The second service
Celia surfaces suspicious application patterns, financial-aid abuse shapes, and data anomalies for your team’s review — computed from the same anonymized signals as everything else CeliaConnect does. Never an accusation, always a signal. Celia never labels a student a fraud.
One platform, two services
Both services ride the same daily Slate round trip — anonymized signals out through your
Slate Query Service, structured ss_celia_*
fields back through Source Format.
Service 1 · Every Flow, every plan
Know which admitted students will melt and what to do about it. Engagement, Readiness, and Yield analyses plus a Risk tier and a recommended next action — written into every student record, every run.
Service 2 · Optional · Enterprise plan
A fourth parallel analysis on Flows where you enable it. Celia examines the same anonymized signals for suspicious patterns across four categories and flags files that deserve a second look — for a human on your team to review.
Why now
Admissions and financial-aid teams increasingly report application pools that contain records no genuine applicant produced — bulk-generated applications, aid-seeking patterns that follow a script, files whose internal data cannot all be true at once. Reviewing for this by hand means re-reading thousands of clean files to find a handful of odd ones.
The dedicated fraud vendors solve it by ingesting full student PII — names, SSNs, device fingerprints — into their systems. For many institutions that trade is the reason the project never clears security review.
Fraud Protection takes the other path: pattern analysis on the anonymized signals Celia already receives, with every flag routed to a human for review. No new data leaves Slate. No verdict is ever automated.
Four detection categories
Every flag names the category (or categories) that fired, so your reviewer knows what kind of pattern they are looking at before they open the file.
application_integrity
Internal signals of an application that does not hang together — timelines that contradict each other, milestone sequences that no genuine applicant produces, activity patterns inconsistent with the stage the file claims to be in.
financial_aid_pattern
Aid-seeking behavior that deviates sharply from your institutional baseline — for example, aid-stage velocity and document patterns that resemble known abuse shapes rather than a student working through FAFSA.
data_consistency
Fields that disagree with each other across the record — stage vs. milestone state, declared program vs. behavioral trail, dates that cannot both be true. Often these are data-entry errors; the flag is how you find out which.
cohort_anomaly
A record that is a statistical outlier against its own cohort — clusters of near-identical applications, activity bursts at improbable hours at improbable scale, patterns that only appear when applications are generated rather than written.
What a flag contains
None · Low · Medium · High · Critical — a five-step severity scale, filterable in Slate list views. Most records, most days, read None.
0.00–1.00. How sure Celia is about the signal itself — so your team can triage a 0.91 flag differently from a 0.55 one.
Which of the four detection categories fired. A flag always tells you what kind of pattern it saw, never just "suspicious."
Up to five specific, anonymized observations behind the flag — the evidence your reviewer starts from, not a black-box score.
A short plain-language summary (300 characters max) of why this record deserves a second look.
A simple yes/no: should a human open this file? Celia recommends review. It never recommends rejection.
The same Slate round trip
Leg 1 — Slate Query with service access, out
Fraud Protection reads the exact same anonymized input as Enrollment Intelligence: anonymous IDs, stage transitions, timestamps, milestone states, and your institutional baselines. No new query, no new fields, and — as everywhere in CeliaConnect — no names, emails, SSNs, or any other PII.
Leg 2 — Slate Source Format, back
On Flows with Fraud Protection enabled, four extra fields ride the same writeback and land on the matching student record:
Your team builds a Slate list view filtered on
ss_celia_fraud_level and works the review queue
inside Slate — the same "no new tool to learn" promise as the rest of CeliaConnect.
The architectural difference
Dedicated fraud-screening vendors ask for the most sensitive data an institution holds — names, SSNs, dates of birth, device fingerprints — because their approach is identity verification. Celia’s approach is pattern analysis: duplicate-shaped application clusters, contradictory timelines, and anomalous aid-stage behavior are all visible in anonymized signals. The suspicious pattern is in the behavior, not the name.
That means Fraud Protection inherits the same architectural guarantee as everything else CeliaConnect does: no student PII ever reaches the AI. Your security review of Fraud Protection is the security review you already did.
A signal surface, never a verdict
Every Fraud Protection output is designed for a human-in-the-loop workflow. Celia writes a severity level, a confidence, the categories that fired, and a short context note — and, where warranted, recommends that a person review the file. What happens next is entirely your institution’s process. Celia never labels a student a fraud, never triggers an automated action against an applicant, and most flags turn out to be exactly what a reviewer would hope: data-entry errors and duplicates worth cleaning up.
Frequently asked
Enterprise plan
Fraud Protection is available on the Enterprise plan as a per-Flow opt-in. Start with the 90-day pilot, see Enrollment Intelligence run on your own Slate data, and we’ll walk your team through what enabling Fraud Protection looks like for your intake Flows.
90-day free pilot
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.