Structured reference. A dense, rule-based version of the persona-setup methodology. Paste this page's URL into an AI assistant and ask it to apply the methodology to your business, or read the prose version.
Hub · Setup · Structured reference

Setting up your personas and buying contexts (structured reference)

Rules for writing personas and buying contexts that produce useful AI research. A persona is who the model is asked to be. A buying context is the moment they're asked about. The most common failure is writing buying contexts as categories ("SOC2 compliance") instead of situations ("First SOC2 attempt to unblock an enterprise prospect"). Categories are always true and trigger nothing. Situations describe what changed.

What every persona setup must satisfy

  • AI gives different answers depending on who it perceives is asking. The persona definition is the input that shapes the answer.
  • A persona is a stable identity (Role or Profile). A buying context is a discrete moment that puts the persona in market right now.
  • Strip-the-brand check applies to both fields: neither the Role/Profile nor the buying context should name the user's category or named competitors.
  • Buying contexts are short phrases (typically 4 to 10 words) that name a situation or trigger. Not categories. Not Jobs-to-be-Done sentences.
  • Embed segment signals (enterprise, mid-market) in role details and trigger specifics, not in labels.

Pick your personas

A persona is one of the distinct buying paths into the user's product. Different buying motions → different personas. Same triggers + outcomes → one persona. Most companies need 2-4. For multi-audience businesses, build parallel persona stacks with at least one persona per side. Two patterns: (1) B2B distribution + B2C consumer (insurance, CPG, healthcare): one B2B set, one B2C set. (2) Two-sided marketplace, network, or platform: retailers and brands (Faire), candidates and employers (job platforms), borrowers and depositors (lending), buyers and sellers (Etsy). Weight by commercial reality (lending often skews borrower-heavy; marketplaces vary).

How to pick

  1. List the distinct buying motions the user serves. Different reasons someone shows up shopping. Two motions with the same triggers and outcomes collapse to one persona.
  2. Identify who decides (or recommends, or blocks) in each motion. B2B → a role type. B2C → a recognizable identity.
  3. Pin them to the kind of organization or household they're in. The combination of role and context is what makes the persona durable across years.
  4. Cover the segments that account for the bulk of revenue or audience. Stop at 2-4 personas. More usually means split where you should have combined.

Field structure

  1. Title — short name in the platform.
  2. Role (B2B) or Profile (B2C) — stable identity.
  3. Buying contexts — three short situation phrases. Covered in Step 2.

B2B format

Title: The First-Time Compliance Builder
Role: Head of Security at 50-500 person SaaS companies
      pursuing their first SOC2 or ISO 27001

Role = role type at plural organizations with one or two qualifiers (size, stage, specialty, geography). Three or more qualifiers stacked makes the persona unwieldy. The same person should remain this persona across years.

B2C format

Title: The HDHP Working Parent
Profile: Late-30s working parents with kids at home,
         covered by a high-deductible health plan

Profile = 2-3 stable axes that drive buying for the category. Common axes: life stage, household structure, employment status, primary coverage type, geography, relationship to the category. Pick axes that drive most of the variance for the user's category.

Strip-the-brand check

Remove the user's category and competitor names from the Role/Profile. What's left should still describe a coherent person worth reaching.

Bad — leads the witness
  • "Marketers researching tools in your category"
  • "Users of [competitor] considering alternatives"
  • "VP of Marketing at SaaS companies evaluating [your category] vendors with a $150K budget"
Good — durable identity
  • "VP of Marketing at Series B SaaS companies"
  • "Senior compliance officer at U.S. community banks with $1B-$10B in assets"

Strip-the-moment check

If the Role/Profile contains "newly," "recently," "post-X," "just-Y," it has a moment baked into the identity. Move the moment into a buying context. The Role/Profile is durable identity only.

Worked examples

Vanta (compliance automation) — three distinct buying motions:

The First-Time Compliance Builder
   Head of Security at 50-500 person SaaS companies
   pursuing their first SOC2 or ISO 27001

The Regulated Mid-Market GRC Lead
   GRC lead at U.S. financial services or healthcare
   firms with 200-5,000 employees

The Enterprise-Ready Security Architect
   VP of Information Security at mid-market companies
   pursuing Fortune 500 customer contracts

Datadog (observability):

The High-Traffic Platform Lead
   Director of Platform Engineering at consumer-facing
   SaaS with seven-figure user counts

The Migration-Stage SRE
   Site Reliability Lead at growing companies migrating
   off self-hosted observability stacks

The Enterprise Observability Architect
   Principal SRE at Fortune 1000 companies running
   heterogeneous cloud infrastructure

Peloton (consumer fitness):

The HDHP Working Parent
   Late-30s working parents with kids at home, dual-income,
   covered by a high-deductible health plan, previously athletic

The Apartment Cardio Seeker
   25-40 year olds in dense urban U.S. metros without practical
   space for traditional equipment, with disposable income

The Post-Pandemic Returner
   35-50 year olds whose pre-COVID fitness habits lapsed
   and who haven't restarted at a gym

Pick your buying contexts

Each persona gets 3 buying contexts, ordered most-frequent first. Each is a short phrase (4 to 10 words) naming a discrete moment that puts the persona in market right now.

The single biggest failure mode: categories vs. situations

A category is a permanent feature of the market — always true for everyone in the segment. A situation is what changed for a specific subset this quarter. Categories look like buying contexts but trigger nothing.

Vanta (compliance automation) Bad — categories
  • SOC2 compliance
  • ISO 27001 readiness
  • Vendor risk management
Good — situations
  • First SOC2 attempt to unblock an enterprise prospect
  • ISO 27001 demand from a new European customer
  • Annual audit prep without a dedicated compliance hire
Datadog (observability) Bad — categories
  • Observability at scale
  • APM tooling
  • Cloud monitoring
Good — situations
  • Production incident took six hours to root-cause
  • Migrating off Splunk after a cost spike
  • Hitting AWS CloudWatch service limits
Peloton (consumer fitness) Bad — categories
  • Home fitness equipment
  • Cardio workouts
  • Weight loss
Good — situations
  • Gym membership lapsed and never restarted
  • Newborn at home, can't get out to work out
  • Doctor flagged a health number at the annual physical

Self-test: if your buying context could appear as a feature on a competitor's website, it's a category. Rewrite it as the moment that made the buyer care this quarter.

Don't lead the witness in the buying context

Strip-the-brand applies here too. Don't name the user's category or solution type inside the situation. "Annual channel planning kicks off; a marketplace bet is up for review" presupposes the buyer is considering a marketplace (the user's category) — rewrite as "Annual channel planning kicks off; needs incremental European reach."

Don't stamp segment labels on the context

If enterprise framing is wanted, don't write "Enterprise orchestration evaluation." Embed enterprise via role details (CDO, model risk management, Chief AI Officer) and trigger specifics (exec mandate, 5x scale, M&A, security renewal). The research reads enterprise from the situation, not from the label.

Two validation tests

In-market test. After this situation fires, is the persona actually in market for the user's product? In underwritten categories (insurance, lending, healthcare), the direct event often disqualifies them — they're calling their existing provider, not shopping. Use adjacent situations: moments where someone close to the persona was hit. Same urgency, no eligibility issue.

Next-action test. What's the literal next thing the persona would do after this situation fires? If it's "call existing provider" or "fight a bill," the situation is too early — they aren't shopping yet. Especially relevant for financial-anxiety situations (deductible spikes, premium increases). Verify the next action is shopping for the user's specific product.

Add your personas to the platform

In the Unusual admin: Settings → Personas → Add new persona. Title → Name field. Role/Profile → Description field. Each buying context → one Add buyer context click. Click Create profile. Repeat per persona. Personas don't need to be perfect at submission; the first round of research will surface refinements.

If/then checklist

Persona-level

  • If the Role/Profile names the brand, category, or competitors → leading the witness. Strip and confirm coherence.
  • If the Role/Profile contains "newly," "recently," "post-X," "just-Y" → move the temporal qualifier into a buying context.
  • If a B2B Role has 3+ orthogonal qualifiers → reduce to one or two.
  • If a B2B Role names a single specific company → widen to the category of organizations.
  • If 2 of 3 buying contexts overlap between two candidate personas → collapse them or differentiate the situations.

Buying-context-level

  • If the context is longer than ~10 words → tighten. If it wants to be a sentence, split into multiple contexts.
  • If the context could appear as a feature on a vendor website ("SOC2 compliance," "multi-region orchestration") → category, not situation. Rewrite as the moment that made the buyer care.
  • If the context names the user's category or solution type → leading the witness. Rewrite without the category framing.
  • If the context stamps "enterprise" / "mid-market" / "startup" as a label → embed the segment signal in the role details and trigger specifics instead.
  • If the situation disqualifies the persona under underwriting rules → swap to an adjacent situation.
  • If the persona's literal next action is "call existing provider" → situation is too early; move closer to the actual search.
  • If only 2 contexts surface as truly distinct → push for more or accept 2; don't pad with categories.
  • Order contexts by frequency of the situation, leading with the most common real-world moment.

Worked example: correcting a flawed submission

Original submission:

Title: SaaS Marketing VP
Role: VP of Marketing at Series B SaaS companies evaluating AI brand
      management platforms with a $150K budget
Buying contexts:
  - AI search optimization
  - Brand visibility in LLMs
  - Currently evaluating vendors

Role: "evaluating AI brand management platforms" names the user's category → leading the witness. "$150K budget" is a third orthogonal qualifier. Strip both:

Role: VP of Marketing at Series B SaaS companies

Buying contexts: all three are categories or ongoing states, not situations.

  • "AI search optimization" → category. Ask: what moment made them care? Possible rewrite: "Board asked about ChatGPT at the quarterly review."
  • "Brand visibility in LLMs" → category. Possible rewrite: "Forwarded Slack screenshot of ChatGPT recommending a competitor."
  • "Currently evaluating vendors" → ongoing state, not a discrete moment. Ask: what triggered the evaluation? Possible rewrite: "Annual marketing planning opens; AI search line item is new."

Re-ordered most-frequent first, the final submission is:

Title: SaaS Marketing VP
Role: VP of Marketing at Series B SaaS companies
Buying contexts:
  - Annual marketing planning opens; AI search line item is new
  - Board asked about ChatGPT at the quarterly review
  - Forwarded Slack screenshot of ChatGPT recommending a competitor

Prefer the prose version?

Same methodology, written as a step-by-step narrative.

Open the prose version