Create Customer Personas from Your Existing Data

You'll end up with: Detailed customer personas built from your actual client data and conversations

Overview
30-45 min
Intermediate
Free
2 tools
Cost breakdown
ClaudeFree tier
Google Sheets / DocsFree
TotalFree
Common mistake

Demographics theater: pretty slides with age and a stock photo, but no link to what people do, ask, or object to in your real data. The fix: every persona must have at least three bullets that map to evidence—themes from tickets, quote patterns, funnel stage, or feature request clusters.

Before you start
  • Export or list 2+ data sources (e.g. CRM + support or survey CSV)
  • Redact or aggregate so you are not pasting real names, emails, or full transcripts
  • Pick one primary goal (messaging vs product vs sales)
  • Open Claude and optionally Google Sheets for tabular data
1

Inventory data you already have (evidence map)

List sources and what each can prove—behavior vs opinion vs firmographics.

Claude.aiFreeOpen Claude.ai
Exact action

1. Go to claude.ai and start a new chat. 2. Paste a short list of your data sources: CRM fields you use, support categories, survey questions, top pages, or deal stages. 3. Ask: "For each source: what can this prove, and what can it NOT tell us? Given our goal (messaging vs product vs sales), which source is highest-signal?" 4. Ask for a ranked list: primary source → secondary → nice-to-have.

You have a ranked evidence map in plain language and you know which 1–2 sources you'll lean on first.
If you only have a vague "we have customers" with no fields—export one real table (even 20–50 anonymized rows) or list 10 real support tags, then run this step again.
2

Export and prep a safe "evidence pack"

Create a small, non-sensitive dataset AI can read—summaries, tags, and paraphrased quotes.

Google SheetsFreeOpen Google Sheets
Exact action

1. Create a new spreadsheet. Name the first sheet "Evidence." 2. Add columns: Source | Segment hint (plan, role, or industry if known) | Theme | Example (redacted) | Frequency (rough). 3. Add at least 15–30 rows across your sources—or paste equivalent summary blocks if you're not using a sheet. 4. Open Claude in a new chat. Paste a sample of the sheet or your summaries. 5. Ask for a 5-bullet "data hygiene" check: PII risk, missing segments, and one suggested merge rule for duplicate themes.

You have a shareable internal pack you could hand to a teammate without privacy panic.
If you're pasting full transcripts or contact lists—stop. Replace each with theme + paraphrase + counts only, then continue.
3

Discover 3–5 segments (themes before personas)

Cluster jobs-to-be-done, objections, and triggers—segments first, polished labels later.

Claude.aiFreeOpen Claude.ai
Exact action

1. In Claude, paste a sample from your Evidence sheet—or key pivots: top tags, cancel reasons, "why now" from sales notes. 2. Prompt for 3–5 candidate segments. For each, ask for: label | defining signal (what you'd look for in data) | non-audience (who to exclude) | one proof quote pattern (paraphrased, no real names). 3. Cap at 5 segments. If you get more, ask which labels to merge and why.

You have a table of 3–5 segments with exclusion rules so they don't all sound the same.
If segments are only industry or geography—add at least one behavioral or workflow differentiator per segment before moving on.
4

Build persona one-pagers (evidence-linked)

Turn each segment into a one-page layout grounded in your pack—not generic market lore.

Claude.aiFreeOpen Claude.ai
Exact action

1. For each segment, prompt Claude using one chat or one thread per segment. Ask for this outline: Who (role) | Job to be done | Top 3 pains | Objections | Triggers / "why now" | Preferred proof (case study, demo, trial) | Where they show up (channel hint from your data). 2. Require "Evidence" bullets under each section: "Seen in: [your source/theme]"—no uncited claims. 3. Avoid fake first names. Use segment title + role (example: "Ops-led SMB — migration anxiety").

Each persona has at least three evidence-linked bullets and a clear job-to-be-done in one sentence.
If a section reads like generic marketing theory—rewrite it with Claude and insist every bullet ties to a theme or row from your Evidence pack.
5

Stress-test: merge, sharpen, and add activation

Remove duplicates, pick one primary use case (e.g. homepage vs sales deck), add team activation and misuse guardrails.

Claude.aiFreeOpen Claude.ai
Exact action

1. Ask: "Which two personas overlap the most? Merge them into one and give a decision rule for when to treat someone as A vs B." 2. Pick one primary use for this pass (messaging, product roadmap, or sales). Ask Claude to sharpen wording for that use only. 3. For each remaining persona, add Activation: one sales discovery question, one support macro or reply pattern, one product tie-in. 4. Add Misuse guardrails: "Never use this persona to claim…" with limits tied to what your evidence actually supports.

You can answer "Which persona is this campaign or pitch for?" without hesitation; overlaps have a clear routing rule.
If you can't choose a primary use—finish activation for one lane first, then duplicate the doc and rewrite headlines for the second lane.
6

Publish the master doc and refresh cadence

One consolidated doc plus a simple review schedule so personas stay tied to real data.

Google DocsFreeOpen Google Docs
Exact action

1. Create a doc titled "[Your business] — Personas (evidence-backed)." Add sections: Evidence map → Segments table → Persona pages → Activation → Review. 2. Paste polished sections from Claude. Add a footer: Sources and date updated. 3. Add Review cadence: e.g. quarterly—or after N new deals or N support tickets, whichever comes first. 4. Share the link with anyone who writes messaging, product specs, or sales scripts.

One link stakeholders can use; sources and date are visible; refresh triggers are explicit.
If the doc balloons—split personas into subpages or tabs, but keep one index page with the segment table and activation summary.

All done!

You now have: Detailed customer personas built from your actual client data and conversations

Explore more guides

Want this workflow built for your business?

Book a free audit