Calibrate Context
Context is the truth layer every surface reads from. Tune its five surfaces in order, foundation docs first and Writing Calibration last, so every output inherits the quality.
Context is the truth layer every other surface reads from. Get it right and every output inherits the quality. Get it wrong and the best pipeline in the world writes confident, generic content. So calibration is the highest-leverage hour you'll spend in the workspace, and it's worth doing in the right order.
Context has five surfaces. Tune them foundation first and Writing Calibration last, because each one feeds the next.
Why the order matters
| Order | Surface | What it captures | Why it comes here |
|---|---|---|---|
| 1 | Foundation docs | Company overview, products, personas, ICP, and an ecosystem map | The brand's identity. Everything below reads from it. |
| 2 | Competitors | A record per rival: tier, a monthly visibility snapshot, an optional reputation brief | You can't position against rivals the system hasn't met. |
| 3 | Taxonomy | The categories you classify work with, and the values inside them | The vocabulary the next two surfaces get organized by. |
| 4 | Content Clusters | The strategic territories you're betting on, each tied to pages and briefs | Clusters need the taxonomy and competitors in place first. |
| 5 | Writing Calibration | Reference writers, a tiered source hierarchy, tone, example pieces | Last, because the writing agent reads everything above it. |
Work top to bottom. Each surface gives the one below it something to build on, and Writing Calibration sits last on purpose: the agent that compiles your voice reads your foundation docs, your competitors, your taxonomy, and your clusters before it writes a word.
Two distinctions worth getting straight
Two pairs inside Context trip people up. Keep them apart and calibration gets simpler.
Personas and ICP. Personas describe the people you sell to, the roles and the humans in them. The ICP describes the companies those people work at, the firmographics. Together they're the full target. You need both because a message that lands with a VP of Engineering at a 50-person startup misses the same role at a 5,000-person enterprise.
Writing Calibration and Writing Profile. Writing Calibration is the input you tune: reference writers, a tiered source hierarchy, tone, example pieces, and how you describe what you sell. The Writing Profile is the compiled output the pipeline writes from. You tune calibration; the pipeline reads the Profile. The auto-pilot seeds the Profile from your foundation docs so you're never starting from blank, and it's opinionated by design, because neutral defaults produce beige content. A profile that takes a stance writes pages worth reading.
Preview your voice before you commit
Before you run a full Page Brief, have the writing agent draft a short sample article from your Writing Profile. Read it the way a buyer would. Does it sound like you? Does it take the positions you'd take? If the voice is off, the fix is in calibration, not in editing the draft. Adjust the reference writers, the tone, and the source hierarchy, then preview again. Tuning the input once saves you editing every page that comes out of it.
A working order for calibration
Review and correct the foundation docs the setup agents drafted. This is the source everything else reads from, so it's worth the most attention.
Confirm your competitors and their tiers. Add the rivals the agent missed.
Approve the taxonomy, renaming categories to match how your team talks about the work.
Name your Content Clusters, the territories you're choosing to compete in.
Tune Writing Calibration, then preview a sample article before you commit to a full piece.
Where to go next
Last updated at June 3, 2026
What happens on day one
Setup agents research your competitors, crawl your site, draw personas from real data, map your taxonomy, and calibrate a writing agent, all before your first call.
Connect your data sources
GrowthOS reads from five inputs: a site crawl, analytics, search console, keyword data, and CheckThat. Connect them and the scoring, Health, and AI visibility have something to read.