Why this dashboard is shaped like this

Admissions does not run on a calendar, it runs on a cycle. One class is recruited over roughly thirteen months, against a fixed number of seats and a fixed reply date, and then the whole thing starts again. Three consequences shape every screen here:

• There is no "last month". Applications in January and deposits in April are not comparable to each other, so nothing is measured month over month. Everything is measured against the prior cycle at the same day.
• Stages are stocks, not flows. "Admitted" is a state a person is in, not an event in a window. Counts only ever grow within a cycle.
• The deadline is the denominator. Being at 60% of target means nothing without knowing where you were at the same point last year.

The as-of control

The slider at the top is not a date filter — it does not narrow a window. It rewinds the entire cycle to a single morning and shows what the office would have seen on that day: the stage counts that had happened by then, the work still sitting in the queue, the pacing against last year. Drag it back to January and the deposits vanish because they had not been paid yet. The marks underneath jump to the milestones — deadlines, decision release, May 1, the start of term.

The stages

• Inquiry — submitted a Request for Information. Applicants who never inquired get an inquiry record created by their application, so the funnel nests.
• Application Started — opened the application and saved at least one section. A partial application.
• Application Submitted — submitted and fee paid. This is the "applications" number a university publishes.
• File Complete — every required document received. Only complete files are read, which is why the admit rate is calculated against this stage and not against submissions.
• Admitted — offered a place. Denies and waitlists are decided the same day and are not counted here.
• Deposited — paid the enrolment deposit. Gross of melt.
• Net of Melt — deposits less the students who have since withdrawn. This is what becomes the entering class.

The stage bars use a square-root scale. The funnel spans 11,000 inquiries down to a few hundred seats; on a linear axis every stage past the application is an invisible sliver, and a log axis misrepresents the distances. Square root keeps the ordering honest and the small stages readable.

The projection

Projected class = net deposits now × (prior cycle's final ÷ prior cycle at this same day). It is a pacing model, not a forecast: it assumes this cycle finishes the way the last one did. It knows nothing about a changed aid strategy, a new competitor, or a bad melt summer. Early in the cycle, when the denominator is tiny, it swings wildly — that is a property of the method, not a bug.

Comparisons and what the filters reach

Filters apply to applicants and inquiries alike, so the top of the funnel moves with everything else. The class target follows the Territory filter — look at one region and it is judged against that region's own goal.

One honest limitation: the prior-cycle comparison is stored as daily stage counts split by type only, not by the full detail. So the Type filter moves it, and the Territory, Program, Residency and Source filters do not. That is a deliberate trade for payload size, and it is the kind of thing worth knowing before quoting a number off a filtered view.

Money

Published tuition and fees are $52,400. Discount rate is institutional aid as a share of that — the number a board and a bond rating look at. It moves against yield: every extra point of discount buys students and gives away revenue, which is the tension the yield and revenue sections exist to show. Recruitment spend is cycle-to-date: only months that have fully elapsed by the selected day are counted, so an early as-of date does not divide a year of budget by six weeks of results.

What the data is built to show

• The class lands just under target. Deposits are up on last cycle and it is still not quite enough — the target moved too. The projection tells you that in March, not in August.
• Visiting campus is the strongest lever there is. Admits who set foot on campus deposit at roughly twice the rate of those who never did. Note the caveat in that panel: visits are self-selected, so some of the gap is students who were already coming.
• Discount buys yield, and it is not linear. Walk the discount bands — the jump between the bottom two is far larger than between the top two. The last ten points of discount buy very little.
• Merit aid is doing exactly what it exists to do. Yield across the academic bands shows no clean gradient — the top band converts at 28%, no worse than the middle. That looks like a non-finding until you notice the discount running alongside it: the strongest admits are discounted 61% and the weakest 22%. Strong admits hold the most competing offers and would yield worst; the money is what flattens the curve. It is an expensive achievement, not an absence of one.
• Parents book the visits; transfers do not need them. Parent-initiated inquiries visit campus at roughly 58% against 12% for transfers, and yield highest of the three. Transfers still out-yield prospective students despite barely visiting — the visit lever hardly moves them, because they arrive having already decided to leave where they are.
• Cost per inquiry is a trap. Sort the source table by Cost/Inq and purchased name lists win easily — around $32 to $50 an inquiry against $74 for a high-school visit. Sort by Cost/Student and it inverts: the visit lands a student for roughly $1,400 where the lists need $1,900 to $2,600, and Google Ads — unremarkable on inquiries — turns out to be the most expensive student in the table at over $4,000. Optimising against the cheap column buys volume that never enrols.
• The admit bar is out of step with the seat plan. The split is clean and it is entirely about selectivity: every program admitting under half its applicants — Nursing, Engineering, Computer Science — finishes short of its seats, and every program admitting around two thirds — Business, Liberal Arts, Education — overflows. Nursing carries the heaviest demand of all at 8.3 applications per seat and still under-fills. Nothing in that panel is a demand problem; it is a decision problem, and it only becomes visible once seats and admits sit on the same row.
• Rewind to mid-January to watch the work queue at its worst: hundreds of partial applications and incomplete files, all still recoverable, with a named counselor against each one.