Initiative 2 · methodology and requirements

Methodology & requirements
The scaled study, end to end

This document is the official methodology and requirements for the scaled student-journey usability study, prepared for the Maricopa ARC steering committee. It explains the persona-per-test model, the scenario bank and how it reaches roughly 5,000 task-runs, the split between human and AI testers, the operations, tracking, and coverage dashboard that keep it coordinated, the no-budget tool stack, and a clear statement of what we need to proceed. It is written to the domain's guardrails: no PII, translate-don't-standardize, an AI-only scope, and human contact as the metric of success.

36research personas
46base tasks
500target scenarios
5,000target task-runs
10colleges

The problem this study solves Why scale, and why now

Across the ten Maricopa colleges, the support that would keep students enrolled already exists on every campus. The failure is findability: students cannot reach advising, financial aid, basic needs, accommodations, or the right person at the moment they need them. A study with three personas walking thirty journeys describes the problem. It cannot map it at the resolution the committee needs to prioritize where AI should close the gaps. This document scales the diagnosis into a structured, repeatable instrument that produces a college-by-college, stage-by-stage map of exactly where students hit barriers, with severity attached.

The persona-per-test model One persona, one campus, one task

The unit of the study is a single run: one persona, at one campus, attempting one task. The tester, human or AI, opens the instrument and receives a brief: this is who you are, these are your barriers, this is the one thing you are trying to do, and this is your campus. Then they attempt it from a real felt need. They are never told where to go. Problems, not destinations: the persona must find the path, the way a real student would, usually starting from a search on the college website.

Why single-task

A single task isolates one barrier, so a failure is attributable. It also makes runs short, parallelizable, and easy to assign to many testers at once, which is what scale requires.

Why a persona library

A page that works for a digitally fluent transfer student can still fail a phone-first, Spanish-dominant parent on a slow connection. Only a varied library of 36 personas surfaces those intersections. The library reflects who actually enrolls: ESL and non-native speakers, veterans, autistic and ADHD students, food-insecure and first-generation students, working parents, returning adults, students who need Disability Resources, swirling multi-campus students, and dual-enrollment high-schoolers.

Why problems, not destinations

Naming the office tests nothing. The whole barrier is that students cannot find the office. By giving a goal and withholding the destination, the run measures findability directly: the path taken, the dead ends, whether they needed a human, and where they finally landed.

The scenario bank and the path to 5,000 runs Base tasks × personas × ten colleges

A taxonomy of 46 base tasks covers the whole journey, from apply and admissions, MEID, residency, orientation, self-enrollment, and the student ID, through financial aid, scholarships, the bookstore, adding and dropping classes and the aid impact, transcripts, tutoring, Disability Resources, counseling, basic needs, campus tech, veterans benefits, transfer, swirling, and graduation offboarding. Each base task is run by several personas, because different barriers reveal different failures, which lifts the bank toward roughly 500 distinct persona-and-task scenarios. Each scenario is then attempted at each of the ten colleges, where the local name and page layout differ. Five hundred scenarios across ten colleges is about 5,000 task-runs. The bank is the data backbone, delivered as a spreadsheet with a row per scenario and fields for status, tester, date, result, and severity.

Human and AI split. The plan assigns roughly 250 scenarios to human testers and 250 to AI testers, each set spread across the ten colleges, for about 2,500 task-runs each. AI testers walk the public, no-login paths at scale and produce a first map fast. Human testers cover the login-walled flows that need an authenticated test account, and they validate what the AI surfaced. The two halves reinforce each other: the AI finds breadth, the humans confirm depth and catch what an agent cannot feel.

Some scenarios cannot be tested until the district's Salesforce advising tool exists: find or get assigned an advisor, book an advising appointment, see the aid impact at the moment of dropping a class, and confirm consortium aid at a second college. These are flagged in the bank as a later wave and excluded from the runs we can start now.

Operations, tracking, and coverage How the work stays coordinated at scale

The coordination problem is real: many testers, one scenario each, no duplication, every result logged, and a clear view of what is left. The approach layers three pieces that already mostly exist.

1 · the bank

Scenario assignment

The scenario bank is the single source of truth. Each scenario has an ID. A college representative assigns one scenario to one tester by putting their initials and the date on that row, so nothing is run twice and coverage is visible in the sheet itself.

2 · the form

Capture, not assignment

The existing Jotform capture form is the right instrument for recording a run: persona, college, task, path taken, time, ease, and severity. It captures results well but does not assign work. We keep it for capture and add the thin assignment layer above in the bank, rather than rebuild the form. The form's entries key back to the Scenario ID.

3 · the dashboard

Coverage at a glance

The coverage dashboard reads the bank and shows, per college and per journey stage, how much has been walked and where the gaps are, plus the worst barriers found. It is how the co-chairs and the steering committee see progress without reading the sheet.

Recommendation: keep the Jotform capture form for recording results, and use the scenario bank as the assignment and coverage layer on top of it. This is the cleanest path. It avoids building a new assignment system, reuses the live form, and keeps one ID linking a planned scenario to its logged result.

Tool stack for a no-budget study Free tiers that fit an unmoderated, multi-tester study at this scale

ToolFree tierProCon
Google Forms & SheetsUnlimited, freeNo tester cap, no run cap, the bank and capture both live here, everyone already has accessNot a purpose-built usability tool; no automatic task metrics
Maze (free)Limited studies and responses per monthReal unmoderated task testing with first-click and path metrics built inFree tier caps monthly responses; built for prototypes more than live sites
Optimal Workshop (free)Small free studiesBest-in-class card sorting and tree testing for findability, exactly our core questionTight participant caps on free; one study type at a time
Useberry (free)Limited monthly testersUnmoderated flows with heatmaps and path analysisLow free-tier limits; prototype-oriented
LookbackTrial only, not a lasting free tierStrong moderated, think-aloud sessions with recordingNo durable free tier; best for a few deep moderated runs, not 5,000 at scale
Top pick: Google Forms and Sheets as the backbone, with Optimal Workshop's free tier for targeted findability checks. For a no-budget, multi-tester, unmoderated study at this scale, nothing beats Forms and Sheets: no response caps, no per-seat cost, and the scenario bank, the capture form, and the coverage dashboard all read from one place. The paid usability platforms throttle exactly the thing we have a lot of, runs, so we reserve their free tiers for small, high-value findability exercises (card sorts and tree tests) where their built-in analysis earns its limits. The live Jotform capture form already follows this model.

Guardrails Non-negotiable, on every run and every document

We hold to

  • No PII. Test accounts and initials only. Personas are research instruments, not real people.
  • Translate, don't standardize. Each college keeps its own names; we test findability, not conformity.
  • AI-only scope. Where the fix is automation, it takes routine work off staff so they focus on the students who need a person. No one is replaced.
  • Human contact is the success metric.

We route out

  • Process or policy barriers go to the owner who can fix them, documented, not solved here.
  • Classroom, curriculum, and faculty-tool issues go to Domain 2 (Teaching & Learning).
  • Data-governance and security questions, including the Salesforce data concern, go to Domain 1.
  • Retention and IR data needs are coordinated with Domain 3.

What we need to proceed The honest list of blockers

Per-campus test accounts

GCC can run now, on an existing GCC student login. The other nine colleges each need one test-student account scoped to that college's live systems, because a student login only reaches its own college. The district Test Student 1, 2, and 3 accounts referenced in our earlier work may be the source; we need confirmation of which test accounts reach each college and whether they can authenticate the login-walled flows. The accounts-we-need list, one line per college, is in the scenario-bank workbook.

Real-student persona validation

The 36 personas are designed from known demographics and the GCC frustrations data, but they should be validated against real students before fieldwork, through the design-studio students and the Student AI Group. Validation confirms the barriers are real and catches any we missed.

The Salesforce-dependent wave

Advising assignment and booking, the aid impact shown at the moment of dropping a class, and consortium-aid confirmation cannot be tested until the district's Salesforce advising tool is in place. These scenarios are flagged in the bank and held for a second wave. The data-governance concern around that tool belongs to Domain 1.

Tester capacity

Reaching about 5,000 task-runs needs the human half staffed across the ten colleges, one scenario per tester at a time. The AI half can begin on the public paths immediately. A college representative per campus to assign scenarios and confirm coverage is the lightest workable structure.