पाठशाला Pathshala · वृद्धि Vṛddhi, Growth · Lesson 11 · Build

Paid acquisition: Meta, Google and the ceiling you will hit

Paid channels work until they do not, and the dashboard average hides the moment. Track CAC by channel and by cohort, compute what the next rupee buys, and stop where it no longer pays back.

Pathshala, The Founder Library · 11 October 2026 · 7 min read

A fuel pump nozzle rests in its holder as the sun sets behind it. — Photograph: Michael Pointner · Pexels

In March a D2C brand in Bengaluru spends ₹5 lakh on Meta and acquires four hundred customers. In April it spends ₹10 lakh and acquires six hundred and fifty. The dashboard reports CAC rising from ₹1,250 to about ₹1,540, an increase the team calls acceptable. The two hundred and fifty customers the extra ₹5 lakh bought cost ₹2,000 each. Nobody computed that number, and it is the only one that mattered.

Paid acquisition is the most controllable growth channel a startup has: spend goes in, customers come out, and the dial can be turned on a Tuesday afternoon. That controllability is also its trap. Every paid channel saturates, the cost of the next customer rises faster than the average shows, and a company that scales spend on the average will cross the point where growth destroys value without seeing it. This lesson is about measuring paid acquisition honestly, by channel and by cohort, and about finding the ceiling before the bank balance does.

Blended CAC flatters, paid CAC tells the truth

Andreessen Horowitz’s note on 16 startup metrics makes the distinction investors care about. Blended CAC divides all sales and marketing spend by all new customers, including the ones who came through referrals, search and word of mouth at no marginal cost. It is not wrong, the note says, but it does not tell you whether your paid campaigns are working or profitable; paid CAC, the spend on a paid channel divided by the customers that channel brought, is more important in judging whether the business is viable. A company with a strong organic base can show a healthy blended number while every paid rupee loses money.

So compute CAC separately for every paid channel, every month, and include everything the channel costs: the media, the agency or freelancer, the creative production, the tools. For a D2C company count customers on delivered orders, not placed ones, because a cash-on-delivery order that is refused and returned has cost the media and the two-way freight and brought no customer. For a subscription company count customers when they first pay, not when they start a trial. The definition matters less than keeping it the same every month.

Track every customer by the month they arrived

A CAC figure on its own says nothing about whether the spend paid back. That depends on what the customers did afterwards, and the only honest way to see it is by cohort: every customer tagged with the channel and the month they were acquired, and each cohort’s gross profit tracked month by month. The sheet has one row per channel-month cohort and one column per month since acquisition. Each cell is that cohort’s cumulative gross profit divided by what it cost to acquire. The month in which a row crosses 1.0 is that cohort’s payback.

The tagging is plumbing and it has to be done before the spend, not after. UTM parameters on every ad link. The campaign and the first-touch source written to the customer record at sign-up or first order. Platform-reported conversions treated as an estimate to reconcile against your own orders, never as the record. Cohorts make problems visible that averages bury: a creative that brought cheap customers who never reordered, a festival-season cohort that paid back in a month and a post-season cohort that never did, a campaign whose customers churned at twice the rate of the rest. They also make the [CAC, LTV and payback](/library/cac-ltv-and-payback-the-three-numbers) arithmetic real, because the lifetime value is now observed rather than assumed.

Meta and Google do different jobs

Google search ads capture demand that already exists: someone typed the problem, and your ad is an answer. Volume is capped by how many people search for it, so the channel is cheap at first for a product people already look for and nearly useless for one they do not know to want. Meta, across Instagram and Facebook, creates demand: it puts the product in front of people who were not looking, which makes it the default for consumer and D2C products and the one whose costs rise fastest as it reaches past the buyers most ready to act. Dropbox is the cautionary case. Drew Houston’s Startup Lessons Learned reports that paid search cost between $233 and $388 to acquire a customer for a $99 product. People were not searching for a product category that did not yet exist. Brian Balfour calls the match between how a product makes money and what a channel costs channel-model fit, and it is decided before the first campaign: a ₹499 product cannot be sold through a channel that costs ₹2,000 a customer however well the campaign is run.

Every channel has a ceiling

The a16z note gives the shape of the problem in one line: costs typically go up as you try to reach a larger audience, with rough figures of a dollar a user for the first thousand, two dollars for the next ten thousand and five to ten dollars for the next hundred thousand. The first buyers a campaign reaches are the ones most ready to buy; each increase in spend reaches people a little less ready, at a higher price in the auction. Andrew Chen’s law of shitty clickthroughs adds the second force: every format that works is copied by competitors and ignored by audiences who learn to scroll past it, so even at constant spend the channel slowly gets worse.

An old outdoor tap lets a single drop of water fall against a grey textured wall. — Open the tap further and a saturated channel still only drips. That is the ceiling and the average will not show it. Photograph: Rajesh S Balouria · Pexels

The consequence is arithmetic. If customers rise more slowly than spend, the average CAC on the dashboard rises gently, but the cost of the last customer bought, the marginal CAC, is always higher than the average and rises faster. In the Bengaluru example the average moved from ₹1,250 to ₹1,540 while the marginal customer cost ₹2,000. A company that decides how much to spend by looking at the average will keep spending well past the point where the marginal customer stops paying back. The figure below makes the gap visible and finds the spend at which it closes.

The average CAC tells you what you have paid. The marginal CAC tells you what the next rupee will buy. Only one of them is a decision.

Reading the ceiling and moving it

Start with what you can afford to pay. David Skok’s SaaS Metrics 2.0 puts the best software companies at five to seven months to recover CAC and calls anything beyond twelve months anaemic; his other guideline is a lifetime value above three times CAC. Multiply monthly gross profit per customer by the payback you can finance, set that as the allowed CAC, and read the ceiling off the figure. Below the ceiling the next rupee pays back inside your target and spending more is right. Above it, each extra rupee buys a customer who costs more than you can recover in time, and the spend should come down to the ceiling however good the top line looks.

The ceiling moves, and there are only four ways to move it. Raise what a customer is worth: a higher price, a better second order, lower churn; each raises allowed CAC and pushes the ceiling right. Improve conversion after the click: a better landing page and a faster first value lower the CAC at every spend level. Refresh the creative and the audiences, which buys time against the law of shitty clickthroughs but does not abolish it. Open the next channel, because once one channel is at its ceiling the cheapest new customer is almost always somewhere else; the [first channel lesson](/library/finding-first-channel-that-works) is the method for finding it.

The weekly spend review and the monthly cohort review

Every Monday, per paid channel: spend, customers, paid CAC, and the marginal CAC of the week’s increase or decrease, computed as the change in spend divided by the change in customers against the previous four-week average. If the marginal CAC is above the allowed figure for two weeks running, cut spend back to the level where it was not. On the first working day of each month, update the cohort sheet: a new row for last month’s cohort per channel, a new column for every older cohort, and the payback month for any cohort that crossed 1.0. Write the ceiling estimate for each channel on one line with the date. Over six months that line is the most useful document the growth team owns, because it shows whether the ceiling is rising, which means the business is getting better, or merely being hit harder.

The saturation model is a simplification and every channel behaves differently; fit it to your own data. The figures quoted are from the sources below. Nothing here is investment advice.