marketing-positioning

Positioning

One workflow: five steps from the buyer's alternatives to one sentence, written to a positioning record. Method: April Dunford, Obviously Awesome — alternatives first, the sentence last. A template filled in from memory produces filler; that is the failure this skill exists to prevent.

Record schema, scope limits, body format, and the resolution order: docs/WORKSPACE.md, "Positioning record". Rules: "Positioning" in AGENTS.md.

When to Use

• No record in workspace/firm/positioning/ yet, or only the prose in firm/profile.md
• A service or department needs its own statement — a firm that sells five things rarely has one sentence true for all five
• After several won or lost outcomes, when the alternatives buyers named have changed
• firm-context setup finished and the profile has no worked positioning

Prerequisites

Brand scope. In a multi-brand workspace (2+ brand homes in workspace/firm/brands/), every path this skill reads or writes is brand-scoped: definitional entities (services, ICP, personas, proof, channels, positioning) live in firm/brands/{brand}/…, working pipelines carry a {brand}/ segment under their entity root, and the working brand comes from the user's choice or B2BFORCE_BRAND — never guessed. Rules: Brand Scope Gate in AGENTS.md; paths: docs/WORKSPACE.md.

1. Firm Context Gate passes: workspace/firm/profile.md and at least one service.
2. The services in scope exist and declare target_icps. If the ICP is missing, run marketing-icp first — never write an ICP inline here.
3. Read, for the services in scope:
   • sales/opportunities/*--{service}--*/outcome.md — competitor, reason_stated, reason_assessed
   • sales/opportunities/*--{service}--*/!_discovery.md — the situation and the buyer's words about what they do today
   • firm/proof/ — records whose service is in scope
   • intelligence/competitors/*/!_profile.md
   • existing records in firm/positioning/
4. Never read examples/ as evidence (Firm Context Gate).

Workflow

0. Choose the scope

Ask which level, unless the user said:

With several services and no records yet, propose a grouping before writing anything: services that share an ICP in target_icps and the same alternatives in their outcomes are one department candidate; services that share neither get their own records. Recommend all.md only when one buyer and one difference really hold across the services. Otherwise say no firm-wide statement is honest yet.

If a broader record covers the scope (a department or all.md above a service), show it. Ask whether the new statement narrows it or deliberately differs. The alternative for one service is often not the firm's alternative, and that is fine.

1. List the alternatives

What did buyers do before buying, or instead of buying, from the firm? Typical answers: their own team, one person "doing marketing", a freelancer, another agency, the vendor's services arm, a hire, or nothing, meaning they kept the current way. "Nothing" is always a candidate.

• Count each alternative across the outcomes and discovery notes in scope, and list the opportunity slugs.
• With no opportunities yet, interview: the last five won and lost deals, what each buyer did instead or before.
• Phantom competitors: a monitored competitor that never appears in any outcome in scope may be one the firm never actually loses to. Ask before listing it. Do not list it silently, and do not drop it silently.

2. List the differences

For each alternative: what does the firm have that this alternative lacks, and what does that give the client? Take 2–3 candidates from the services' Differentiators and the interview, not from website copy. Also name what the firm sells in words the buyer would use; this becomes the category in the sentence.

For a department, keep only differences true for every service in it.

3. Attach proof to each difference

Search firm/proof/ (Proof Gate in AGENTS.md). Each difference gets a record, or it moves to ## Dropped differences with proof_needed and what would prove it. Never estimate, illustrate, or borrow a result.

• A record with usable_publicly: false may support the statement, which is internal. Say that a public headline cannot lean on it.
• Multi-brand: a sibling brand's record only via {brand}/{slug} with cross_brand: true, attributed to the delivering brand.

4. Name for whom and in what situation

From the differences that survived step 3, pick the client type who cares most, the moment they look for help, and the alternative from step 1 that this client considers most often. icp must be an existing ICP in target_icps of every service in scope. If it is not, the scope is too wide: split it.

5. Write the sentence

For [whom], who [situation] and today [alternative], we are [difference], as shown by [proof].

Use content_language, one client group, and one or two differences. If the sentence needs more to make sense, go back to step 4 and narrow the group.

6. Write and verify

1. Write the record (schema: docs/WORKSPACE.md). Use status: draft until the user confirms the sentence, then set active.

2. Writing all.md replaces the old source in the same change. In single-brand, the profile's ## Positioning becomes one line pointing at firm/positioning/all.md. In multi-brand, remove the brand's positioning: field.

3. Run:

   scripts/validate-positioning.sh
   

   A non-zero exit is a hard stop.

Output

Multi-brand: workspace/firm/brands/{brand}/positioning/…, same filenames.

Report in chat: the sentence, the dropped differences with what would prove them, and which proof records are internal-only.

Quality checks

• Every alternative has its source: opportunity slugs, or "interview" and a date
• Phantom competitors were asked about, not assumed
• Every difference in ## Parts has a Proof N row with a real record
• The statement names one client group and at most two differences
• No second source left: profile prose or the brand's positioning: field, if all.md was written
• scripts/validate-positioning.sh passes

Related Skills