Skip to content

Checks

Pipeline report

Reports pipeline per channel by opportunity and buying group, not by lead, and fails the report when missing sources or duplicates could change which channel comes out on top.

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 pipeline-report.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 bowtie model
Needs
Opportunities with account, created and close dates, amount, stage, source and type
Works with
Claude, ChatGPT, Codex, Cursor and any agent that reads SKILL.md
Version
1.0.0 · Last checked · License MIT
Maintained by
Sabih AhmedMarkOps
SkillMarkOps

Pipeline report

Example output

Pipeline you can act on
ChannelPipelineGrade
Organic search$98,000Trusted
Paid social$82,000Caution
Unknown$12,000Do not use

Reports pipeline per channel by opportunity and buying group, not by lead, and fails the report when missing sources or duplicates could change which channel comes out on top.

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
Doesn't need the internet.

3 test cases

pipeline-report/SKILL.md

In the folder

  • pipeline-report/SKILL.md
  • pipeline-report/references/stage-model.md
  • pipeline-report/scripts/pipeline_report.py
---name: pipeline-reportdescription: >-  Reports a B2B SaaS company's pipeline per channel by opportunity and buying group, not by lead, from a CRM export:  opportunities, pipeline, win rate, days to close, contacts and roles per deal, cost per opportunity and pipeline  against target, with new business and expansion kept apart. Flags where the data can't be trusted (missing source,  duplicates, stage skips, date errors) and fails the report when those gaps could change which channel comes out on  top. Use it for a monthly or quarterly channel review, before moving budget between channels, or before a board  update. Not for checking tracking (use tracking-audit) or for scoring leads (use lead-scoring).license: MITcompatibility: >-  Works in Claude, ChatGPT, Codex, Cursor and any agent that reads SKILL.md. Reads a connected CRM read-only if the  agent has one, otherwise CSV exports or pasted rows. Python 3 is optional, for the report and the trust checks.metadata:  author: SHAKEUP  version: "1.0.0"  stage: check  plugin: markops  homepage: https://www.shakeuphq.com/skills/pipeline-report--- # Pipeline report Shows which channels create pipeline, counted the way deals are bought, and says how far the figures can be trusted. ## Purpose - One job: one pipeline report per channel for one period, with a trust check on the data behind it.- Stage: check. It ends in a verdict on whether the report can carry a budget decision.- Unit: the opportunity. It never counts leads as pipeline, never adds new business and expansion together, and never adds self-reported and software channels together.- It never invents a win rate, deal size, cycle or benchmark. Every figure comes from the export or the profile, and a missing one is `not available`. ## Use when / not | Situation | Use || --- | --- || Monthly or quarterly channel review | this skill || Before moving budget between channels | this skill, after `tracking-audit` || A board or CFO asks for pipeline by channel | this skill || Platform figures and CRM figures disagree | `tracking-audit` || UTMs land in the wrong channel | `utm-check` || Deciding which leads sales calls first | `lead-scoring` | ## Inputs | Input | Required | Ways to provide, in order | If missing || --- | --- | --- | --- || Opportunities: id, name, account, created date, close date, amount, stage, source, type; the period plus 2 earlier quarters | yes | a connected CRM, read-only; a CSV export; pasted rows | `BLOCKED`: ask for the export || Contacts linked to each opportunity: opportunity id, created date, source, role or title | no | the same CRM; a contacts export with the deal association | channel from the opportunity's source; buying group `not available` || Stage date fields on opportunities | no | the same export | stage skip check `not checked` || Spend per channel for the period | no | ad platform exports; `## Channels` monthly budgets | cost per opportunity `not available` || `self-reported-attribution` handoff | no | its handoff block; a self-reported column in the export | self-reported column `not available` || `utm-check` handoff | no | its handoff block | channels named as the CRM writes them || `tracking-audit` handoff | no | its handoff block | platform-sourced figures shown without a tracking grade | ## Profile read From `company-profile.md`: `## Targets` (Pipeline target, Opportunity to closed-won rate, as targets to compare against, never to replace measured rates), `## Company` (Average deal size, Sales cycle, as stated values to compare with the measured ones), `## Buyers` (the buying group roles, to map titles to roles), `## Channels` (channel names and budgets), `## Accounts and data` (CRM), and every `## Learnings` entry for `pipeline-report`. No profile: ask for the pipeline target for the period and the buying group roles, in one message; continue without them if the reader doesn't know. Say that `company-profile` saves this next time. ## Steps 1. **Apply learnings.** A channel mapping or rule the reader accepted overrides the defaults. Name each one used. 2. **Set the period.** Default: the last full quarter, counted by opportunity created date (pipeline created). Keep 2 earlier quarters in the export for the cycle and the comparison. 3. **Give each opportunity 1 channel** (rule in `references/stage-model.md`): the earliest contact in its buying group created on or before the opportunity, else the opportunity's own source. Group source values with the `utm-check` scheme when its handoff exists, so the CRM and GA4 name channels alike. 4. **Run the report and the trust checks.**    ```   python3 scripts/pipeline_report.py --opps opps.csv --contacts contacts.csv --start 2026-07-01 --end 2026-09-30 \       --stage-cols "discovery_date,proposal_date" --target 1800000 --spend spend.csv   ```    Without Python, by hand: drop opportunities whose close date is before their created date (count them); list later copies of the same account with the same amount within 30 days, or the same name and account, and leave them out of totals; put missing sources (blank, unknown, not set, none, other, offline, import) in `unknown`; then per channel, sum opportunities and amounts, count won, lost and open, and take the median days from created to closed-won. Say which way you ran it. 5. **Run the trust checks.**    | Check | Fires when | Severity | Why |   | --- | --- | --- | --- |   | `headline_unsafe` | The `unknown` pipeline is at least the gap between the top 2 channels | critical | The unknown pipeline could belong to the runner-up, so the top channel isn't known |   | `ranking_changed` | Removing duplicates changes the order of the top 5 channels | critical | The ranking depends on records counted twice |   | `missing_source` | Any opportunity in `unknown` | warning | Shown as its own row, never spread across channels |   | `duplicate` | Same account and amount within 30 days, or same name and account | warning | A re-entered deal is far more common than 2 equal deals in a month; the list is for review |   | `date_error` | Close date before created date | warning | Left out of every figure |   | `stage_skip` | A won opportunity has a later stage date with an earlier one missing | warning | Stage conversion for that channel is `not available` |   | `zero_amount` | No amount | warning | Counted as an opportunity, adds nothing to pipeline |   | `no_contacts` | No linked contact | warning | Buying group unknown for that deal |   | `low_information` | Source is direct | warning | Kept as a channel; usually a missing tag, not a channel |   | `win_rate_immature` | A channel has more open than closed opportunities | warning | Its win rate will move as deals close; shown, labelled | 6. **Read the buying group.** Map each contact's role or title to the roles in `## Buyers`, `### Buying group` (match on the role's words; unmatched is `other`). Per channel: median contacts per opportunity and the share with 2 or more roles. Overall: win rate with 2 or more roles against 1 role or none, with the closed counts. Under 10 closed deals in either group, write the rate and say it is too few to compare. 7. **Compare with the target and the timing** (`references/stage-model.md`): pipeline created against `Pipeline target`; opportunities needed at the measured median amount; and when this period's pipeline should close, from the measured median days to close. Where the profile's stated cycle or deal size differs from the measured one by a quarter or more, show both. 8. **Grade each channel.**    | Grade | Condition (first match wins) |   | --- | --- |   | `do not use` | Its rank could flip with the `unknown` pipeline, or half or more of its opportunities carry a flag |   | `caution` | Any flag touches it, or its win rate is immature |   | `trusted` | No flag touches it | 9. **Write the read.** 3 to 5 lines, each with its figure, its source and the channel's grade. Never recommend cutting or growing a channel graded `do not use`, and never when `headline_unsafe` fired. ## Check and verdict | Verdict | Condition (first match wins) || --- | --- || `BLOCKED` | No opportunities in the period || `FAIL` | `headline_unsafe` or `ranking_changed` fired: the report can't carry a budget decision until the data is fixed || `PASS WITH NOTES` | Warnings only || `PASS` | No flags | Named stop: `BLOCKED` asks for the opportunity export. No other stops: a FAIL still delivers the report, labelled, with the fix list. ## Learn - The reader corrects a channel mapping ("Partner referral and Partner are one channel"): write a `fact`.- The reader overrides a grade or a check: write a `proposal` with the reason.- Each run's totals per channel (opportunities, pipeline, won): write a `fact` so the next run can show the trend. Example: `- 2026-10-24 | pipeline-report | fact | Q3 new business: paid social 14 opps, 336,000; organic search 9 opps, 198,000; unknown 4 opps | Why: CRM export, 2026-10-01 | Status: noted` ## Output ```Verdict: <PASS | PASS WITH NOTES | FAIL | BLOCKED> · Coverage: <complete | partial (what's missing)> · Report: <script | by hand>Period: <dates> · Opportunities: <n> (<n> after duplicates, <n> date errors left out) · Channel rule: earliest contact <n>, opportunity source <n> New business:| Channel | Opps | Pipeline | Share | Won | Win rate (closed) | Median days to close | Contacts per opp | 2+ roles | Cost per opp | Self-reported | Grade |(by pipeline; "unknown" as its own row) Expansion: <same table, or "none in period">Buying group: win rate with 2+ roles <r> (<n> closed) vs 1 role or none <r> (<n> closed)Against target: <pipeline created> of <target> (<share>) · opportunities needed at the median amount: <n> · this period's pipeline closes around <date range>Trust: <each fired check with its count and the records>Read: <3 to 5 lines>Learnings applied: <list, or none>``` Then the handoff block. ## Hand off ```handoffskill: pipeline-reportversion: 1.0.0verdict: PASS WITH NOTEScoverage: partial (no spend for organic channels)for: <company> new business, 2026 Q3outputs:  pipeline_by_channel: {paid social: 336000, organic search: 198000, unknown: 72000}  opportunities_by_channel: {paid social: 14, organic search: 9, unknown: 4}  win_rate_measured: 0.24  median_days_to_close: 71  median_won_amount: 23000  trusted_channels: [paid social]next: [lead-scoring, lead-value-math]open: [4 opportunities with no source]``` `lead-scoring` reads the measured win rate and the won and lost deals. `lead-value-math` (paid media collection) reads `win_rate_measured`, `median_won_amount` and `median_days_to_close` instead of the profile's stated targets. ## Failure modes | Problem | Do this || --- | --- || The export holds leads, not opportunities | Stop counting pipeline: ask for the opportunity export. Leads can't be pipeline || Amounts in several currencies | Report per currency, or convert only at a rate the reader gives, named || Most opportunities have no linked contacts | Use the opportunity source; buying group `not available`; recommend linking contacts at creation || A channel has 1 or 2 opportunities | Show the counts; no win rate comparison || The reader asks for "industry average" win rates | None in this skill; compare with their own earlier quarters || Sources are free text with many spellings | Group them with the `utm-check` scheme or ask the reader for the mapping; record it as a `fact` | ## Influenced by - Winning by Design's bowtie model, the idea that the customer journey continues past the sale into expansion and renewal, measured with one data model on both sides: https://winningbydesign.com/press/winning-by-design-releases-updated-bowtie-model-and-benchmarks-to-provide-go-to-market-teams-with-a-standardized-customer-journey-and-data-model/- SHAKEUP's pipeline target and pipeline timing calculators: the target worked back to opportunities from the company's own rates, and pipeline read 1 sales cycle ahead.

How it thinks

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

What we built

We count opportunities, because one deal can hold 4 contacts from 4 channels. The report checks itself, and it fails when its own data gaps could change the answer to which channel works.

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

    Channel mappings your team confirmed in earlier runs are applied before anything is counted.

  2. 02

    Give each opportunity 1 channel

    The source of the earliest contact in its buying group, created before the deal. Without contacts, the deal's own source. The report says which rule decided each count.

  3. 03

    Count opportunities, not leads

    Pipeline, win rate, days to close and contacts per deal, per channel. New business and expansion stay in separate tables.

  4. 04

    Check the data behind it

    Missing sources, duplicates, close dates before created dates, stage skips and deals with no amount, each counted and listed.

  5. 05

    Test whether the ranking holds

    If the unknown pipeline could cover the gap between the top 2 channels, or removing duplicates reorders them, the report fails.

  6. 06

    Compare with your target

    Pipeline created against your quarterly target, and when this quarter's pipeline should close, from your own measured cycle.

What you need

  • Required

    Opportunities with account, created and close dates, amount, stage, source and type

    A connected CRM, read-only → A CSV export of deals → Rows you paste

  • Optional

    Contacts linked to each opportunity, with created date, source and role

    The same CRM → A contacts export with the deal association

  • Optional

    Spend per channel

    Ad platform exports → Monthly budgets in your company profile

  • Optional

    The self-reported attribution and UTM check results

    Their handoff blocks

What it can return

  • PASS

    No flags on the data behind the report.

  • PASS WITH NOTES

    Flags that don't change the ranking, each listed with its records.

  • FAIL

    Missing sources or duplicates could change which channel comes out on top.

  • BLOCKED

    No opportunities in the period.

Where it stops for you

  • When there's no opportunity export for the period

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 deciding whether to shift budget toward paid social. Its CRM held 12 deals created in the quarter.

Verdict

FAIL

Coverage: partial (no spend)

Report: script

ChannelOppsPipelineWonWin rate (closed)Contacts per oppGrade
Organic search4$98,00022 of 31.5Trusted
Paid social3$82,0000None closed yet1Caution
Paid search2$34,0000None closed yet0Do not use
Unknown1$12,0000None closed yet0Do not use
Why it fails

a re-entered deal for the same account and amount puts paid social on top. Without it, organic search leads by $16,000.

Channel rule

one deal moved from paid social to organic search, because its first contact arrived through organic search 6 weeks before a second contact came from paid social.

Expansion

1 deal, $8,000, reported apart. Against target: $226,000 of $1,800,000.

The handoff block the next skill reads
skill: pipeline-report
version: 1.0.0
verdict: FAIL
coverage: partial (no spend)
for: new business, 2026 Q3
outputs:
  pipeline_by_channel: {organic search: 98000, paid social: 82000, paid search: 34000, unknown: 12000}
  median_days_to_close: 58.5
  trusted_channels: [organic search]
next: [lead-scoring, lead-value-math]
open: [review the duplicate deal, add a source to 1 deal]

What good looks like

  • Every figure counts opportunities, never leads
  • An unknown row, not missing sources spread across channels
  • Duplicates listed and left out of totals
  • New business and expansion in separate tables
  • A clear fail when data gaps could change the top channel

Install

Use it in the tool you already have.

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

  1. 01
  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 opportunities and not leads?

A B2B deal is bought by several people who arrive through different channels. Counting leads credits every channel with the same deal. The report counts each opportunity once and shows its buying group beside it.

Why would the report fail itself?

When the pipeline with no known source is larger than the gap between your top 2 channels, the top channel isn't actually known. A failed report says what to fix before any budget moves.

Does it include expansion deals?

Yes, in their own table with the same columns. Adding them to new business would credit a channel with revenue from customers you already had.

Can it work without contact data?

Yes. Each deal takes its own source, and the buying group columns say not available.