Skip to content

Decides

Positioning

Works out what buyers compare you with, what only you have, who cares most and the category you win in, from your won and lost deals.

Other tools

Add this skill to Claude

Claude takes skills as a ZIP file. Your download has started; if it didn't, use the button below.

  1. 1In Claude, open Customize, then Skills.
  2. 2Choose +, then Upload a skill, and pick positioning.zip.
  3. 3Ask for it by name, or let Claude use it when a task matches.

Skills need code execution turned on in Claude's settings. On a Team or Enterprise plan, an admin can share it with everyone.

Download the folder

Try it now, without installing

See the skill work on a made-up input before you add it. Nothing is installed, and you can swap in your own input.

Open in the Claude app

Opens a new chat in the Claude desktop app with the prompt filled in. Nothing is sent until you press enter.

Includes the whole SKILL.md and the sample input. Paste it into Claude, ChatGPT or any chat.

Influenced by
The positioning method from Obviously Awesome
Needs
Won and lost deals from the last 12 months · Your product facts
Works with
Claude, ChatGPT, Codex, Cursor and any agent that reads SKILL.md
Version
1.0.0 · Last checked · License MIT
Maintained by
Sabih AhmedCreative
SkillStrategy and messaging

Positioning

Example output

Test each attribute against the main alternative
AttributeProofCall
One calendar for every site7 accounts, product docsLead
Online bookingEvery alternative has itTable stakes
Most flexible schedulerNo sourceCut

Works out what buyers compare you with, what only you have, who cares most and the category you win in, from your won and lost deals.

The skill

The whole skill, in plain text.

Nothing is hidden. Read it, change it to fit your team, or take the whole folder.

Scripts
Optional Python scripts for the counting. It works without them and says which way it ran.
Network
Opens web pages or your tools if your agent can. Works from pasted data if not.

3 test cases

positioning/SKILL.md

In the folder

  • positioning/SKILL.md
  • positioning/references/evidence-rules.md
  • positioning/scripts/deal_patterns.py
---name: positioningdescription: >-  Decides what a B2B SaaS company is compared to, what only it has, who cares most and the market category it should  be judged in, from its own won and lost deals and buyer evidence. Every attribute needs proof and every segment  needs a win-rate pattern, so a crowded market gets a position the evidence supports. Returns NEW, KEEP, REVISE,  NEEDS DECISION, NEEDS EVIDENCE or BLOCKED and writes the result into the profile's Positioning section. Use it  before a messaging house, a homepage rewrite or a category decision, or when win rates shift. Not for writing  taglines or pillars (use messaging-house) or for finding the moments that start a search (use buying-moments).license: MITcompatibility: >-  Works in Claude, ChatGPT, Codex, Cursor and any agent that reads SKILL.md. Uses a connected CRM or call recorder if  the agent has one, otherwise exports or pasted deal notes. Python 3 is optional, for the deal counts.metadata:  author: SHAKEUP  version: "1.0.0"  stage: decide  plugin: strategy-messaging  homepage: https://www.shakeuphq.com/skills/positioning--- # Positioning Finds the frame a B2B SaaS company wins in, from the deals it actually won and lost, so every message after it stands on proof. ## Purpose - One job: decide the 5 positioning fields (competitive alternatives, unique attributes, value, best-fit customers, market category) from evidence, and write them to `## Positioning` in `company-profile.md`.- Stage: decide. It ends in a verdict.- It never invents a win rate, a customer, a competitor fact or a proof. It never writes copy: headlines, taglines and pillars belong to `messaging-house`.- It never picks a category because it sounds bigger or newer. The category is the one the evidence says the company wins in. ## Use when / not | Situation | Use || --- | --- || No positioning yet, or it was written from opinion | this skill || Win rates or deal mix changed, a new competitor shows up in deals, or the product changed | this skill, to confirm or revise || A messaging house or homepage rewrite is next | this skill first, then `messaging-house` || You need the ICP, the buying group and the moments that start a search | `buying-moments` (run after this skill; it reads best-fit customers) || You need copy tested on buyers | `message-test` || Filling the rest of the profile | `company-profile` | ## Inputs | Input | Required | Ways to provide, in order | If missing || --- | --- | --- | --- || Won and lost deals, last 12 months: outcome, what they compared or lost to, why, segment fields, deal size, cycle | yes (won deals at least) | a connected CRM (closed-won, closed-lost, loss reason, competitor field); a CRM export (CSV); a pasted table; the reader's account of the last 10 deals, labelled `recalled` | `NEEDS EVIDENCE`, with the exact report to pull || Buyer evidence: win and loss interviews, call transcripts, sales notes, reviews | no | a connected call recorder or CRM notes; uploaded transcripts; pasted excerpts | value stays `inferred`; coverage `partial` || Competitor facts (what each alternative does and lacks) | no | the competitors' own pages and docs, if the agent can browse; pasted pages | uniqueness stays `unverified` against that alternative || Product facts | yes | `## Product` in the profile; the product site; pasted docs | ask once | ## Profile read From `company-profile.md`: `## Company` (category, business model, ACV, sales cycle), `## Product`, `## Buyers` (ideal customer, bad fit), `## Positioning` (the current one, if any), `## Claims and proof`, `### Never say`, `## Competitors` (mention policy and how buyers compare), and every `## Learnings` entry for `positioning` with status `noted`, `fact` or `accepted`. No profile: ask in one message for what the product does, who it sells to, and the deal list (or the last 10 deals from memory). Say that `company-profile` saves this step next time. ## Steps 1. **Apply learnings.** Read the learnings for this skill first. An accepted learning overrides the default rule it names; say so in the output when it fires. 2. **Build the evidence table.** Give every item an id: `D1` for deals, `I1` interviews, `C1` call notes, `R1` reviews, `N1` sales notes. For each deal record: outcome, alternatives (compared, used before, lost to), reason tags, segment fields (industry, size band, stack, region, sales motion), ACV, cycle days, the role that led the buy.   - A `no decision` is a loss to the status quo. Count it as a loss; if it names no alternative, record `status quo` as its alternative.   - A customer that churned in its first year counts against fit for its segment, not in the win rate.   - A `recalled` deal counts, but a pattern that rests only on recalled deals is a hypothesis.    Then run the counts, naming each segment column as it appears in the reader's headers:    ```   python3 scripts/deal_patterns.py --csv deals.csv --segment employees --segment crm   ```    Read the `warnings` first. A missing outcome, alternative or reasons column means the counts are wrong: rename that column (`outcome`, `alternative`, `reasons`) and run again.    Without Python, count by hand: win rate = won / (won + lost + no decision); per alternative, the deals it appears in and the outcome of each; per segment value, deals, wins, win rate and one deal's swing (100 / deals, in points); per reason tag, the distinct accounts that gave it. Say which way you ran it. 3. **Grade the evidence.**    | Evidence | complete | partial | too thin |   | --- | --- | --- | --- |   | Won deals with a reason | 10 or more | 5 to 9 | under 5: with fewer, any 3-deal pattern is most of the sample, so it can't separate a segment from the whole base |   | Lost or no-decision deals | 5 or more | 1 to 4 | 0: alternatives come from won deals only, so the loss side is blind |   | Buyer words (interviews, transcripts, reviews) | 5 or more sources | 1 to 4 | 0: value stays `inferred` |    Under 5 won deals ends at `NEEDS EVIDENCE` after step 9: finish the draft as a labelled hypothesis, but don't write it to the profile. 4. **Name the competitive alternatives.** What would buyers use if the company didn't exist? Take them from the deals, not from the competitor list.   - Always test the status quo forms: a spreadsheet or manual process, a feature of a tool they already pay for, hiring a person or an agency, doing nothing.   - Keep an alternative that appears in 2 or more deals. Rank by appearances across won, lost and no-decision deals. The main alternative is the one that appears most often; on a tie, the one buyers chose in more lost and no-decision deals.   - With 5 or more deals against an alternative, give the win rate against it. Below 5, write the count only.   - Mark the market `crowded` when 3 or more products from other companies each appear in 2 or more deals. A feature inside a tool the buyer already pays for counts as one of those products. Step 8 then applies. 5. **Test the unique attributes.** Candidates come from `## Product`, won-deal reasons, and what buyers said the alternatives lacked. Different must pass against the main alternative; against the next 2 alternatives, record pass or fail, because it shows which comparisons the attribute wins:    | Test | Passes when | If it fails |   | --- | --- | --- |   | Different | the main alternative lacks it or does it measurably worse, shown by a competitor source or deal notes | table stakes: list it, never lead with it |   | Proven | it has proof a buyer could check, from the first 4 kinds in `references/evidence-rules.md` (product fact, customer data with a source, a cleared customer, a third party). Deal reasons prove that it mattered, not that it's true | `claim to confirm`: keep it out until a source exists |   | Mattered | it appears in won-deal reasons or buyer words from 3 or more different accounts | `unvalidated`: it can't carry the position |   | Lasting | the alternative can't match it inside one sales cycle with a setting or a price cut | keep it, flagged `fragile` |    A unique attribute passes Different, Proven and Mattered. Lasting only flags risk.    Pick the lead attribute: one that passes all 4 tests (if none passes Lasting, the strongest of the rest, flagged `fragile`). On a tie, lead with the one named in more deals won against the main alternative, then the one named by more accounts. 6. **Write the value.** For each unique attribute, chain it: attribute, then what it lets the buyer do, then the result the buyer measures, in the buyer's words where you have them. Value needs buyer evidence (interviews, transcripts, reviews, or deal notes in the buyer's words) from 2 or more sources; with less, mark it `inferred`. An inferred value doesn't block a verdict; it makes coverage `partial`. Group the chains into 2 or 3 value themes. Each value line ends with its proof id. 7. **Find the best-fit customers.** Who cares most about that value?   - Candidates: segment values with 5 or more decided deals whose win rate beats the overall rate by more than one deal's swing. The rule stops 1 lucky deal from defining the segment.   - Rank by lift, then by median ACV of won deals, then by shorter median cycle.   - Describe the segment by what makes it care, not only by firmographics: the situation, the trigger and the value theme that wins it ("clinic groups with 5 or more sites whose managers run a calendar per site"). Made-up example, not a real company.   - Candidates from different segment fields that share half or more of the smaller one's deals (say "5 or more sites" and "on a shared practice system") describe one group. The larger is the best fit; the shared deals are its core, reported with their own counts.   - Add the bad-fit pattern: segment values with churn, or with a win rate below overall by more than one deal's swing.   - Compare with `## Buyers` Ideal customer. A best fit inside it is a narrowing: note it for `buying-moments`. A best fit outside it (most of its won deals fall outside the stated ideal customer) is a conflict: verdict `NEEDS DECISION`.   - 2 candidates that share fewer than half their deals and sit within one swing of each other tie: verdict `NEEDS DECISION`. 8. **Apply the crowded-market rules** when step 4 marked the market crowded:   - Position against the alternative buyers choose most often, which is often the status quo, not the loudest competitor.   - Lead with the lead attribute from step 5. Move every shared attribute out of the lead.   - Narrow the best-fit segment until its lift clears the swing rule. A narrower position that wins beats a broad one that ties.   - Every comparative word in the fields ("only", "faster", "more") needs a proof that compares against a named alternative.   - If no attribute passes, the company has a hypothesis, not a position: `NEEDS EVIDENCE`, with the evidence that would prove one. 9. **Choose the market category.** First match wins:    | Condition | Frame |   | --- | --- |   | Buyers in won deals already name a category, and the win rate against that category's main vendors is at or above overall | Head to head in that category |   | A known category exists, but wins concentrate in the best-fit segment for a reason that segment states | That category, narrowed: "<category> for <segment>" |   | Buyers place the product in a category whose default buying criteria it loses on (the loss reasons are that category's checklist) | An adjacent category whose default criteria match the unique attributes |   | No category in the buyers' words fits, and 3 or more buyers couldn't say what it was | A new category: `NEEDS DECISION`, because the company would fund the category's education, which takes longer and costs more than any other frame |    Then test the chosen frame: buyers use the term (interviews, review-site categories, the reader's search data); the frame's default expectations match the unique attributes; the main alternative looks weaker or out of place inside it. A frame that fails 2 of the 3 goes back to the table. A test with no evidence either way is `untested`: it isn't a fail, but say so and mark coverage `partial`. Details and examples: `references/evidence-rules.md`. 10. **Add a market change only with a source.** Include a trend only when a source the reader can open shows it and it makes the value more urgent for the best-fit segment. Otherwise leave it out. 11. **Draft the section and compare.** Fill the 5 fields exactly as the profile template names them. Write the one-line statement `messaging-house` builds on: "For <best fit> who <problem>, <company> is the <category> that <value>. Unlike <main alternative>, it <lead attribute>." If the profile already has positioning, compare field by field: same, sharpened (same meaning, better evidence) or contradicted. 12. **Write it back.** After the verdict, per the stops below, replace `## Positioning` with the new fields. Each unique attribute and value carries its proof in brackets. Attributes that failed Proven go to open items for `company-profile` to record as claims to confirm, never into the section. If the agent can't edit the profile file (a chat app, or a profile pasted into the chat), give the section in one code block ready to paste and set `profile_section_written: false (ready to paste)`. ## Check and verdict | Verdict | Condition (first match wins) || --- | --- || `BLOCKED` | No product description, or no deal or buyer evidence of any kind after asking once || `NEEDS EVIDENCE` | Fewer than 5 won deals; or no attribute passes Different, Proven and Mattered || `NEEDS DECISION` | The evidence-based best fit conflicts with the profile's ideal customer; or 2 segments or 2 frames tie; or the frame would be a new category || `REVISE` | The profile has positioning and the evidence contradicts at least 1 field || `KEEP` | The profile has positioning and every field holds against the evidence (wording may sharpen) || `NEW` | The profile had no positioning; every field is filled from evidence | Named stops: - `BLOCKED`: ask for the missing input, naming the CRM report to pull (closed-won and closed-lost, last 12 months, with loss reason and competitor).- `NEEDS EVIDENCE`: stop. Show the draft labelled `hypothesis`, the tests it failed, and the evidence to collect (which deals, which interviews). Don't write the profile.- `NEEDS DECISION`: stop. Show the 2 options side by side with the deals behind each, what each wins and what each gives up. Write nothing until the reader picks.- `REVISE`: stop before overwriting. Show the old and new field by field with the evidence; write after the reader says yes. `NEW` and `KEEP` write without a stop. ## Learn Append to `## Learnings` in the profile: - the reader picks the other option at `NEEDS DECISION`, or rejects a `REVISE`: a `proposal` with their reason;- the positioning is set or changes (including a first `NEW`): a `fact` with the old (or none) and new category and best fit, so later runs know what changed and why;- a result comes back, such as the win rate in the best-fit segment 2 quarters later, or `message-test` or a buyer panel showing buyers don't recognise the category: a `fact` with its source. Example: `- 2026-10-24 | positioning | fact | Best fit narrowed from all clinics to groups with 5 or more sites | Why: win rate 82% vs 61% overall on 11 deals (reader's CRM, 2026-Q3) | Status: noted` ## Output ```Verdict: <NEW | KEEP | REVISE | NEEDS DECISION | NEEDS EVIDENCE | BLOCKED> · Coverage: <complete | partial (what's missing) | blocked> · Counts: <script | by hand>Evidence: <n> deals (<won> won, <lost> lost, <n> no decision, <n> churned), <n> buyer sources · Overall win rate: <n>% · Market: <crowded | not crowded> Competitive alternatives| Alternative | Deals | Won against | Lost to | Win rate against | Note | Unique attributes| Attribute | Different | Proven (proof id) | Mattered (accounts) | Lasting | Call | Value| Value theme | Attribute it rests on | Result the buyer measures | Proof | Best-fit customers| Segment | Deals | Win rate | Lift | One deal's swing | Median ACV won | Why they care |Bad fit: <segments and the evidence> Market category: <frame> because <the rule that fired>. Tests: buyers use it <yes/no>, expectations match <yes/no>, alternative weaker inside it <yes/no> Statement: For <best fit> who <problem>, <company> is the <category> that <value>. Unlike <main alternative>, it <lead attribute>. ## Positioning   (the section as written to the profile, or "not written: <stop>")Competitive alternatives: ...Unique attributes: ...Value: ...Best-fit customers: ...Market category: ... Changed from the profile: <field: old, new> (or "none" or "no previous positioning")Learnings applied: <list, or none>``` Then the handoff block. ## Hand off ```handoffskill: positioningversion: 1.0.0verdict: NEWcoverage: partial (no loss interviews; competitor docs not opened for 1 alternative)for: <company>outputs:  main_alternative: a spreadsheet the operations team keeps  lead_attribute: one calendar across every site  best_fit: clinic groups with 5 or more sites  market_category: scheduling software for multi-site clinic groups  statement: For clinic groups with 5 or more sites who run a calendar per site, ...  profile_section_written: truenext: [buying-moments, messaging-house]open: [a source for "most flexible", or drop it]``` Made-up values for illustration. For `NEEDS EVIDENCE` and `NEEDS DECISION`, `profile_section_written` is false and `next` is `[]` until the reader acts. `messaging-house` reads `statement`, `lead_attribute` and the written section; `buying-moments` reads `best_fit`. ## Failure modes | Problem | Do this || --- | --- || The CRM has no loss reason or competitor field | Use call notes or ask the sales lead for the last 10 losses; mark those deals `recalled` || Every loss reason is "price" | Price is often the stated reason for a weak value case. Read the notes for what the buyer compared; if they compared a cheaper category (the status quo), the frame is the problem, not the price || The reader wants an attribute in that has no proof | Keep it out of the section; list it under open items as a claim to confirm, with the proof that would let it in || The founder rejects the result after seeing the deal evidence | The deals win. Show the gap with the deal ids; if the founder still disagrees, stop at `NEEDS DECISION`. A request to add an unproven claim is the next row, not this one || One big customer dominates the wins | Count accounts, not deals or revenue, for every pattern; note the account's share of won ACV || The reader asks for "the leading" or "the best" in the category | Not a position: there is no proof that compares. Use the lead attribute against the main alternative instead || Only won deals exist (a new company) | Alternatives come from what won customers used before; mark the loss side blind and coverage `partial` || The reader asks to write without waiting | Named stops hold. Show exactly what you would write and the choice it needs, then wait; don't pick for them | ## Influenced by - April Dunford's positioning method from Obviously Awesome, where competitive alternatives, unique attributes, value, best-fit customers and the market category are worked out in that order: https://miro.com/templates/april-dunfords-positioning-template/ and https://www.aprildunford.com/books- What we changed: every field must come from won and lost deals, an attribute must pass 4 tests (different, proven, mattered, lasting), a segment must beat the overall win rate by more than one deal's swing, crowded markets get their own rules, and the result is written into the company profile for the next skills to read. We don't use her canvas or wording.

How it thinks

Inputs in, a verdict out, and a stop only where you decide.

What we built

Every field must come from won and lost deals, and every attribute needs proof before it can lead. We added a win-rate test so a lucky deal can't define your best fit, and rules for crowded markets where most features are shared.

It reads your company profile if you have one, or asks for only what it needs. Then it works through its steps, checks its own work and ends in a verdict.

The steps

  1. 01

    Read your learnings first

    Rules your team accepted for this skill change the defaults before it starts.

  2. 02

    Count the deals

    It counts every deal by outcome, alternative, reason and segment, and treats a stalled deal as a loss to the status quo. A script does the counts, or the skill counts without it and says so.

  3. 03

    Name what buyers compare you with

    Alternatives come from your deals, not your competitor list. A spreadsheet or doing nothing can be the main one.

  4. 04

    Test every attribute 4 ways

    Different from the main alternative, proven by a source, named in deals from 3 or more accounts, and hard to copy within one sales cycle. Only attributes that pass the first 3 can lead.

  5. 05

    Find who cares most

    A segment counts only when its win rate beats your overall rate by more than one deal's swing, so 1 lucky deal can't define your best fit.

  6. 06

    Pick the frame you win in

    Head to head, narrowed to a segment, an adjacent category or, rarely, a new one. A new category always stops for your decision.

  7. 07

    Write it to your profile

    The 5 fields go into your profile's positioning section, each with its proof, ready for the messaging house.

What you need

  • Required

    Won and lost deals from the last 12 months

    A connected CRM with loss reasons and a competitor field → A CRM export (CSV) → A table you paste → Your account of the last 10 deals, marked as recalled

  • Optional

    Buyer evidence, such as win and loss interviews, call notes and reviews

    A connected call recorder or CRM notes → Transcripts you upload → Excerpts you paste

  • Optional

    What each alternative does and lacks

    Their own pages and docs, if your agent can browse → Pages you paste

  • Required

    Your product facts

    The product section of your company profile → Your product site → Docs you paste

What it can return

  • NEW

    Your profile had no positioning, and every field now comes from evidence.

  • KEEP

    Your current positioning holds up against the deals.

  • REVISE

    The deals contradict at least 1 field. It shows old and new before writing.

  • NEEDS DECISION

    The best fit sits outside your stated ideal customer, 2 options tie, or the frame would be a new category.

  • NEEDS EVIDENCE

    There are fewer than 5 won deals, or no attribute passes the tests. You get a labelled hypothesis and a list of what to collect.

  • BLOCKED

    There's no product description or no deal evidence at all.

Where it stops for you

  • When there's no deal or buyer evidence after one question
  • When the evidence is too thin, with a list of what to collect
  • When the best fit sits outside your stated ideal customer, or 2 options tie
  • Before it overwrites positioning you already have

Example output

What you get back.

A run for a made-up company. Your output depends on your inputs.

Here's a run for a made-up B2B SaaS company that sells scheduling software to clinic groups. Its profile had no positioning yet.

Verdict

NEW

Coverage: partial (no loss interviews; 1 alternative's docs not opened)

Evidence

23 deals (14 won, 7 lost, 2 stalled) and 6 buyer interviews. The overall win rate is 61% (14 of 23). It's a crowded market, with 3 vendors in 2 or more deals each.

AlternativeDealsWon againstLost or stalled
A spreadsheet the operations team keeps972
The booking feature in their practice software835
Scheduling vendor A532
AttributeDifferentProvenMatteredCall
One calendar across every siteThe booking feature works per siteProduct docs7 accountsLead
Moves staff between sites in the rotaVendor A's docs show it can'tProduct docs4 accountsUnique
Online bookingEvery alternative has itProduct docs9 accountsTable stakes
The most flexible schedulerNot shownNo source0 accountsCut, claim to confirm
SegmentDealsWin rateLiftOne deal's swing
Groups with 5 or more sites1182% (9 of 11)+21 points9 points
Single-site clinics729% (2 of 7)-32 points14 points
Market category

scheduling software for multi-site clinic groups. Wins concentrate in groups with 5 or more sites, for a reason those buyers state: one calendar for every site.

Statement

For clinic groups with 5 or more sites who run a calendar per site, the company is the scheduling software that runs every site from one calendar. Unlike the operations team's spreadsheet, it moves staff between sites inside the rota.

Bad fit

single-site clinics, who chose the booking feature they already pay for in 5 of 7 deals.

The handoff block the next skill reads
skill: positioning
version: 1.0.0
verdict: NEW
coverage: partial (no loss interviews; 1 alternative's docs not opened)
for: clinic scheduling company
outputs:
  main_alternative: a spreadsheet the operations team keeps
  lead_attribute: one calendar across every site
  best_fit: clinic groups with 5 or more sites
  market_category: scheduling software for multi-site clinic groups
  profile_section_written: true
next: [buying-moments, messaging-house]
open: [a source for "most flexible", or drop it]

What good looks like

  • Alternatives taken from your deals, including the status quo
  • Every unique attribute with its proof beside it
  • A best fit with deal counts, not just a percentage
  • A category chosen by a stated rule, with its 3 tests
  • A clear stop when the deals and your stated ideal customer disagree

Install

Use it in the tool you already have.

Pick your tool. Each tab has the shortest path we know of.

  1. 01
    Download positioning.zip.
  2. 02
    In Claude, open Customize, then Skills.
  3. 03
    Choose +, then Upload a skill, and pick the ZIP. Code execution must be on in your settings.
  4. 04
    Ask for it by name, or let Claude use it when a task matches.

Questions

Before you add it.

How many deals does it need?

At least 5 won deals to run, and 10 won with 5 lost or stalled for a complete result. With fewer, it gives you a labelled hypothesis and the evidence to collect, and it leaves your profile alone.

Why does a stalled deal count as a loss?

Because the buyer chose the status quo. Leaving stalled deals out hides the alternative that often wins most, which is doing nothing or keeping a spreadsheet.

Will it use our founder's claims?

Only with proof. An attribute needs a product fact, a sourced customer result, a third party or deal reasons from 3 or more accounts. Anything else goes on a list of claims to confirm.

Does it write our homepage?

No. It decides the positioning and writes it to your profile. The messaging house skill turns it into a core message and pillars, and copy comes after that.