Welcome to Agentic Roommate Matching thatmatches on the factors that actually end roommate pairs.
Most pairs are made by a six-question form filled out optimistically in July. Risely's agents match on the factors that actually end pairs, check both sides at week 3, and route friction to the RA early. The residence staff make every call.
Matching run · incoming class
comparing
incoming first-year · profile A
incoming first-year · profile B
midnight-plus night owl
sleep ·midnight-plus night owl
an honest six
cleanliness ·an honest six
sometimes, ask first
guests ·sometimes, ask first
weighing the factors that predict
The match runs on what predicts peace.
Sleep schedules, cleanliness honesty, and guest tolerance are the factors that actually end roommate pairs. Risely's agents match on those three, and treat the rest of the July form as the optimism it is.
What the match is allowed to care about
predicts the pair's year
asked in July, predicts nothing
The pairs that end always end over the first three. The matching weights them accordingly.
An honest six beats a nine paired with a three.
Two students who both admit to average tidiness will hold. The spotless one paired with the chaotic one files a room-change request in October, and the matching knows that in July.
Cleanliness, as the matching reads it
Agreement predicts peace better than tidiness does. The matching pairs the honest answers, and it never pairs the extremes.
Week 3 is the check that saves October.
Risely's agents check in with both roommates separately. A friction word routes to the RA early, and the pairs that strain get help while the fix is still a conversation.
The first six weeks run on a schedule
running
October's breaking point stops being the first signal anyone hears.
Next year's matching learns from this year.
Every match factor ties to the outcome it produced. The pairs that strained teach next year's matching, and the questions that predicted nothing get retired from the form.
The loop the office never had time to run
October's outcomes tie back to July's factors
every mediation and every quiet pair
The questions that predicted nothing get retired
the form gets shorter and sharper
Next year's matching starts smarter
trained on this campus and nowhere else
Next year's algorithm learns from this year's Octobers.
One thing moves. The match re-runs.
A friction word in a week-3 reply, a room-change request, a cancelled contract. Any of them changes a pairing, and every one runs the same way.
A week-3 reply lands
one friction word in it
The pair flags
read on both sides, separately
The RA hears it early
while the fix is a conversation
The mediation drafts itself
held for the RD to approve
The outcome feeds next year
this pair teaches the matching
What changes, and how it runs.
Good pairs stay quiet
The 88% match never appears in a mediation queue.
Strained pairs get help at week 3
October stops being the first time anyone hears about it.
The matching gets smarter every year
Outcomes tie back to the factors that predicted them.
The technical read
Connects read-only to your housing system, SIS, and roommate questionnaire to start
An RD approves every re-match and mediation
Every action lands on an append-only log
Runs alongside what you have. Nothing rips out.