Skip to content

Setup & integrations

Data dictionary

The per-Org map from Slate field names to semantic meaning. This is what lets Celia reason about meaning instead of opaque field codes.

Who uses this
Every authenticated team member.

What it is

Your institution's translation layer — and the access-control list for what reaches Celia. It does two jobs:

  1. Meaning — Celia ingests Slate field names verbatim, then looks them up here to know what each one represents. Without an entry, Celia sees the raw column name (fine for canonical names, useless for custom SS_* fields).
  2. Access — before a student batch is sent to Celia, CeliaConnect strips any column not in this dictionary (and not a standard Celia signal key). A field your Slate query returns but that has no dictionary entry never reaches the AI. Adding an entry immediately unlocks that field for scoring.

The same dictionary also controls positive-cohort learning: when computing what historically converted students look like, CeliaConnect only tallies fields that appear here. A field with no entry contributes nothing to the institution-specific priors Celia receives at score time.

What you can do here

Anatomy of an entry

  • Field name — exact Slate column name (e.g., SS_FAFSA_STATUS).
  • Label — human-friendly version ("FAFSA submission status").
  • Type — text, number, date, boolean, enum.
  • Description — what the value represents semantically.
  • Allowed values — for enum types, the canonical value set.

Common workflows

Stand up a brand new dictionary

  1. Export your Slate field schema as CSV.
  2. Add a Label, Type, and Description column.
  3. Import via /settings/data-dictionary/import.

Fix a single bad description

  1. Search for the field name.
  2. Click into it.
  3. Edit the Description and save. Celia uses the new text on the next analysis.

Common gotchas

Celia keeps using the field name itself

You haven't added a dictionary entry for that field yet, or the field name in the entry doesn't exactly match the Slate column name.

Talk-to-Celia says "I don't know that field"

Same root cause — add a dictionary entry and refresh.

A custom field isn't influencing scores

Two conditions must both be true: the field must have a dictionary entry, and your In Progress Slate query must SELECT it. Meeting only one has no effect. Verify both, then click Refresh fields on the Flow and run again.

Positive-cohort priors aren't showing my new field

Priors are computed from the Positive Outcomes query run at the time of each Flow run. After adding a dictionary entry, the field will appear in the priors on the next scheduled run — or immediately if you click Run now.