Skip to content

Decides

Channel fit

Decides which channels your price, deal size, sales cycle and market can pay for, sizes each first test, and splits spend between brand and demand.

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 channel-fit.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
Market, product, channel and model fit, The 95:5 rule
Needs
Business model, deal size or pricing, sales cycle and regions
Works with
Claude, ChatGPT, Codex, Cursor and any agent that reads SKILL.md
Version
1.0.0 · Last checked · License MIT
Maintained by
Sabih AhmedPaid media
SkillStrategy and messaging

Channel fit

Example output

Give each channel a call
ChannelCost against ceilingCall
Google Ads$913 against $1,000Core
Microsoft AdsUntestedTest
Meta AdsBuyers aren't on itSkip

Decides which channels your price, deal size, sales cycle and market can pay for, sizes each first test, and splits spend between brand and demand.

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

channel-fit/SKILL.md

In the folder

  • channel-fit/SKILL.md
  • channel-fit/references/channels.md
  • channel-fit/scripts/channel_math.py
---name: channel-fitdescription: >-  Decides which marketing channels can work for a B2B SaaS company's price point, deal size, sales cycle and  market, and how to split spend between creating demand (brand) and capturing it. Sets cost ceilings from the  profile's own ACV, conversion rates and targets, sizes the smallest honest test per channel, checks each of 11  channels against what it needs to work, and orders the tests. Returns PLAN, MEASURE FIRST, NEEDS DECISION or  BLOCKED. Use it when planning a year or quarter, adding a channel, or when spend is shifting without a reason.  Not for planning one channel in depth (use linkedin-plan, reddit-plan or chatgpt-ads-check) or for setting the  most a lead can cost in detail (use lead-value-math).license: MITcompatibility: >-  Works in Claude, ChatGPT, Codex, Cursor and any agent that reads SKILL.md. Uses connected ad accounts, a CRM or  analytics if the agent has them, otherwise exports or pasted figures. Python 3 is optional, for the unit math.metadata:  author: SHAKEUP  version: "1.0.0"  stage: decide  plugin: strategy-messaging  homepage: https://www.shakeuphq.com/skills/channel-fit--- # Channel fit Works out which channels a company can afford to win in, what each must prove in its first test, and how much goes to buyers who aren't looking yet. ## Purpose - One job: decide the channel set, each channel's call, the order of tests and the brand and demand split for one company.- Stage: decide. It ends in a verdict with the rule that fired.- It never invents a benchmark: no industry cost per click, conversion rate or "typical B2B split". Every threshold comes from the profile (ACV, rates, targets) or the reader's own channel data, with its reason. A missing figure is `not available` with what would provide it.- It doesn't build the campaign. Each channel it picks goes to that channel's skill. ## Use when / not | Situation | Use || --- | --- || Planning the next quarter or year's channel mix | this skill || Deciding whether to add a channel, or which one first | this skill || Asked how much should go to brand and how much to demand | this skill || Setting the most a lead can cost and the smallest campaign budget for one channel | `lead-value-math` || Planning LinkedIn audiences, formats and the split inside LinkedIn | `linkedin-plan` || Deciding whether ChatGPT Ads fit yet | `chatgpt-ads-check` || Leads can't be traced to opportunities | `tracking-audit` first | ## Inputs | Input | Required | Ways to provide, in order | If missing || --- | --- | --- | --- || Business model, ACV (or pricing), sales cycle, regions | yes | `## Company` in the profile; the reader | ask once; `BLOCKED` without ACV or pricing || Lead-to-opportunity and opportunity-to-won rates (or first conversion to paying customer, for product-led) | no | `## Targets`; a connected CRM; a CRM export | ceilings from a payback target only, or `not available`; coverage `partial` || Target cost per lead, or a CAC payback target in months | no | `## Targets`; the reader | no ceilings, so no test floors; coverage `partial` || Monthly budget and the pipeline target | no | `## Channels` and `## Targets`; the reader | skip the target check; coverage `partial` || Channel results, last 90 days: spend, leads, opportunities, customers per channel | no | connected ad accounts plus CRM; exports; pasted figures | every active channel is judged on fit only, marked `untested` || The facts each channel needs (a nameable account list, search demand, site traffic, video, a willing founder, communities, sales follow-up time) | no | the profile; `buying-moments` and `keyword-selection` outputs; the reader | ask the 5 that decide the most channels; the rest stay `unknown` || Market size and how often buyers switch (accounts that fit, years between purchases) | no | the reader | skip the in-market estimate | ## Profile read From `company-profile.md`: `## Company` (business model, pricing, ACV, sales cycle, regions), `## Buyers` (ideal customer, buying group), `## Buying moments` (with what buyers search or ask), `## Positioning` (best fit), `## Channels` (status, budget, owner), `## Targets`, `## Accounts and data`, and every `## Learnings` entry for `channel-fit` with status `noted`, `fact` or `accepted`. No profile: ask in one message for the business model, ACV or pricing, sales cycle, regions, monthly budget and the channels running now with their spend, leads and opportunities. Say that `company-profile` saves this next time. ## Steps 1. **Apply learnings.** Accepted learnings for this skill override the defaults below (for example a test window the reader set). Name each one that fires. 2. **Gather the facts.** Fill the inputs table from the profile and the reader's data. For the channel needs in `references/channels.md`, ask at most 5 questions, choosing the facts that decide the most channels first: can you name your target accounts; do buyers search for the problem or category (evidence, not belief); monthly site visitors; can sales follow up within 2 days; is the founder willing to post in their name. Map the profile's channel names to the 11: a founder's LinkedIn posts are founder-led content, and "Reddit" covers both Reddit rows unless the profile splits them. 3. **Measurement gate.** If channels are active and no lead can be traced to an opportunity or a customer (no CRM link, no rates, no channel-level opportunities), stop at `MEASURE FIRST`: any split would be judged on form fills, which teaches each platform to find more form fillers. Hand to `tracking-audit`. 4. **Do the unit math.** Run:    ```   python3 scripts/channel_math.py --json inputs.json   ```    When the reader gives budgets per campaign and the campaigns do different jobs (cold account campaigns and retargeting), enter each campaign as its own row with its own job. Otherwise a channel with 2 jobs is split evenly between them, and the output says so.    Without Python, work it by hand and say so:    | Figure | Rule | Why |   | --- | --- | --- |   | Value per lead | ACV x lead-to-opportunity x opportunity-to-won (or ACV x first-conversion-to-customer) | What a lead is worth in first-year revenue |   | Ceilings | From the target cost per lead: cost per opportunity = target / lead-to-opportunity; cost per customer = target / (lead-to-opportunity x opportunity-to-won). Else from a payback target: most a customer may cost = ACV / 12 x payback months x gross margin (1 if not given), then down the funnel | The company's own limit, not a market average |   | Implied payback | Cost per customer / (ACV / 12 x gross margin) | Shows what the target cost per lead really allows |   | Test unit | Opportunity when lead-to-opportunity is known, else customer | Read a test on the earliest stage that tracks revenue |   | Test floor | 5 units x the ceiling per unit, spread over the test window (90 days by default, a quarter, the period pipeline targets are set in) | With fewer than 5 units, one more or less moves the cost by 20% or more |   | Read tests on | Opportunities when the sales cycle is longer than half the window, else closed-won | A lead from mid-window can't close inside it |   | Opportunities needed | Pipeline target per quarter / ACV | Assumes an opportunity is valued at ACV; say so |   | In-market estimate | Share entering the market per quarter = 1 / (years between purchases x 4) | The 95:5 logic with the reader's own cycle; always labelled an estimate |    - Ceilings use the profile's `## Targets`, which are the reader's plan. If the last 90 days show a different lead-to-opportunity rate, report both and name any call that would flip on the observed rate. Don't move the ceiling unless the reader updates Targets.   - If the implied payback is shorter than the sales cycle, say so: the target cost per lead may be too tight to scale, and that's the reader's call. 5. **Check each channel's needs.** For each of the 11 channels in `references/channels.md`, mark every need `met`, `not met` or `unknown` with the fact behind it. Then give each channel one call (each campaign, when a channel's campaigns do different jobs), first match wins. The budget left for tests is the total budget minus funding orders 1 and 2 in step 7:    | Call | Condition |   | --- | --- |   | `skip` | The profile marks it `never`; or a "skip when" condition holds; or a need can't be built (no account list exists for an account-based channel; nobody can answer in their own name for a community channel) |   | `later` | Not running yet, and a need is missing that can be built, or a need is `unknown` (name it, so the reader can confirm it); or a channel it depends on isn't proven yet (Microsoft Ads waits on Google Ads within the ceiling) |   | `fix` | Active, needs met, a demand channel with cost per unit above the ceiling: no new budget until it's back under |   | `core` | Active, and a demand channel within the ceiling; or an active brand channel (if its reach into the target accounts isn't measured, say so and ask for it); or a channel producing opportunities whose cost isn't tracked (say so, and ask for the cost) |   | `test` | Needs met, not yet run, and the test floor fits the budget left for tests; or no floor can be set (no ceiling), in which case the reader sets the cap and the output says the floor is `not available` |   | `later (budget)` | Needs met but the test floor is more than the budget left for tests |    A brand channel (`create`) is never cut on in-window cost per opportunity alone. Judge it on reach into the target accounts and its pipeline over 2 or more sales cycles: most buyers it reaches enter the market later. A demand channel (`capture`, `retarget`, `answer`) is judged on cost per unit against the ceiling. 6. **Check the model against the market.**   - Compare opportunities needed per quarter with the opportunities the channels made in the last 90 days. The gap is what new spend or tests must find; no gap means the plan protects what already works.   - If opportunities needed per quarter exceed the in-market estimate of accounts per quarter, the target can't come from in-market buyers alone: `NEEDS DECISION`.   - If the spend to reach the target at the ceiling is more than the quarter's budget: `NEEDS DECISION`.   - If the business model is product-led with no sales team, skip offers that need a sales call (demo requests, lead forms); channels must carry a self-serve offer. 7. **Set the split.** Fund in this order. Each amount carries the rule that set it.    | Order | Fund | Up to | Why |   | --- | --- | --- | --- |   | 1 | Demand channels within the ceiling | Their demand limit: where search impression share lost to budget reaches 0, or where the last added spend costs more than the ceiling per unit | The cheapest pipeline that exists now. Capture can't buy more searches than buyers make |   | 2 | Demand channels marked `fix` | Their current spend, not more | Fix before scaling a channel that costs more than a unit is worth |   | 3 | 1 test at a time | Its test floor | 2 tests at once split the budget and blur which one worked |   | 4 | Brand channels marked `core` or `test` | Everything left | Demand channels reach only the few buyers in market now; brand reaches the rest before they look |    - Without the demand limit (no impression share or search demand data), hold demand channels at their current budget, label the limit `not available` and ask for it.   - If step 1 uses the whole budget, the brand share is 0: write it as an open trade-off. Every future buyer would have to find the company on their own; give the amount that would fund 1 brand channel at its test floor so the reader can choose.   - If no channel can take what is left, hold it, and name the need that would release it once confirmed.   - If the test pool comes out of an active brand channel's budget, show the amount and offer to delay the test instead; the reader decides.   - Report the split now and proposed: brand and demand in amounts and shares. 8. **Order the tests.** List the `test` channels in the order they run, 1 at a time: why this one first (the need it meets best, the buying moment it serves), its budget at the test floor, what it's read on, the success line (5 or more units within the ceiling inside the window), the stop line (fewer than 5 units, or above the ceiling at the end of the window), and the skill that plans it. ## Check and verdict | Verdict | Condition (first match wins) || --- | --- || `BLOCKED` | No ACV or pricing, or no business model, after asking once || `MEASURE FIRST` | Channels are active and no lead can be traced to an opportunity or customer || `NEEDS DECISION` | Opportunities needed exceed the in-market estimate; or reaching the target at the ceiling costs more than the budget; or no channel passes its needs; or the reader's data conflicts with the profile on budgets or channel status || `PLAN` | Otherwise: every channel has a call, the split has amounts, the tests have an order | Named stops: - `BLOCKED`: ask for ACV or pricing and the business model.- `MEASURE FIRST`: stop the split; hand to `tracking-audit` with what is missing.- `NEEDS DECISION`: stop. Show the gap with its inputs and the options: widen the market (a new segment or region), raise ACV, lower or lengthen the target, raise the budget or accept a higher ceiling. Give what each option changes in the math. Write no split that pretends to reach the target. `PLAN` runs through with no stop. ## Learn Append to `## Learnings`: - a test ends: a `fact` with the channel, units, cost per unit, the ceiling and the call (kept or stopped), with the source;- the reader overrides a call or the split: a `proposal` with their reason;- a brand channel's later effect shows up (more branded search, more self-reported mentions, a higher win rate in reached accounts): a `fact` with its source;- a ceiling or the test window proves wrong for this company twice: a `proposal` to change it. Example: `- 2026-10-24 | channel-fit | fact | Microsoft Ads test stopped: 3 opportunities at $1,400 against a $1,000 ceiling in 90 days | Why: Microsoft Ads export and CRM, 2026-Q3 | Status: noted` ## Output ```Verdict: <PLAN | MEASURE FIRST | NEEDS DECISION | BLOCKED> · Coverage: <complete | partial (what's missing) | blocked> · Math: <script | by hand> EconomicsValue per lead: <n> · Ceilings (<source>): cost per lead <n>, per opportunity <n>, per customer <n> · Implied payback: <n> monthsTest floor: <n> per test (<n> per month over <n> days), read on <opportunities | closed-won> Channels| Channel | Jobs | Now | Needs met / not met / unknown | Cost per unit vs ceiling | Call | Why |(all 11, core first, skip last) Target check: <n> opportunities needed per quarter · in-market estimate <n> accounts per quarter (estimate) · spend at the ceiling <n> vs budget <n> · Flags: <list or none> Split| | Now | Proposed | Rule || Demand | <amount, share> | <amount, share> | <order step> || Brand | <amount, share> | <amount, share> | <order step> || Test pool | | <amount> | <order step> | Test order: 1. <channel>: <why first>, <budget>, read on <..>, success <..>, stop <..>, plan with <skill> Not available: <figure: what would provide it>Learnings applied: <list, or none>``` Then the handoff block. ## Hand off ```handoffskill: channel-fitversion: 1.0.0verdict: PLANcoverage: partial (no impression share data, so the demand limit is not available)for: <company>, <quarter>outputs:  ceilings: {cost_per_lead: 250, cost_per_opportunity: 1000}  core: [Google Ads, LinkedIn Ads]  fix: []  tests_in_order: [Microsoft Ads]  later: [Google Display: confirm monthly visitors]  skip: [Meta Ads, Reddit Ads]  split_proposed: {demand: 14000, tests: 1667, brand: 8333}next: [lead-value-math, linkedin-plan, microsoft-ads-b2b]open: [add a pipeline target to run the target check]``` Made-up values for illustration. `lead-value-math` takes the ceilings for each campaign; the channel skills named in `next` plan each `core` and `test` channel. ## Failure modes | Problem | Do this || --- | --- || The reader asks for "the usual B2B split" | There isn't one this skill will give. Show the split from the funding order and the reader's own figures || Platform-reported conversions disagree with the CRM | Trust the CRM for opportunities and customers; note the gap || A channel has under 5 units in 90 days | Mark the result `small sample`; keep or stop on fit and cost per lead, and extend the test only once || Only 1 channel is active and it is all capture | Expected for young companies. The split shows brand at 0 with the amount 1 brand test would need || The sales cycle is longer than a year | Read tests on opportunities and pipeline stages; brand channels on reach into accounts. Say closed-won won't show inside any test || A channel's rules or availability changed | Check the channel's current documentation; `references/channels.md` notes the date its facts were checked | ## Influenced by - Brian Balfour's Four Fits, especially channel-model fit (what a customer is worth limits the channels that can pay for them) and product-channel fit (channels don't bend to products): https://brianbalfour.com/four-fits-growth-framework- The 95:5 rule from Professor John Dawes of the Ehrenberg-Bass Institute, written for LinkedIn's B2B Institute: most business buyers aren't in the market at any moment, so advertising mostly works by building memory for later: https://marketingscience.info/news-and-insights/advertising-effectiveness-and-the-95-5-rule-most-b2b-buyers-are-not-in-the-market-right-now and https://business.linkedin.com/advertise/resources/b2b-institute/b2b-research/trends/95-5-rule- What we changed: ceilings and test floors come from the company's own ACV, rates and targets with no benchmarks, each of 11 channels has written needs to check, the split comes from a funding order instead of a fixed ratio, and the in-market share is estimated from the reader's own purchase cycle.

How it thinks

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

What we built

We found no good public method for this in B2B SaaS, so we built one. Ceilings and test budgets come from your own deal size, rates and targets, never from benchmarks, and the split comes from a funding order instead of a fixed ratio.

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

    Results from past tests and rules you accepted change the defaults before it starts.

  2. 02

    Check the tracking

    If no lead can be traced to an opportunity, it stops. A split read on form fills teaches each platform to find more form fillers.

  3. 03

    Set your ceilings

    From your deal size, conversion rates and target cost per lead or payback, it works out the most an opportunity and a customer can cost. It uses no industry averages.

  4. 04

    Check what each channel needs

    11 channels, each with written needs, such as a nameable account list for LinkedIn Ads or real search demand for Google Ads.

  5. 05

    Give each channel a call

    Core, fix, test, later or skip, with the need that decided it. Brand channels are read over 2 or more sales cycles, not on this month's cost per opportunity.

  6. 06

    Check the target against the market

    If the target needs more opportunities than the buyers entering the market, or more spend than the budget, it stops for your decision.

  7. 07

    Set the split and the test order

    Proven demand channels come first, up to the demand that exists. Then 1 test at a time, and brand gets the rest.

What you need

  • Required

    Business model, deal size or pricing, sales cycle and regions

    The company section of your profile → Your answers

  • Optional

    Conversion rates from lead to opportunity to customer

    The targets section of your profile → A connected CRM → A CRM export

  • Optional

    A target cost per lead, or a payback target in months

    The targets section of your profile → Your answer

  • Optional

    Channel results for the last 90 days (spend, leads, opportunities)

    Connected ad accounts and your CRM → Exports you upload → Figures you paste

  • Optional

    The facts each channel needs, like a target account list or real search demand

    Your profile and the buying moments and keyword skills' results → Up to 5 questions it asks you

What it can return

  • PLAN

    Every channel has a call, the split has amounts, and the tests have an order.

  • MEASURE FIRST

    Channels are running but no lead can be traced to an opportunity or a customer.

  • NEEDS DECISION

    The target needs more opportunities than the market or the budget can give, or no channel meets its needs.

  • BLOCKED

    It has no deal size, pricing or business model after one question.

Where it stops for you

  • When it has no deal size or pricing
  • When leads can't be traced to opportunities
  • When the target can't come from the market or the budget

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. Next quarter's budget is $24,000 a month.

Verdict

PLAN

Coverage: partial (no pipeline target in the profile, so the target check didn't run)

Ceilings, from a $250 target cost per lead

$1,000 per opportunity and $3,333 per customer, which pays back in about 2 months. A test needs 5 opportunities, so $5,000 over 90 days. Tests are read on opportunities, because the 50-day cycle is longer than half the window.

ChannelJobsCost against ceilingCallWhy
Google AdsCapture$913 against $1,000CoreWithin the ceiling, with no impression share lost to budget
LinkedIn AdsCreate$2,000 per opportunity so farCoreBrand: read on reach into the 900 target accounts over 2 cycles
Microsoft AdsCaptureUntestedTestGoogle Ads is within the ceiling, so the same searches can extend to Bing
Founder-led contentCreateNo spendCoreThe founder posts weekly and buyers are on LinkedIn
Meta AdsCreateUntestedSkipBuyers are operations directors, not founders or small teams
Reddit AdsCreateUntestedSkipNo active communities for clinic operations

The other 5 channels get a call and a reason in the full output.

SplitNowProposedRule
Demand70% ($14,000)58% ($14,000)Google Ads is at its demand limit
Test poolnone7% ($1,667)1 test at its floor: Microsoft Ads
Brand30% ($6,000)35% ($8,333)Everything left goes to LinkedIn's account campaigns
The handoff block the next skill reads
skill: channel-fit
version: 1.0.0
verdict: PLAN
coverage: partial (no pipeline target in the profile)
for: clinic scheduling company, next quarter
outputs:
  ceilings: {cost_per_lead: 250, cost_per_opportunity: 1000}
  core: [Google Ads, LinkedIn Ads, Founder-led content]
  tests_in_order: [Microsoft Ads]
  split_proposed: {demand: 14000, tests: 1667, brand: 8333}
next: [lead-value-math, linkedin-plan, microsoft-ads-b2b]
open: [add a pipeline target to run the target check]

What good looks like

  • Ceilings worked out from your own figures, with the math shown
  • A call for all 11 channels, with the need that decided it
  • 1 new test at a time, with a success line and a stop line
  • A split in amounts, each with the rule behind it
  • A clear stop when the target can't come from the market

Install

Use it in the tool you already have.

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

  1. 01
    Download channel-fit.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.

Why no industry benchmarks?

A benchmark is someone else's average. Your ceiling comes from what a lead is worth to you, from your deal size, your conversion rates and the target you set.

How does it split brand and demand?

In a funding order. Proven demand channels get budget up to the demand that exists, then 1 test runs at its smallest honest budget, and brand gets the rest.

Why won't it cut LinkedIn for a high cost per opportunity?

Brand channels reach buyers who aren't looking yet, and most of them buy later. It reads them on reach into your target accounts and pipeline over 2 or more sales cycles.

What's the 95:5 rule?

Research by Professor John Dawes for the LinkedIn B2B Institute argues that most business buyers aren't in the market at any given time. The skill estimates your own share from how often your buyers switch, and labels it an estimate.