> ## Documentation Index
> Fetch the complete documentation index at: https://docs.granvl.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Strategy fields
> Every field on every Strategy record: what it asks for, why it matters, and an example.
The [Strategy concept page](/concepts/strategy) covers the shape. This is the field reference: what each one is asking for, and what a good answer looks like.
Every field is asking one question. **What do you actually know about this, and how would the reader say it themselves?**
Examples below are illustrative, built around a home-warranty offer. They are not from any live account.
## Build Strategy in this order
The five tabs form one chain. Build from the thing you sell toward the copy a visitor sees. Each step gives the next one something concrete to attach to.
**What to do.** Open **Strategy → Products → New product** and name what you sell. Add its current capabilities, planned capabilities, limitations, and qualified outcome when you have them.
**Why it matters.** The Product is the root of the tree. Every ICP, Angle, Research entry, and copy line needs something real to point back to.
**What to do.** Open **Strategy → ICPs → New ICP** and describe the person in terms of goals, triggers, beliefs, objections, and language rather than only demographics.
**Why it matters.** Copy briefs are built for a named audience. Without an ICP, the agent has no specific reader to write for, and copy aimed at everyone reads as if it is aimed at no one.
**What to do.** Open **Strategy → Offers/Angles → New angle** and connect a Product to an ICP with one persuasive argument. State the problem, desired outcome, mechanism, proof, qualification, and disqualifiers.
**Why it matters.** The Angle is the thesis a page or ad argues. Two angles with different mechanisms create a useful test; two rewrites of the same claim usually do not.
**What to do.** Open **Strategy → Research → New research** and save the source, evidence, and takeaways. Link the entry to the Product, ICP, or Angle it supports.
**Why it matters.** Research is where a claim earns the right to be made. It preserves the original evidence and connects a strategy decision to an actual receipt.
**What to do.** Open **Strategy → Copywriting** to review headlines, subheadlines, and ad lines written by an agent. Filter by Product, ICP, Angle, status, kind, or surface.
**Why it matters.** Saving copy makes the library compound. A proven line can be adapted, a weak line can be archived, and the rationale preserves why each line was written.
There is no blank “New copy” form. Lines are generated by an agent and save as `testing`. Only warehouse-backed experiment results can promote a line to `winner`.
The sections below are the field-by-field reference for each step.
## Products
**Strategy → Products → New product.** The root of the tree. Every ICP, angle, and copy line traces back to one.
| Field | What it asks for | Example |
| ------------------------ | -------------------------------------------------- | ------------------------------------------------------------------------------------------------ |
| **Name** (required) | A short label for what you sell | "Home Warranty Plans" |
| **Description** | What it is and who it serves, in a sentence or two | "Annual coverage for home systems and appliances, for homeowners past their builder's warranty." |
| **Destination URL** | Where a visitor lands to buy or learn more | The offer page this product sells |
| **Current capabilities** | What the product does **today** | "Covers HVAC, plumbing, electrical, and six named appliances." |
| **Planned capabilities** | What is coming but has not shipped | "Roof coverage, Q4." |
| **Limitations** | What it does not do, honestly | "No coverage for pre-existing failures or cosmetic damage." |
| **Qualified outcome** | What counts as a qualified lead or purchase | "Homeowner, property over 5 years old, not currently covered." |
| **Status** | `draft`, `active`, or `archived` | |
Current and planned capabilities are never merged when a brief is built. This is the mechanism that stops copy promising a roadmap item as a shipped fact. Put a future feature in **planned**, not in **current**, even when it is two weeks out.
## ICPs
**Strategy → ICPs → New ICP.** Every field here is a question about a real person, not a demographic bucket.
### Who they are
| Field | What it asks for | Example |
| ------------------- | -------------------------------------------------- | ----------------------------------------------------------------- |
| **Name** | One line, plain language, who this is | "Renewal-deadline procrastinators" |
| **Summary** | A one-line read of this person | "Knows the renewal is coming, keeps not dealing with it." |
| **Primary goal** | What they are actually trying to accomplish | "Compare options and pick one before the renewal deadline hits." |
| **Buying trigger** | The specific event that turns interest into action | "Renewal notice just arrived in their inbox." |
| **Current state** | Where they are today, before your product | "Auto-renewing on a plan they never compared." |
| **Desired outcome** | The outcome they are actually buying | "Confidence they are not overpaying, without a research project." |
### What stops them
| Field | What it asks for | Example |
| ------------------------------------- | ---------------------------------------------------- | ----------------------------------------------------------------- |
| **Customer beliefs** (audience truth) | What they believe, true or not, in their terms | "All providers are basically the same, so price is what matters." |
| **Core objections** | Reasons they talk themselves out of it, one per line | "Every provider claims to be cheapest. Why believe this one?" |
| **Fears** | What they are afraid of, one per line | "Signing up and finding the thing that breaks is excluded." |
| **Language to mirror** | Their words, one per line | "Didn't realize it auto-renewed until the charge hit." |
| **Evidence read** | What the two lists above rest on, cited | Links or references to the research behind them |
### What wins them
| Field | What it asks for | Example |
| ------------------ | ------------------------------------------- | ------------------------------------------------- |
| **Offer** | The promise currently pointed at this ICP | "Lock in your current rate before renewal." |
| **Role in market** | What that promise competes against | "Versus doing nothing and auto-renewing." |
| **Use cases** | Concrete situations where they would use it | "30 days out from renewal, comparing two quotes." |
### Status
| Field | What it asks for | Example |
| -------------------- | ----------------------------------------------------------- | -------------------------------------------------------------- |
| **Current decision** | Keep, expand, or retire, and why | "Expand. Best conversion rate of any ICP in the last 30 days." |
| **Confidence** | How well evidenced this is: `strong`, `moderate`, `limited` | |
| **Status** | `proposed`, `draft`, `active`, or `archived` | |
| **Notes** | Anything else worth keeping | "Skews older. Avoid overly casual copy." |
| **Product(s)** | Which products this ICP applies to | "Home Warranty Plans" |
It is tempting to fill an ICP in with a job title and an age range. The fields are deliberately shaped around goals, triggers, beliefs, and objections instead. A demographic tells the ad platform who to target. This tells your agent what to say.
## Offers/Angles
**Strategy → Offers/Angles → New angle.** The persuasive thesis: one or more products, one or more ICPs, and one argument.
| Field | What it asks for | Example |
| ----------------------------- | ------------------------------------------------------- | ----------------------------------------------------------------------------------------- |
| **Name** | A short label for the argument | "Beat the auto-renewal" |
| **Problem / pain point** | The specific problem this angle leads with | "Their plan renews automatically at a higher rate if they miss the cancel window." |
| **Desired outcome** | The dream state the promise delivers | "Lock in today's rate with zero risk of an unwanted renewal." |
| **Mechanism** | How the product actually gets them there | "We read your renewal date off the policy and quote against it before the window closes." |
| **Proof** | Evidence the promise holds | "Average \$340 saved across 1,200 switchers last year." |
| **Objection handled** | The objection this angle exists to defuse | "Everyone claims cheapest, so the claim is doing no work." |
| **Qualification** | Who this offer is for | "Homeowners with a renewal inside 60 days." |
| **Disqualifiers** | Who it is not for | "Renters, and anyone mid-term with no renewal in sight." |
| **Incentive** | The sweetener, if any | "Price-match guarantee for 12 months." |
| **Price / terms** | Price and terms as stated to the buyer | |
| **Confidence** | `strong`, `moderate`, or `limited` | |
| **Status** | `proposed`, `draft`, `testing`, `active`, or `archived` | |
| **ICP(s)** and **Product(s)** | What this angle targets and sells | An angle must name at least one of the two |
**Mechanism** is the field that separates a real test from a reworded one. Two angles arguing different mechanisms teach you something when one wins. Two angles rewording the same claim teach you which sentence scanned better.
## Research
**Strategy → Research → New research.** Evidence, not opinion. The thing an angle points back to when someone asks how you know that.
| Field | What it asks for | Example |
| ------------------------- | --------------------------------------------------------------- | --------------------------------------------------------------------------------- |
| **Title** | A short label for this piece of evidence | "Reddit thread: renewal price shock complaints" |
| **Kind** | The category: transcript, competitor, survey, review, and so on | "Transcript" |
| **Source URL** | Where it came from, if anywhere | A link to the original thread or interview |
| **Body** | The actual quotes, findings, or excerpts | "Didn't realize it auto-renewed until the charge hit. Wish there was a reminder." |
| **Takeaways** | The findings, one per entry, written to be read on their own | "Surprise is the emotion, not price. Lead with the date, not the discount." |
| **ICP / Angle / Product** | What this evidence supports. Defaults to unassigned | Linked to the matching ICP |
Takeaways are what a brief leads with; the body is the long-form backing detail. An entry with a body and no takeaways still reads as its body, so nothing is lost by adding evidence fast and summarizing it later.
## Copy lines
**Strategy → Copywriting.** There is no creation form here. Lines are written by your agent, either directly or while building a page. Clicking a line opens its detail panel.
| Field | What it asks for | Example |
| ------------- | --------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------- |
| **Body** | The copy itself | "Lock in your rate before the renewal window closes." |
| **Kind** | `headline`, `subheadline`, `ad_primary`, `ad_headline`, or `ad_description` | |
| **Surface** | Where it runs: `landing` or `ad` | A headline is written differently for a page than for a feed |
| **Status** | `testing`, `winner`, or `archived` | |
| **Source** | `agent`, `human`, or `imported` | |
| **Rationale** | Why this line was written this way | "Mirrors the current ad headline closely to reduce post-click message mismatch while keeping the qualification marker." |
| **Batch** | A free label for lines drafted in one pass | "hero test 3" |
| **Targeting** | The product, ICP, and angle it argues from | The linked triple |
The rationale is the difference between *the agent wrote this* and *the agent wrote this because*. It is what a human reads when deciding whether a line deserves promotion.
Copy is not signed off, it wins a test or it does not. Only warehouse-backed experiment results move a line to `winner`; your agent cannot promote its own work. Archive retired lines rather than deleting them, since the library is also the record of what was tried.
## For your agent
| Tool | What it does |
| -------------------- | --------------------------------------------------------------------------------- |
| `get_strategy_brief` | One bounded product + ICP + angle brief with linked evidence and explicit gaps |
| `get_copy_brief` | The same brief plus existing copy lines and the exact copy gaps for an ad or page |
| `list_copy_lines` | Read the library, filtered by kind, surface, status, angle, or ICP |
| `create_copy_lines` | Save a batch of new lines, each with a rationale |
| `update_copy_line` | Edit one line or change its status |
Products, ICPs, angles, and research are authored in the dashboard, by you or by granvl's [built-in agent](/concepts/built-in-agent), which has its own tools for creating and updating them. Your MCP-connected agent reads them through the brief tools and tells you what is missing.