Profit Leak Diagnoser

Profit-Leak Diagnoser

The first instinct in almost any profitability conversation is to cut costs. That instinct is wrong about half the time. Sometimes the drain is open, yes. But sometimes the tap is just running slow, and cutting costs when the real problem is falling volume makes things worse. This skill enforces the discipline of checking both sides before concluding anything.

The method

Profit has only two sides: revenue and cost. Isolate which side the problem is on first, then drill.

• Revenue splits into price × volume. Volume splits into customers × purchase frequency.
• Cost splits into fixed and variable. Variable splits into cost per unit × units.

Drill until you reach a hypothesis you can test. "Profit declined" is a statement. "Profit declined because volume fell in the mid-tier cohort while fixed costs held constant" is a hypothesis worth investigating.

Then run the unit-economics check, because a business can look fine on the tree and still be structurally broken per transaction. Five metrics, all of them, not just the ratio:

1. CAC — cost to acquire one customer.
2. LTV — value of one customer over the whole relationship.
3. LTV:CAC — the headline ratio; roughly 3:1 is a common health benchmark in subscription and

SaaS businesses, but healthy ratios differ by industry, margin structure, and stage, so treat it as a heuristic, not a law.

1. Payback period — how long to recover the acquisition cost.
2. Contribution margin — does each transaction make money before fixed costs?

A 3:1 LTV:CAC means nothing if payback is four years. You can be structurally sound and still run out of cash before the math works. Check all five.

Output format

• Profit tree: indented decomposition, with the suspected leak branch flagged.
• Unit-economics check: the five metrics, with whatever the user can supply and what's missing.
• Hypothesis: one testable sentence naming the likely leak and its driver.
• Fastest test: the single check that would confirm or kill the hypothesis first. (This is the

hypothesis-first-coach move; hand off to hypothesis-first-coach if the testing discipline needs its own pass.)

How to run it

Default to producing the diagnosis: build the tree and run the check, but always show that you looked at both revenue and cost so the reasoning is visible. Switch to coaching when the user signals they want to think it through: ask them which side they believe the problem is on and why, then push them to check the other side before committing.

Where this breaks

The tree assumes profit drivers are independent, but they interact: a price cut lifts volume, a volume push can raise variable cost per unit. When drivers are clearly coupled, say so rather than treating each branch as separate. Unit economics also assume you can cleanly attribute cost and value per customer, which breaks in businesses with heavy shared costs or long, fuzzy customer relationships. Flag low-confidence numbers instead of presenting them as precise.

Style

Plain language, define any metric on first use (CAC, LTV, contribution margin), no em dashes, short paragraphs. Never open with "cut costs."

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: AnugamChakra
• Source: AnugamChakra/think-like-a-strategy-consultant
• License: MIT

Install and usage instructions live in the source repository linked above.