Factual circumstances
Factual circumstances are represented in questions a product asks users about their actual situation, not their mental state. "Are you traveling with kids or solo?" and "Have you booked accommodation yet?" are factual; "What kind of trip are you hoping to have?" and "How adventurous are you feeling?" are not. Users tend to answer factual questions more readily and reliably than questions about feelings, self-image, or goals.
Definition
A design commitment for sorting hat questions and other intent-detection instruments: ask about a user's actual situation, not their mental state, and infer intent classification from circumstantial answers rather than requesting it directly.
What users answer reliably
Users are reliably bad at self-classifying their own intent and reliably good at answering factual questions about their actual situation. Asked what kind of onboarding they're trying to do, what level of expertise they have, or what they're hoping to accomplish, users project an identity, sort themselves into categories they may not actually fit, or answer aspirationally. Asked whether their product is launched, whether they're tracking retention, or whether they currently have an onboarding project, they answer the situation in front of them.
The translation craft in sorting hat design is wording each question so that an honest factual answer is the easy answer to give. Awkward, leading, or self-presentation-inviting questions produce instruments that classify users by their projected identity rather than their actual circumstances.
Classification by inference
Intent classification gets inferred from circumstantial answers, not asked for directly. A user who reports their product is post-launch and they're tracking retention has revealed more about their actual intent than the same user answering "I want to optimize my conversions." The optimization answer passes through the user's self-presentation; the factual pair reports the user's situation regardless of how they prefer to describe it.
Where the commitment applies
The commitment applies to any intent-detection signal stream that takes direct user input. Sorting hat instruments are the primary case because the team designs the questions and the user types or clicks the answers. Other signal streams in intent detection — search-term referrals, acquisition-channel data, campaign attribution — are circumstantial by nature; the user generated the signal in the course of their actual activity rather than in response to being asked.
Example
Two versions of a sorting hat instrument for an activation-consulting offering, both trying to classify whether the user is pre-launch, post-launch with conversion problems, or post-launch with retention problems.
Mental-state version:
- "What stage is your company at?"
- "What's your biggest growth challenge right now?"
- "What kind of help are you looking for?"
Factual-circumstances version:
- "Is your product launched?"
- "Are you currently tracking retention?"
- "Are you working on an onboarding project?"
The mental-state version invites users to position themselves — "growth challenge" invites a self-narrative, "what kind of help" invites a category-selection — and the resulting classification reflects how the user presents their situation. The factual-circumstances version asks about the situation directly and lets the classification fall out of the answers: a no/no/no user is pre-launch, a yes/yes/yes user is in retention-work territory, and the offering classifies the user without the user having to self-classify.