Unit Economics Agent Skill

Unit economics

The question is simple and usually unanswered: does one more customer make the business better off,

and how long does that take?

Build it in this order

• Define the unit. A customer, an account, a seat, an order. State it, because most

disagreements about unit economics are disagreements about the unit.

• Contribution margin — revenue per unit minus the costs that vary with it. Include everything

that actually varies: payment processing, hosting attributable to usage, support load, delivery,

third-party fees. Understating variable cost is the single most common error and it flatters

everything downstream.

• Acquisition cost — fully loaded. All sales and marketing spend, including salaries, divided

by customers acquired in the same period. Excluding people costs is the second most common error

and typically understates by half or more.

• Payback period — acquisition cost divided by monthly contribution. This is the number that

governs how fast you can grow without financing it.

• Lifetime value — contribution × expected lifetime, from observed retention. Not from a

churn-rate assumption chosen because it produces a good ratio.

Read it honestly

• Ratios hide the constraint. A healthy lifetime-value ratio with a long payback still means

growth consumes cash faster than it produces it. Payback governs the growth rate; the ratio

governs whether it is worth doing at all.

• Segment before concluding. Blended economics almost always conceal one segment subsidizing

another. The average is the least useful number.

• Use cohorts, not averages. Retention improves or decays over time, and blended figures mask

which. If early cohorts retain better than recent ones, the business is deteriorating while the

average looks stable.

• Do not extrapolate lifetime beyond your data. A twelve-month-old company cannot observe a

three-year lifetime, and assuming one is how unprofitable businesses appear profitable.

What it should change

Good unit economics by segment tell you where to spend. Bad ones tell you to fix the model before

scaling — no acquisition efficiency rescues a negative contribution margin, it only reaches the

loss faster.

Never

• Compare acquisition cost against revenue rather than contribution.
• Report lifetime value without stating the retention data behind it and its observation window.
• Treat improving ratios as progress without checking whether the mix simply shifted.