The Situation
Greg oversees billions of dollars in scientific research grants for the U.S. Office of Naval Research. Part of his job is going to Congress and justifying that spending, including the research that isn’t flashy, doesn’t make headlines, and sounds a little ridiculous out of context.
The clearest way to understand the problem turned out to be a story. Sarah mentioned a news headline she’d seen years earlier, something like “the Navy spends $1 million studying ketchup.” Greg corrected her: it was mayonnaise. It turns out mayonnaise has the same consistency as a highly volatile jet fuel the Navy uses. To figure out the safest and most efficient way to move large quantities of that fuel, and to measure everything that could go wrong along the way, researchers used mayonnaise as a stand-in. It saved time and money, and it produced a system the Navy still uses.
To a headline writer, it’s a mayonnaise study. To the Navy, it’s the reason jet refueling works. The problem Greg brought to Sarah’s MBA capstone was how to prove that at scale: how to show that research that looks boring or trivial is the foundation every impressive result stands on, and put a dollar figure on it.
What was built
Paper Trail, a working prototype that takes a single study or patent, traces the research it was built on, and adds up what that research cost, so the value of foundational science shows up as a number instead of an argument
- Start with one ID — enter the PubMed ID of the study or patent you want to investigate. That’s the only input.
- Automatic research lineage — the tool pulls every study the starting paper cites, then every study those papers cite, building a two-generation family tree of the research it depended on.
- A dollar figure for the lineage — using grant reporting data linked to each study, the tool estimates what it cost to fund the research that made the starting result possible.
- A map of where the money went — a geographic view of funding across a lineage, so reviewers can spot trends and see which regions and institutions are getting left out. –
- Built for people who aren’t data scientists — an interactive Dash dashboard instead of spreadsheets or raw database queries.
How a reviewer uses it:
- Look up the PubMed ID for the study or patent in question.
- Enter it into Paper Trail.
- The tool traces the cited research two generations back and pulls the grant data for each study.
- Review the estimated funding behind the lineage and the map of where that money went.
One study in, its whole funding history out. The “boring” research finally shows up on the balance sheet.
How It Was Done
Sarah built Paper Trail from concept to working MVP in Python, using Dash and Plotly for the interactive dashboard and map. This was 2024, when AI coding assistants were brand new, and she used them the way a good developer uses a senior colleague: to speed up the build while still writing and understanding the code herself.
We started with PubMed because it had the three things the idea needed: research records, well-documented citation links, and grant reporting numbers that could be tied back to funding. The full version would have added more databases, including ones internal to the military, for a much bigger picture.
Her fellow students acted as the customers, and she built in feedback from several people who do scientific research as part of their jobs, so the tool answered the questions researchers and funders actually ask rather than the ones that seemed interesting in a classroom.
The Outcome
Paper Trail proved the core idea worked: starting from one study, you could trace its research lineage, attach real grant costs to it, and see where the money went. The approach was novel enough that Sarah explored patent protection and was advised it was viable.
The next version would have run in the other direction too: starting from a foundational study and finding everything that cited it, then everything that cited those, to measure the research it went on to make possible. That’s the number that shows what a “mayonnaise study” is really worth.
The project stopped at the MVP stage. After the 2024 election, federal research funding priorities shifted sharply and much of the grant infrastructure the tool depended on was scaled back, so the decision was made not to take it further. The problem it solves hasn’t gone away. If anything, the case for showing what foundational research is worth is stronger now.
What Made this project work
- The problem had a story. The mayonnaise example turned an abstract funding question into something anyone could picture, and it shaped every design decision after that.
- It started with one well-documented data source. Building on PubMed first proved the concept without waiting for access to every database the full version would need.
- Real researchers shaped it. Feedback from people who do this work daily kept the tool focused on the questions funders and scientists actually ask.
Every organization has work that’s essential but invisible: the maintenance, the groundwork, the research nobody puts on a slide. When it’s the foundation everything else stands on, it’s worth finding a way to measure it, because what you can’t show, you eventually lose the budget for.