Track 3 · Engine · lesson 5
Unit economics
14 min
Track 3 · Engine · lesson 5
14 min
Two delivery businesses are both growing twenty percent a month. One of them loses 4 on every order and is therefore losing more money every month than the month before, on purpose, in public, to applause.
You cannot tell which is which from the growth rate. You can tell instantly from one customer.
Unit economics is the practice of shrinking a business down to a single customer and asking what happens when you run the machine once.
What does this one person pay? What does it cost to serve them? What did it cost to get them here? How long do they stay?
Four numbers. Everything else about the business — the office, the funding, the headcount, the story — is downstream of whether these four have the right signs.
Price. What one customer pays, per period.
Cost to serve. What it costs to deliver to that one customer, per period. Not the rent. Not the salaries of people who would be there anyway. The costs that exist because this customer exists.
Cost to acquire. Everything spent winning customers, divided by the number of customers won. Advertising, sales salaries, discounts, the free trial that never converted. All of it, including the money spent on people who said no.
How long they stay. Expressed as churn: the share who leave each month. A five percent monthly churn is a twenty-month average life.
From those you get two figures. Lifetime value is the contribution per period multiplied by the number of periods. And payback is how long you wait before a customer has repaid what they cost to win.
Predict
The model runs a single customer from the day they cost you money to the day they leave. The controls are the four numbers.
Unit economics
Hypothetical model, not a forecast
What one customer contributes, what one costs to acquire, and how long the two take to meet.
It assumes
It ignores
Find a churn rate at which this customer never repays what they cost to acquire — not in thirty-six months, not ever, however patient you are.
| Month | Net position per customer | Contribution collected |
|---|---|---|
| 0 | -300 | 0 |
| 1 | -240 | 60 |
| 2 | -186 | 114 |
| 3 | -137 | 163 |
| 4 | -93.66 | 206 |
| 5 | -54.29 | 246 |
| 6 | -18.86 | 281 |
| 7 | 13.02 | 313 |
| 8 | 41.72 | 342 |
| 9 | 67.55 | 368 |
| 10 | 90.79 | 391 |
| 11 | 112 | 412 |
| 12 | 131 | 431 |
| 13 | 147 | 447 |
| 14 | 163 | 463 |
| 15 | 176 | 476 |
| 16 | 189 | 489 |
| 17 | 200 | 500 |
| 18 | 210 | 510 |
| 19 | 219 | 519 |
| 20 | 227 | 527 |
| 21 | 234 | 534 |
| 22 | 241 | 541 |
| 23 | 247 | 547 |
| 24 | 252 | 552 |
| 25 | 257 | 557 |
| 26 | 261 | 561 |
| 27 | 265 | 565 |
| 28 | 269 | 569 |
| 29 | 272 | 572 |
| 30 | 275 | 575 |
| 31 | 277 | 577 |
| 32 | 279 | 579 |
| 33 | 281 | 581 |
| 34 | 283 | 583 |
| 35 | 285 | 585 |
| 36 | 286 | 586 |
Not there yet — keep moving the controls.
Above about twenty percent monthly churn, the customer is a permanent loss. Not a slow payback — a loss. The business can grow at any rate it likes and every point of that growth makes the hole deeper.
Notice also the two payback figures in the summary. Ignoring churn, this customer repays in five months. Allowing for churn, seven. The second number is the true one, and the first is the one that appears in plans, because it is the one you get by dividing two numbers you already have.
Not the same thing
Whether a customer is worth more than they cost, summed over their whole life.
How many months pass before a customer has returned what they cost to win.
Which is which? Put each one on a side.
Deciding whether to keep spending on a marketing channel this quarter
Deciding whether the business is fundamentally viable at all
Explaining why a profitable-looking company keeps needing money
Everything fixed. The rent, the founders' salaries, the accounting software, the office plants. None of it appears in the four numbers, and that is not an oversight — it is the whole design. Unit economics answer one question, which is whether the marginal customer is worth having, and mixing in the cost of existing makes that question unanswerable.
Which means the reverse trap is real too.
Check
Acquisition cost is the softest of the four, because its denominator is a judgement.
Divide marketing spend by all new customers and you get a flattering number, because some of those customers would have arrived anyway — through word of mouth, through search, through a friend. Divide it by only the customers the spending caused and you get an honest number, and a much larger one, and no reliable way to identify them.
Nobody has to be dishonest for the first number to be the one in the deck. It is easier to compute, it is defensible, and it is smaller.