Rotational Focus Strategies

Focus strategies—Pomodoro, chunking, environmental shifts, checklists — are effective ADHD management tools. The problem is adherence. Strategies lose novelty over time, and novelty-seeking is a defining trait of ADHD. People with ADHD aren't failing to try; they're running into a structural mismatch between how static interventions are designed and how ADHD brains respond to repetition.

Kovacs et al. (2018) found that rotating online behavior-change interventions increased effectiveness even as it increased attrition—a tradeoff worth examining for ADHD specifically. Could a lightweight app that randomly assigns a different focus strategy each day improve perceived productivity and satisfaction for people with ADHD?

The study was conducted as a solo graduate course project in Designing Digital Health Systems (Spring 2025), covering everything from literature review and study design through engineering, data collection, and analysis.

Designing for this population meant the app had to be genuinely low-friction — forgiving of missed days, usable across devices, and unobtrusive enough not to get in the way of what it was measuring. Those constraints shaped every technical and design decision downstream.

The study drew on eight papers spanning ADHD prevalence and employment outcomes, novelty-seeking as a core ADHD trait, rotating behavior-change interventions, self-efficacy and productivity, and software engineers with ADHD. The Kovacs et al. finding was the theoretical anchor: rotation increases effectiveness, and the ADHD population's novelty-seeking disposition makes that effect worth testing directly.

Design constraints first, then stack

Before writing any code, I worked through what the app actually needed to be: platform-independent, low in daily demand, forgiving of forgetfulness, and simple enough not to confound what it was measuring.

Google Forms would have been faster to build. I ruled it out because randomized daily strategy assignment needs logic that static forms can't support, and I needed persistent user state across a multi-day study. A bad UX would have been a research problem, not just an engineering one.

I tested a paper prototype with two peers before committing to the build. The format couldn't simulate the multi-day structure, so findings were limited—but it confirmed the core interaction model and made clear that instructions needed to be more explicit about what was expected at each phase. With that settled, I finalized the stack: React, TypeScript, Next.js, PostgreSQL, Supabase, and Clerk Auth, deployed on Vercel. Once the stack decisions were made, implementation took about four days.

Piloting and iterating

Before the main study I ran a digital pilot with 3 participants over 3 days — 1 diagnosed with ADHD, 2 self-reporting suspected ADHD. The sample was small and the window was short, but the pilot surfaced real problems. Database double-submissions from rapid clicks got fixed by disabling the submit button after the first click. Users had no way to see what strategy they'd previously been assigned, so I added assignment history to the dashboard. Instructions were unclear about which survey to complete and when, which I addressed with phase-specific accordion sections. The exit survey was being skipped entirely because it looked identical to the daily survey, so I distinguished its appearance.

One participant wanted to select their own strategy each day rather than have one randomly assigned. After talking it through, it was clear they didn't have strong feelings about it—and allowing self-selection would have undermined the novelty mechanism the study was built around. Random assignment stayed.

Study design

The main study ran for 7 days per participant: 3 baseline days of end-of-day Likert surveys on satisfaction and productivity, followed by 4 days of randomly assigned focus strategy use, with an exit survey on the final day.

Recruitment was via convenience and snowball sampling—ADHD forums require IRB approval, so I reached out through personal networks. 7 participants enrolled; 3 completed more than 80% of surveys (P1, P3, P5) and were included in the primary analysis. Analysis combined Pearson's correlation on Likert responses with thematic analysis of open-ended daily and exit survey responses.

Quantitative findings

Among participants with more than 80% completion, mean satisfaction scores rose from 4.0 at baseline to 4.33 during focus strategy days. Mean productivity scores rose from 3.33 to 4.25.

Pearson's r for focus strategy use and daily productivity was 0.38, a weak positive correlation. For satisfaction, r = 0.079, which falls below the threshold for even a weak correlation by standard benchmarks. With only three participants in the primary analysis, neither result is conclusive. The directional signal is consistent, but this is a pilot study, not a proof.

Thematic findings

Thematic analysis of 12 open responses surfaced 6 themes. Breaking up tasks or the day to aid completion was the strongest (n=4)—participants described chunking and task-switching as concrete mechanisms behind their better days, not just general positivity about the app. One participant noted that rotating between strategies correlated with their best day of the study, which is about as direct a confirmation of the Kovacs et al. hypothesis as you can get from a single data point.

The more interesting finding came from the exit surveys. One participant wrote:

"Even if I was not employing the strategy at the particular time, reminding myself that I HAD a focus strategy for the day was helpful."

That suggests a self-efficacy effect that operates independently of whether the strategy is actually used—having a plan improved how people felt about their day even when the plan wasn't followed. A second participant said they wanted to keep journaling and monitoring their productivity after the study ended, which suggests the intervention produced some behavior change beyond the study window.

The biggest lesson from this project was about scope discipline. Building a full-stack web application was the right choice—a form-based alternative couldn't have supported randomized daily assignment, and offloading that randomization to the app meaningfully reduced participant burden (no rolling dice, no remembering which strategy to use). The flexibility to submit surveys late was equally important for an ADHD population where rigid daily deadlines would have increased dropout. The tradeoff was time: development consumed more than it should have, leaving less room for piloting and a shorter study window than I would have liked.

The dropout rate was anticipated—longitudinal studies with ADHD populations have known retention challenges—but 4 of 7 participants completing fewer than 80% of surveys still limited what could be concluded. A follow-up study with better retention mechanisms, whether push notifications, shorter surveys, or a buddy system, would be worth running.

What I'd preserve: the decision to keep random assignment despite a participant's preference to self-select. Accommodating that request would have undermined the mechanism the study was testing. Knowing when a design decision is a constraint worth holding is as much a research skill as an engineering one.