Faculty Workload Models That Reflect the Actual Work

Walk into most provost's offices and ask how faculty workload is calculated, and you'll get a formula that hasn't changed in twenty years: credit hours taught, multiplied by a standard factor, plus a fixed allowance for "service." It fits on one line of a spreadsheet. It is also, almost without exception, wrong — not wrong in the sense of a rounding error, but wrong in the sense that it measures a different thing than the thing anyone actually needs measured, which is how much of a given faculty member's working capacity a given semester consumes.

We've built workload models across three campuses now, and the pattern that shows up every time is the same: the credit-hour formula undercounts advising, undercounts research supervision, undercounts committee and accreditation work, and — this is the one that surprises administrators most — undercounts the variance between two faculty members teaching the identical course. A section of Intro Statistics taught by a second-year assistant professor who's never taught it before consumes meaningfully more hours than the same section taught by someone on their ninth iteration, and no formula built on credit hours alone will ever see that difference, because credit hours are a property of the course, not the person teaching it.

What the formula leaves out, specifically

Four categories consistently account for the largest gap between modelled load and actual load:

Advising, uncapped. Most institutions assign advising as a flat stipend or a rounding-error line item, regardless of caseload. A faculty member advising 12 majors and one advising 45 are doing structurally different jobs, and the formula treats them identically. When we've instrumented advising-session logs against actual time spent — not self-reported estimates, but calendar and LMS-messaging timestamps — the range between a light and heavy advising load within the same department is routinely 15-20 hours a month, unmeasured and unrewarded.

Research supervision that doesn't show up as "teaching." A faculty member directing four thesis students is running four independent, asynchronous mentoring relationships, each with its own cadence of meetings, drafts, and revisions. Institutions that count this as a flat "service" credit are pricing a semester's worth of one-on-one supervision the same as sitting on the library committee.

Committee and accreditation load that clusters unevenly. Service isn't distributed evenly across a department in any given year — it clusters around accreditation cycles, search committees, and curriculum reviews, and it clusters on specific people, usually the same people, because institutional knowledge concentrates. A formula that assumes a flat service allowance every semester misses both the clustering and the fact that some years are genuinely heavier than others for the same person.

Course prep variance by iteration count, not just credit hours. A new course prep — first time teaching it, no existing materials — runs somewhere between 1.5x and 2.5x the ongoing-maintenance load of a course in its fifth iteration, by every time-diary study we've run against faculty self-tracking. Treating "a course" as a fixed unit of work regardless of how many times it's been taught is the single largest source of the gap between what department chairs plan for and what actually happens each semester.

Why the fix isn't a more granular formula

The instinct, once you've catalogued what's missing, is to build a bigger formula — more line items, more granular weights, a coefficient for advising caseload and another for prep iteration. We've tried this and watched it fail in a specific way: a formula with twenty inputs is no more trusted by faculty than a formula with two, because faculty don't evaluate a workload model on its architecture. They evaluate it on whether it matches their lived experience of the semester, and a twenty-variable formula that still gets their load wrong is worse than a two-variable one that's honestly approximate, because it claims a precision it doesn't have.

What actually earns trust is a model built from instrumented data — actual advising caseloads, actual committee assignments pulled from governance records, actual prep-iteration counts pulled from course history — rather than self-reported estimates or departmental folklore, combined with a deliberately visible process for faculty to see and contest their own inputs before the model is used for anything consequential like course-release decisions or equity reviews. The model doesn't need to be more sophisticated than the credit-hour formula. It needs to be built from data faculty recognize as true, and it needs a correction path when it isn't.

The deployment lesson

Across the three campuses where we've deployed workload models, the rollout that worked started with transparency before optimization: publish what the model currently measures, let department chairs and faculty senate committees challenge specific inputs for a full semester before the model touches a single staffing or course-release decision. The rollout that struggled tried to solve equity and planning at the same time as building trust in the underlying numbers, and faculty — reasonably — treated the whole system as an administrative tool for extracting more teaching load, because that's what workload formulas have historically been used for. A model faculty don't trust gets gamed, ignored, or quietly overridden by the same department-chair judgment calls it was built to replace, and at that point you've spent a semester of political capital building a spreadsheet nobody uses.