mechanisms matter more than total dollars.

Chasing "hot" topics on trend data alone

By the time a topic's award curve looks spectacular, the labs winning those awards built their preliminary data years earlier. Trend data is best used to position work you are already equipped to do, not to pick a field from scratch. We cover this in depth in our guide to positioning research with funding trends.

What Changed in 2026: No More Paylines

For decades, the companion question to "what's the success rate?" was "what's the payline?" — the percentile cutoff an institute used to fund applications more or less automatically. That era ended in January 2026, when NIH eliminated institute paylines under the Unified Funding Strategy. Funding decisions now weigh scientific priority, portfolio balance, and institutional and geographic diversity alongside percentile scores, rather than applying a hard numeric cutoff.

Two practical consequences for anyone reading success-rate data today. First, historical payline-by-institute tables — a staple of grant-advice blogs — are now obsolete as planning tools. Second, published success rates remain meaningful (they describe what actually happened), but the mapping from "your percentile" to "your funding probability" is fuzzier than it used to be. A strong score is still the main driver, but it is no longer the whole story. For the full picture, see our guide to the Unified Funding Strategy and the end of paylines.

Frequently Asked Questions

Where can I find the official NIH success rate for my institute and mechanism?

The NIH Data Book at report.nih.gov publishes success rates by institute and by activity code for each completed fiscal year. That is the authoritative source; everything else is derived from it or invented.

Does NIH publish anything topic-level at all?

Yes — the RCDC (Research, Condition, and Disease Categorization) system reports annual spending by category (for example, "Diabetes" or "Neurodegenerative"). These are dollars awarded, not success rates, and the categories are broad. They are useful for tracking NIH's investment priorities, not your personal odds.

Should I pick my research topic based on funding data?

No — pick your framing, mechanism, and target institute based on funding data. The science itself should come from your expertise and preliminary data. Funding data is a positioning tool: it helps you present work you can already do in the venue where it competes best.

If success rates are around one in five, should I expect to submit five applications?

Success rates are population averages, not personal probabilities — and they collapse resubmissions in ways that understate the odds of a persistent, improving applicant. Many funded projects succeeded on resubmission (A1). The realistic plan is not five different applications; it is one strong application with a resubmission strategy built in from day one.