How I model a consumer business

Most annual plans start with the revenue target and work backward, which is how bad CACs get rationalized: the target needs the customers, so the spend gets approved, and the price paid becomes whatever the target required. I model the other direction. Start with the customer base you already own: cohort by cohort, what do repeat rates and LTV curves say the returning population will spend next year with no heroics at all? That number is the floor.

The gap between the floor and the target is the acquisition mandate, and now it has a shape: how many new customers, at what blended CAC, with what first-year value. If the mandate implies buying customers above what the margin can carry, the plan is wrong, not the marketing. You find that out in the spreadsheet in January instead of in the P&L in September.

Spend, revenue, return. Each channel in the model carries all three, split across new and returning customers, and the two directions of that table are two different instruments. Rolled up, it's the P&L: total spend against total revenue at the profitability the year requires. Rolled down, it's management: each agency partner's targets fall out of their channel's row, so the paid media agency's ROAS goal, the affiliate agency's revenue-share target, and the direct mail program's cost per acquired customer all trace to the same sheet.

This is what makes vendor accountability arithmetic instead of vibes. When a channel misses, the conversation is about a number both sides agreed to in January, derived from a model both sides can read. When the model itself is wrong, that shows up too, and it gets fixed in the model, not argued about in the meeting.

Paid acquisition does not scale linearly. The market holds a limited supply of your customers at any given price, and pushing spend past that supply buys the next tranche at a worse rate. At some point the blended CAC crosses what the margin structure can absorb, and that point is a property of the business, not a bad quarter. I've watched a brand hit the same ceiling three times across seven quarters, at almost exactly the same indexed CAC each time. Twice might be luck. Three times at the same number is a wall.

The discipline: know the ceiling before you hit it, and pull back when you do. Spending through it converts growth into losses one customer at a time, invisibly, because topline still rises while unit economics rot underneath. The pullback feels like failure and is actually the model working. And a confirmed ceiling re-aims the whole growth effort: the question stops being how hard you can push spend and becomes which of the model's underlying assumptions you can break — a new channel, a different product, an offer that changes first-order value. You move the wall, because you now know you can't run through it.

Email looks free, so the reflex is to send more. But an email list is a finite attention reservoir, and volume past its band degrades the channel itself: inbox placement falls, opens fall, revenue per send falls, and the extra sends cannibalize the sends that were working. In five years of one brand's data, the lowest-volume quarter earned the best revenue per thousand sends and a quarter with five times the volume earned the worst. The industry-wide panel shows everyone making the same overshoot.

So owned channels get modeled like paid ones: with a return curve, not a lever. The question is never "how much can we send" but "where does the next send stop paying," and the answer moves with list health, which is downstream of acquisition, which is bounded by the ceiling above. It is all one system.

Revenue is a lagging output. By the time it moves, the cause is months or years old. The leading indicators live in the customer base: cohort repeat rates, the shape of new-customer quality, the ratio of new to returning revenue, list growth against traffic. A business can post flat revenue while its base quietly rots (aging cohorts propping up a shrinking intake), or post declining revenue while getting structurally healthier (every efficiency metric improving through a deliberate contraction). The topline number cannot tell those two stories apart. The base can.

A model nobody operates from is a spreadsheet. What makes it a management system is the cadence that keeps the numbers in front of the people spending against them: the daily revenue read against plan, the weekly budget review where money actually moves, and the annual rebuild where the whole structure gets re-derived from fresh cohort data instead of last year's assumptions plus ten percent. Ran daily for years, this loop is also what catches the ceiling early, spots the band overshoot, and notices base rot while it is still cheap to fix.