Case studies
Three scores. Three problems solved.
Engagement tells you who just went quiet. Readiness tells you whose file is stuck. Yield tells you who is actually movable. Each score answers a different question — and each one drives a different counselor workflow. Here is what that looks like in practice.
All scenarios are composites. No real student records are referenced anywhere on this page — consistent with our no-PII architecture.
Case study 1 — Engagement
Stopping the stealth melt at a private liberal arts college
~3,400 undergrad · selective admissions · Midwest
The problem
Every spring, high-yield applicants who had been warm through February would quietly go silent by late March. By the time a counselor noticed, the student had deposited elsewhere. With Slate showing static snapshots, counselors couldn't distinguish a student who had always been quiet from one who had just gone quiet — and there was no recoverable intervention window.
The Celia signal
Celia computes an engagement_delta on every overnight run — the change since the prior run, not just the current level. When a high-yield student's Engagement dropped 20+ points in a single week, the team's 'Watch: Quiet Drop' Slate query surfaced them by morning.
The counselor workflow
- Daily Slate query: Yield ≥ 0.55, Engagement < 40, engagement_delta < −20, risk IN (high, critical)
- Counselors only acted on flags where ss_celia_engagement_conf ≥ 70 — low-confidence drops skipped
- Next action token stealth_melt_recovery triggered a personal call, not a mass email
- Celia's next_action_why field gave each counselor a natural, student-specific conversation opener
Celia outputs used
+41
Deposits above prior-year baseline
63%
Re-engagement rate on flagged students
+7.9%
Year-over-year deposit improvement
Key insight
The Engagement score alone is not what moved the needle. It was the delta. Counselors had always known some students were quiet. What they didn't have was a signal that said 'this specific student just went quiet this week.' That change-detection is what created a recoverable intervention window.
Case study 2 — Readiness
Clearing the document bottleneck at a regional Catholic university
~6,800 undergrad · moderately selective · Northeast
The problem
Hundreds of engaged applicants had incomplete files — missing transcripts, pending FAFSA verification, overdue recommendations. Counselors knew the problem existed but couldn't efficiently find who needed which intervention. Mass "checklist reminder" emails generated ~18% open rates and counselor fatigue. Students who wanted to enroll were falling off the funnel because no one reached them with the specific next step.
The Celia signal
The Readiness score measures how close a file is to decision-ready. The addressable cohort was defined by the intersection of high Engagement (≥60) and low Readiness (<40) — students who wanted to complete their application but had hit a bureaucratic wall. Celia's next_action_1 token (request_document) and next_action_why field named the specific missing item for each student.
The counselor workflow
- Priority queue: Engagement ≥ 60, Readiness < 40, Yield ≥ 0.50, risk IN (medium, high), readiness_conf ≥ 65
- Next action token request_document pre-filtered to only students where Celia was confident about the gap
- Each counselor email named the exact missing item from next_action_why — not a generic checklist blast
- Counselors averaged 22 targeted nudges per day — operationally sustainable at scale
Celia outputs used
+18 pts
Document completion rate (61% → 79%)
63%
File-forward rate after targeted nudge
+89
Additional deposits from recovered files
Key insight
Readiness without Engagement would have been noise — incomplete files often belong to disengaged students who have already moved on. The intersection of high Engagement + low Readiness identified students who wanted to complete their application. Generic reminders treat them the same as the disengaged ones. Celia separates them.
Case study 3 — Yield
Prioritizing deposit-close calls at a regional public university
~14,200 undergrad · open-access with selective programs · Southeast
The problem
With a 1:800 counselor-to-applicant ratio in the deposit-close window, each counselor could make 25–35 meaningful calls per week. Without a prioritization signal, they used admission date or program interest — both poor proxies. Counselors estimated 35–40% of calls reached students with very low enrollment intent. The movable students — the ones a meaningful conversation could actually close — sat uncalled.
The Celia signal
The Yield score is a probability calibrated against the institution's own past depositor cohort — not a generic model. The insight: the highest-leverage calls are not to the very-high-Yield students (they're coming anyway) or the very-low-Yield students (they've decided against you). The 0.35–0.69 "swing case" band is where a meaningful conversation changes the outcome.
The counselor workflow
- Swing cases (0.35–0.69 Yield, risk = medium, yield_conf ≥ 70): personal phone call ordered by Engagement
- Next action token schedule_call confirmed the counselor action; send_admit_packet for the upper swing band
- High Yield (≥ 0.70): deposit packet + automated follow-up — counselor time reserved for swing cases
- 87 swing-case students reviewed for merit aid bump; 29 of 44 who received awards deposited before May 1
Celia outputs used
+3.6 pts
Deposit yield rate (31.2% → 34.8%)
+111
Additional deposits year-over-year
~$4.2M
Est. 4-year tuition revenue impact
Key insight
The Yield score's highest-leverage use is not identifying who will enroll — it's identifying who is movable. Celia identifies the swing-case band precisely because it's calibrated against your institution's own historical depositor cohort. A generic model applied to your population would blur that line.
What the three cases share
Three scores. Used together. That's the system.
Scores compound
High Yield + low Engagement is a stealth-melt risk. High Engagement + low Readiness is a document bottleneck. None of the three scores work in isolation — the risk tier and recommended action are computed from their combination.
Deltas beat snapshots
In all three cases, the insight that drove action was change, not absolute level. A student at Engagement 38 who was at 72 last week is a completely different situation from a student who has always been at 38. The delta and stability_warning fields are what make Celia a time-aware system.
Counselor agency stays intact
Every team that saw strong outcomes treated Celia's next_action_why text as a conversation starter, not a script. Counselors who know a student's off-Slate context can override any recommendation. The data anchors judgment — it doesn't replace it.
See it end-to-end
Want to see the exact output Celia writes into your Slate?
One anonymized student, run through Celia's nightly pipeline. The exact 33
ss_celia_* fields, the PII-free input, and what
a counselor opens the next morning.
90-day free pilot
Pilot Celia on your Slate. Free for 90 days.
Tell us about your institution and your Slate setup. We'll book your kickoff, provision your workspace on the call, and you have 90 days to verify Celia moves the metric you care about. No card. No commitment. Real Slate integration.