Glean Foundations and Search Intent
Core concepts and intent framing
You'll be able to
- Define enterprise search intent and relevance
- Explain baseline retrieval flow
Read first
Day 1 Lab
- Define three user intents for a single topic.
- Write expected top-3 results for each intent.
- Compare with actual returned results.
Working Example: Intent vs. Ranking Mismatch
One query, three intents — and how to tell *which kind* of failure you're looking at when the top result is wrong. This is the distinction the check-for-understanding asks for.
Query: "vacation policy"
Intent Expected top result
-------- --------------------------------------------
fact → a snippet: "Employees accrue 15 days/year"
document → the official PTO Policy PDF
person → the HR partner who owns the policy
Diagnosing a bad #1 result:
Wrong TYPE comes back (a doc when you wanted the person) -> INTENT mismatch
Right type, but the best item sits at #6 instead of #1 -> RANKING mismatchIntent mismatch is fixed by *reframing the query* (or picking the right result type); ranking mismatch is fixed by *signals and tuning*, not by rewording. Naming which one you have is the whole skill of Day 1.
Check for understanding
How would you distinguish intent mismatch from ranking mismatch?
Check yourself
1. What does Glean's Enterprise Graph primarily connect?
2. When is a natural-language question better than keywords?
3. Your top result is the wrong TYPE of thing entirely. That is a...
4. What does a result's snippet tell you?
5. What does `type:document onboarding` do?
6. `app:slack quarterly roadmap` returns...
7. In one sentence, what does `from:me` do?
In 10 seconds
“Great Glean search starts before you type: decide if you want a fact, a document, or a person. That one choice shapes your keywords — and explains every result.”
Reading Results and Relevance