If you were responsible for deciding where to place hand sanitizer dispensers around a large university building, say, during a global pandemic, it may sound intuitive to put them near the doors. However, Professor Allen wanted to test it.

In the first study, titled “Design and Analysis of Facility Location Experiments Applied to Sanitizer Dispensers,” Allen and his collaborators explore the surprisingly complex problem of facility placement, meaning deciding not only how many facilities (such as dispensers, charging stations, restrooms or even restaurants) should exist, but exactly where they should go to best serve people.

Traditionally, such problems have been solved through intuition, convenience, or rough estimates. However, as Allen’s research points out, location affects demand. Where you put something changes how much people use it, and how they feel about it. This insight introduces what researchers call “demand endogeneity,” meaning that the decision itself (where to place something) actually influences the very thing being measured (how much people use it).

The researchers tested the placement of hand sanitizer dispensers at a university student union. The team placed two types of dispensers, open-faced and closed-faced, across 40 possible sites over multiple time periods. They recorded not just where people used them most, but how the configuration of dispensers influenced overall usage.

The results were both intuitive and counterintuitive. For instance, placing two dispensers together didn’t split demand as one might expect, it increased total usage by over two times. This “network effect,” as Allen calls it, shows that facilities can reinforce each other: seeing more sanitizers might make people more aware, more trusting, and more likely to use them.

Behind these findings lies a powerful mathematical engine. The researchers used what’s called a split-plot experimental design and a modified exchange algorithm, which are techniques that let computers assess experimental setups to find the most informative combinations. The result wasn’t just a better sanitizer plan, but a model that can apply to countless real-world systems, including bus stops, recycling bins, EV chargers, or anywhere placement and behavior interact.

If the sanitizer study was about where to put things, Allen’s second study is about what to say, and also where to say things on the map. Like dispensers of hand sanitizing solution, advertisements can be viewed as facilities that provide services located in space, i.e., a campaign might target just one county in a state. Then, people in other counties would not receive the service unless they saw the ad while traveling.

Titled “Using Optimization to Increase Vaccination Rates Through Educational Campaigns,” this study was conducted with FactSpread, a nonprofit dedicated to improving public understanding of scientific facts. In the shadow of the COVID-19 pandemic, Allen and colleagues faced a pressing question: how can data and optimization increase trust, not just efficiency?

Public health campaigns have long relied on intuition and mass messaging. But Allen and colleagues realized that the tools of optimization, such as clustering algorithms, experimental design, and causal impact analysis, could revolutionize how we communicate about health.

The researchers ran two major educational campaigns about COVID-19 vaccine efficacy, reaching over a million people across eight U.S. states and generating nearly four million ad impressions. Their process was both careful and ethical: before launching the full campaign, they conducted small-scale experiments to ensure their messages did not inadvertently harm any demographic group.

The first campaign tested different ad styles: one that explained the “death probability” for unvaccinated people, one that highlighted the Delta variant, and one that focused on basic vaccine facts. The ads were designed to be brief, factual, and visually clear, often inviting viewers to “fact-check us yourself.”

What made this approach special was its scientific backbone. Using clustering algorithms, the team grouped counties with similar demographics, such as similar education levels or percentages of people of color, to ensure fair comparisons. Then they used optimal experimental design to decide which clusters should receive which ads, ensuring the results would be statistically meaningful.

The data told a compelling story. The campaigns likely caused between 7,000 and 24,000 additional people to get vaccinated, and possibly influenced hundreds of thousands more through word-of-mouth effects. These estimates are still being evaluated by peer review but seem promising.

However, the study didn’t shy away from uncomfortable truths. One advertisement, the “Delta” ad, appeared to have negative effects among counties with higher percentages of people of color, even while helping others. Instead of ignoring this, the team used it as a learning moment: they redesigned their second campaign to focus on positive messages, avoiding potential harm.

This kind of ethical optimization, balancing effectiveness with inclusivity, is what sets the study apart. The team also used advanced statistical tools to estimate the real-world impact of their work. Even after accounting for confounding factors, the results held: data-driven messaging can move public behavior, even in a polarized and information-saturated world.

Allen’s study concludes with humility. The emergence of the Omicron variant, which reduced vaccine efficacy, made later campaigns less effective. Still, the methodology, combining optimization, empathy, and experimentation, proved remarkably powerful. It showed that science can learn not only what works, but for whom and why.

At first glance, hand sanitizers and vaccine ads seem to inhabit different universes. Yet both studies reflect the same worldview with resources located in space influencing decision-making. It’s the idea that decisions can be treated as experiments, and that the best decisions come from balancing data with human understanding.

In the sanitizer study, optimization helped people use hygiene tools more effectively, reducing waste and improving health. In the vaccination study, optimization helped people understand complex scientific information, building trust and saving lives.

Both projects took everyday systems, such as the layout of a building or the scrolling of a social media feed, and transformed them into laboratories for learning about human behavior.

In each case, Allen’s methods helped make invisible patterns visible: how people move, how they think, and how information spreads. The sanitizer project discovered the hidden “network effects” of spatial design; the vaccination project revealed the delicate social dynamics of persuasion.

So, what can the rest of us learn from Professor Allen’s research? First, every decision is an opportunity for experimentation. Whether you’re arranging furniture, designing a website, or launching a campaign, you can test, measure, and learn. Second, optimization isn’t just about numbers, it’s about people. The best algorithms don’t simply maximize clicks or minimize costs; they help us understand human needs and respond to them with care.

Moreover, small experiments can have big effects. Moving a sanitizer station or testing a single advertisement might seem trivial, but such micro-interventions can ripple through systems, shaping behaviors across communities and increasing impact.

Finally, ethics and empathy belong in every model. Allen’s vaccine study didn’t just aim to raise vaccination rates; it aimed to avoid harm, to respect differences, and to make knowledge more accessible.

Looking ahead, the implications of Allen’s research stretch far beyond universities or public health. Cities could use similar methods to plan bus routes or public parks. Companies could test environmental nudges to reduce energy use. Nonprofits could optimize donation campaigns to increase equity and trust.

In a world increasingly driven by data, Allen’s work reminds us that data alone isn’t enough. What matters is how we design the experiments that generate it, and how we interpret it with humility and purpose.