Your “30% margin” is hidingthe deals losing money.
A blended margin number is an average — and averages hide the tail of deals running thin, below floor, or outright negative. The Margin Leak Auditor sweeps your whole deal book and pulls those deals out by name, on the real pocket margin after every discount, freight charge, and rebate.
Plus four skills to stop the leak — guard discounts, review quotes, trace cost-change impact, diagnose where margin’s going. It defends the prices you’ve set; it doesn’t decide them.
The leak you can’t see in the average.
You know your blended margin. What you don’t know is which deals are dragging it down — because the healthy ones average out the bleeding ones, and the number looks fine. McKinsey’s research on the price waterfall found off-invoice leakage alone averages around 16% of list price, and once on-invoice discounts are added, total erosion often tops 40%. The deals that crossed the line are sitting inside your average right now, invisible.
And the leak compounds quietly: one more point of discount to close, free freight on a small order, a bundled give-away, a customer who negotiated net-60 years ago and now pays in 70 days. None of it shows on the invoice. All of it comes out of what you actually pocket.
A blended '30% margin' is a mean, and a mean hides its tail. The losing deals don't show up — the healthy ones average them out — so a book that looks fine can be quietly carrying deals below cost.
McKinsey's price-waterfall work found off-invoice leakage — freight, rebates, fees, terms — averages roughly 16% of list price, none of it visible on the invoice. Most teams never read the price past the invoice line.
One more point of discount to close, free freight on a small order, a give-away, a stale net-60 that's now net-70. Each is small; together they ratchet the realized margin down where no single report shows it.
Five skills for margin defense.
Sweep the book, guard the discount, vet the quote, model the cost change, diagnose the cause. One runnable centerpiece and four skills around it — every one single-purpose and composable.
“audit my margins,” “which deals are losing money,” “are we below floor”
Ships a real, runnable script. Sweeps your deal book or price list and flags every line whose realized (pocket) margin is below floor — LOSS, BELOW-FLOOR, THIN, DEEP-DISCOUNT. Quantifies the dollars leaking, contrasts it with your blended margin, and returns HOLD / CLEAR. Exits non-zero so it gates quote approval.
“can I give this discount,” “is this approved,” “customer wants X% off”
Checks a discount before the quote ships: does it keep pocket margin above floor, and is it within the rep's authority? Returns approve / escalate / below-floor — and the largest discount that would still hold the floor, so there's a number to counter with.
“review this quote,” “is this quote profitable,” “before I send this”
Reviews a single quote line by line for all-in pocket margin — catching the freight, rush, give-aways, services, and payment terms that don't show on the unit price. Shows the apparent margin vs the real one, and names the lines below floor.
“our supplier raised prices,” “costs went up,” “what does this do to margins”
When a cost moves — supplier increase, freight, FX — recomputes pocket margin and names exactly which SKUs and deals just flipped underwater or thin. Quantifies the hit and shows what price move would restore the floor (as information, not a directive).
“why is our margin down,” “where are we leaking,” “what's driving the erosion”
Traces a margin gap to its cause — discounting, cost creep, mix shift, give-aways, freight, or payment terms — biggest driver first. Separates a mix problem from a pricing problem (opposite fixes), so you treat the cause, not the symptom.
The auditor actually runs. Here’s real output.
This is the included margin-leak-auditor.py run against the six-deal sample book at a 30% floor — not a screenshot of a promise. Look at the headline: the book blends to 17.4% pocket margin, but two deals are actually losing money and one sits at 9.8%. It named them and quantified ~$4,975 leaking below floor.
$ python scripts/margin-leak-auditor.py deals-sample.csv --floor 30 --max-discount 25
HOLD lines: 6 | loss 2 below-floor 1 thin 2 deep-disc 0 ok 1
blended pocket margin 17.4% | margin leaking below floor: 4,975.00
--------------------------------------------------------------------------------
FIX BEFORE YOU BOOK IT (loss / below floor) (3)
• [LOSS] D-102 — Big-box account: LOSS: pocket margin -800.00 (-17.0%).
Selling at or below all-in cost — reprice or walk.
• [LOSS] D-104 — Loss leader: LOSS: pocket margin -170.00 (-37.8%).
Selling at or below all-in cost — reprice or walk.
• [BELOW-FLOOR] D-105 — Volume deal: BELOW FLOOR: pocket margin 9.8% vs
floor 30% (short 20.2 pts). Leaking 2,460.00 vs floor.
REVIEW (thin, deep discount, data gaps) (2)
• [THIN] D-101 — Acme reorder: THIN: pocket margin 32.9% is within 5 pts
of the 30% floor. One cost bump from breaching it.
• [THIN] D-103 — Standard SMB: THIN: pocket margin 34.1% is within 5 pts
of the 30% floor. One cost bump from breaching it.
Verdict: HOLD. ~4,975.00 in margin is below floor — recover it by repricing
the lines above.
$ echo $?
1Reads a CSV exported from your CRM or ERP, matches your column names flexibly, separates fix-before-you-book losses from review items, and quantifies the leak. JSON mode for tooling; --strict makes thin and deep-discount lines block too.
Built on the price waterfall — the realized-margin framework.
The method comes from McKinsey’s Marn & Rosiello (Harvard Business Review, 1992). A price doesn’t stay at list — it cascades down to what you actually pocket. Most teams stop reading at the invoice price; the auditor follows it all the way down.
The published / catalog price — where most analysis starts.
Volume breaks, promos, customer-specific deals.
What the customer is billed. Most teams stop here. That's the mistake.
Freight, rebates, fees, allowances, payment-terms cost.
What you actually keep. Pocket margin = (pocket − cost) ÷ pocket.
A sourced primer is included — and the engine works in margin (% of price), never silently mixing in markup (% of cost), a common and expensive error.
Set the price, or defend it — two different jobs.
If you’ve seen our “Should I Raise My Prices?” Decision Kit, here’s how this is different. Most businesses want both.
- ·Models a deliberate price increase
- ·Forward-looking, occasional decision
- ·“By how much, and what loss can I absorb?”
- ·A spreadsheet you sit down with
- ·Sweeps actual deals for margin below floor
- ·Ongoing, operational check on real transactions
- ·“Which deals are violating the price I set?”
- ·Skills + a runnable auditor on your CSV
The Kit decides where to set the price; the Auditor catches every deal quietly violating the price you already set. Neither replaces the other.
Clear about the lane. No inflated promises.
- 5 production-grade SKILL.md skills with NEVER/ALWAYS guardrails and worked examples.
- A real, runnable margin-leak-auditor.py — stdlib, HOLD/CLEAR, leak quantification, exit codes, flexible CSV.
- A sourced pocket-margin / price-waterfall primer, with the margin-vs-markup trap spelled out.
- Per-skill references: discount authority, all-in cost checklist, cost-change playbook, leakage sources.
- An AGENTS.md companion for standing margin discipline — pocket margin, no fabricated costs, stay out of price-setting.
- A README covering Claude Code, Codex, Cursor, Gemini, and Copilot installation.
- Not a price-setting tool — it defends your floor; it won't tell you what to charge or to raise prices.
- No invented numbers — a missing cost is flagged, never guessed into a fake margin.
- No markup/margin confusion — it works in margin (% of price) throughout.
- Not an accounting system or ERP — it does the margin judgment around your data, not the books.
- No SaaS, no monthly fee, no telemetry — your deal and cost data never leave your environment.
If you discount or hold a price list, margin leaks.
Anyone who discounts
If reps or you give discounts to close, margin leaks at the discount line. Guard it before the quote ships, and sweep the book for what already slipped through.
Distributors & resellers
Thin margins, freight, and rebates make pocket margin the only number that matters — the average flatters the losers. The auditor pulls them out by name.
Agencies & services firms
Bundled hours, scope give-aways, and blended rates hide losing engagements inside a healthy-looking average. The auditor names them.
Manufacturing & Distribution Ops Pack
$89When a supplier cost moves — its territory — the cost-change impact analyzer here names which SKUs flipped underwater. Ops that moves the stock, margin defense that protects what it sells for. A natural pair.
“Should I Raise My Prices?” Decision Kit
$49The other half of the pricing question: when you want to DECIDE a deliberate increase and model the volume you can absorb. This auditor then defends whatever floor you set. Different jobs, both wanted.
AI Service Pricing Kit
$79Sets the rates and pricing model in the first place — sourced market rates, model decision tree, margin defense playbook. Price it here, then defend the realized margin deal by deal with the auditor.
The questions finance and sales leads actually ask.
It finds the deals and SKUs that are quietly losing money — the ones your blended margin number hides. Five Claude skills: a runnable auditor that sweeps your whole deal book or price list for realized (pocket) margin below floor; a discount guard that checks a discount before a quote goes out; a quote margin reviewer that catches hidden cost drivers in a single quote; a cost-change impact analyzer that finds which products flip underwater when a cost rises; and a leak diagnosis skill that traces a margin gap to its cause. It runs on a CSV you export from your CRM, ERP, or spreadsheet.
They do opposite jobs, and most businesses want both. The Decision Kit helps you DECIDE a price — it models a deliberate increase and tells you the volume loss you can absorb. The Margin Leak Auditor DEFENDS the price you already set — it sweeps your actual deals and finds the ones quietly running below floor because of discounting, freight, rebates, or cost creep. One is a forward-looking pricing decision you make occasionally; the other is an ongoing operational check you run against real transactions. Put simply: the Kit decides where to set the price; the Auditor catches every deal that's violating the price you already set. Neither replaces the other.
Pocket margin is what you actually keep after a price erodes through every deduction — the framework comes from McKinsey's price waterfall (Marn & Rosiello). A list price drops to an invoice price after on-invoice discounts, then to a pocket price after off-invoice deductions: freight, rebates, fees, payment terms. Pocket margin = (pocket price − cost) ÷ pocket price. Most teams stop at the invoice price and never see the off-invoice leakage — and off-invoice leakage alone averages around 16% of list price per McKinsey. The auditor runs the whole waterfall, so it catches the margin that dies below the invoice line.
Because averages hide the tails. A business that reports '30% margins' is almost always carrying deals that run thin, below floor, or outright negative — and they're invisible inside the blend. In the auditor's own sample run, a deal book that blends to 17% pocket margin contains two deals that are actually losing money and one sitting at 10%. The whole point of a line-by-line sweep is to pull those losing deals out by name and quantify the leak, which a single average number can never do.
No — that's deliberate, and it's the line that keeps this distinct from a pricing-strategy tool. Every skill stays in the margin-defense lane: it flags which deals are below your floor and quantifies the leak, but it does not set your prices or recommend an across-the-board increase. When it shows the price move that would restore the floor on a specific line, that's information for your decision, not a directive. Deciding what to charge — and whether to raise prices generally — is your call (and the job of a different tool).
No. The auditor computes margin only from the data you give it, and if a line is missing a cost it skips it and says so rather than guessing a cost that would fabricate a margin. The skills never invent a cost, freight figure, discount, or rebate — a missing number gets flagged for you to confirm. It also works strictly in margin (percent of price), never silently mixing in markup (percent of cost), which is a common and expensive error. The honesty spine runs all the way through: it reports the real realized margin and says HOLD when a line is below floor.
You don't need to be technical — the skills run through normal conversation in Claude, and the one runnable piece (the auditor) takes a CSV exported from your CRM, ERP, or spreadsheet; in Claude Code or Cowork the assistant runs it for you and explains the output in plain English. Refund: thirty days, no questions. The honest test — export a deal book or price list and run the auditor against your real floor. If it comes back CLEAR, you've confirmed your book is healthy for free. If it surfaces even one deal quietly losing money, the recovered margin dwarfs the price. If it doesn't earn its keep, email us and we refund in full.
Margin is one lever; cash is the whole game. Cash-Flow Sentinel reads your actual inflows and obligations and flags a runway risk weeks out — the founder-grade Executive Suite that grades the whole company, not one leak. $249.
Run it on your own deal book.
Export a deal book or price list to a CSV and run the auditor against your real floor. If it says CLEAR, you’ve confirmed your book is healthy. If it surfaces one deal quietly losing money, the recovered margin dwarfs the price. Installs in Claude Code, Pro, Codex, and Cursor. Thirty-day refund either way.
Sold by RedHub AI LLC · Secured by Stripe · redhub.ai