How Compass guides students

How Compass guides students

Compass is the student-facing finder in Pathways. A prospective student answers a
few questions and gets back a short list of programs the college actually offers,
with credentials, an estimated plan of study, and a cost estimate. This article
explains what is happening underneath: how Compass turns three simple choices into
a set of real programs, why the same answers always return the same result, and
why it cannot recommend a program the college does not offer. It does not cover how
to publish or configure Compass; that belongs to the how-to guides.

A guided finder, not a conversation

It helps to be clear about what Compass is not, because the word “guide” invites a
comparison it does not deserve. Compass is not an AI chatbot. There is no language
model, no free-text box where a student types a question, and nothing that composes
an answer on the fly. A student cannot ask Compass “what should I do with my life”
and receive an essay. What Compass offers instead is a structured path through three
questions, asked in order: first an interest, then a goal, then a pace.

• Interest is a program area, chosen from the areas the college has defined
  (for example, health sciences or business). Behind the scenes this is the
  program_area taxonomy, the same one that organizes the catalog.
• Goal is the kind of credential the student is after: a certificate, a diploma,
  an associate degree, or a transfer path. Transfer is treated as an associate-level
  goal, because that is the credential a transfer student earns before moving on.
• Pace is full-time or part-time, which changes how the plan of study is spread
  across semesters but never changes which programs qualify.

Each answer narrows the field. The interest decides which program area to look in,
the goal decides which credentials within those programs count, and the pace shapes
how the resulting plan is laid out. The reason for this fixed order is that it mirrors
how the catalog itself is arranged, from area down to program down to credential. A
student walking through Compass is walking down the same tree that a catalog editor
built, one branch at a time.

Why the same answers always return the same programs

Compass is deterministic. Give it the same interest, goal, and pace, and it returns
the same programs in the same order, every time. This is not a tuning goal or a
promise about model behavior; it is simply a property of how the code works. When a
student submits their three choices, Compass runs an ordinary database query. It asks
WordPress for published programs in the chosen area, then for each program it looks up
the published credentials whose titles match the chosen goal, then it arranges the
plan of study for the chosen pace. Matching a goal is a plain text comparison: a goal
of “certificate” matches a credential whose title contains the word “certificate.”
There is no ranking model, no randomness, and no external judgment call in the loop.

Because the logic is fixed, the only thing that changes the output is the catalog
itself. Publish a new credential, change a program’s area, or unpublish a program, and
Compass reflects that on the next search. Nothing has to be retrained or re-indexed.
This is a direct consequence of Compass reading the live catalog rather than a
separate copy of it. The catalog is the single source of truth, and Compass is a lens
onto it, not a second store that can drift out of sync.

The ordering students see follows the same principle. Programs a college has marked as
featured come first, and everything else is alphabetical. That is the whole rule.
There is no popularity score or personalization weighting quietly reordering the list.

Why Compass cannot invent a program

The most important thing this design buys is trust. Compass can only return programs
that exist as published records in the college’s own catalog, and only credentials
attached to those programs. It has no vocabulary of programs beyond what the college
has entered. If a college does not offer a welding certificate, no combination of
answers will produce one, because there is no welding certificate record for the query
to find. A student can never be pointed at a program that is not real, mispriced
against a program that does not exist, or shown a plan of study assembled from nothing.

This matters most in exactly the situation where a generative chatbot is least
trustworthy: the gap. When a student’s interest has no matching program, Compass
returns an empty result rather than inventing a plausible-sounding option to fill the
silence. An enrollment office can stand behind every result Compass shows, because
every result traces back to a record staff can open and check.

There is a cost to this design, and it is worth naming. Compass is narrower than an
open conversation. It answers only along the three axes it knows about, and only from
what is published. A student with an unusual or cross-disciplinary interest may not
see a natural home in the list, and Compass will not improvise one for them. Some
teams would prefer a more exploratory, conversational tool that can wander outside the
catalog. That is a legitimate preference, but it is a different tool with a different
risk profile. Compass makes the opposite trade on purpose: it would rather show less
than show something that is not true.

Where the numbers come from

The programs Compass returns are the college’s own. The labor-market figures attached
to them, such as wages and job outlook for the related occupations, are not. Those
come from public data sources, mapped to each program through standard occupation
codes, and they carry their own rules about freshness and gaps. That is a separate
topic with its own limits, and it is worth understanding on its own terms; see
What demand signals are and are not for how
that data behaves and where it stops.

Cost estimates follow a similar honesty rule. Compass only shows a dollar figure when
the college has configured a tuition rate. When it has not, Compass shows a note
pointing the student to admissions rather than guessing a number. The pattern is the
same throughout: show what is known, say so plainly when something is not.

The boundary of this explanation

This article is about how Compass decides what to show and why those decisions are
reliable. It deliberately stops short of the mechanics. How a college turns Compass on,
which page it lives on, how program areas and credentials get entered, and which
settings control the cost estimate are all operational tasks covered by the Pathways
how-to guides and setup material. The exact routes Compass calls, including the public
explore endpoint, are listed in the REST API reference.
What you should take away here is narrower and more durable: Compass is a finder driven
by the catalog, not a chatbot, and its reliability is a property of that design rather
than a feature bolted on top of it.