Notes · The numbers · June 19, 2026

They told me the lifetime value metric was flawed.

Years ago I ran growth for a brand, and our distribution partner owned the reporting. They steered the whole operation by one figure: blended MER. When I pushed to look at cohort lifetime value and real acquisition cost instead, I was told, in writing, that the lifetime value metric was flawed. The line back to me: "LTV is included in daily MER."

Here's what that one number actually did. It left entire channels of ad spend out of the math. And the distribution partner was a high-eight-figure business that, when I asked plainly what it cost to acquire a customer, could not give me an answer. Not wouldn't. Couldn't.

I kept waiting for the clever version behind the curtain. There wasn't one. Just a big, comfortable average with the expensive parts left out, and nobody willing to look behind it. That isn't sophistication, and it isn't bad luck. Running on a number that hides its own costs isn't efficiency. It's negligence with a tidy dashboard.

That isn't one partner's problem. It's the most common mistake in ecommerce, and it has a thousand quieter versions. The loudest is ROAS.

Your ad platform reports a 4x and you feel good about your day. Fair enough. But it only told you what came in. It has no idea what went out: product cost, shipping, card fees, the discount code, what it cost to get that customer through the door. It subtracts none of it. So a 4x that looks like profit can be a quiet loss. You won't catch it on the dashboard. You'll catch it in the bank.

Your accountant's P&L is right, and two weeks too late.

You do have a true profit number. It lives in your accountant's P&L, and it's correct. It's also finalized ten to fifteen days after the month closes. Fine for the taxman, useless for Tuesday. You cannot run ads on a two-week delay. By the time the statement says a campaign was underwater, you've spent another three weeks feeding it. A real P&L is necessary, and it is not a steering wheel. It's the rear-view mirror.

So I run a second one.

Alongside the accountant's version, I run a live marketing P&L. Same data, different job: it updates daily, and it makes one move the accountant's never does. It pulls ad spend above the line. Standard accounting files advertising as overhead, a budget you set and sit beneath. In ecommerce that's nonsense. Stop paying Meta and your sales stop the same afternoon. Advertising isn't overhead. It's a cost of the sale, same as the box and the postage.

The number that actually matters.

Once ad spend sits in the costs, you can finally see contribution: what's left from a sale after you've made the thing, shipped the thing, and bought the customer who ordered it. That is the number that pays your rent, your team, and you. Not revenue. Not ROAS. What's left.

And it reorders everything. A campaign with a lower ROAS on a cheaper-to-make product can beat a flashier one. A "winning" 30%-off code can be the exact thing bleeding you. You stop optimizing for the number on the ad platform and start optimizing for the number in your bank account.

This is where the marketing-versus-finance war ends.

For years the marketer and the finance lead fought, because they were looking at different numbers in different timeframes, and neither could see the other's truth in time to agree. Put ad spend above the line and suddenly there's one number on the table. Finance sets the floor: to cover our fixed costs and the profit we want, every dollar of revenue has to leave this much contribution behind. The marketer gets total freedom underneath it: scale as hard as you like, as long as the floor holds. No budget meetings. No permission slips. One shared line you both trust.

So why hire anyone?

Here's the part the software won't tell you. Building this is cheap now. A spreadsheet, a few connectors, an afternoon with AI, and you've got a live marketing P&L. The tooling was never the hard bit. The hard bit is reading it. That distribution partner had the data. The reports, the dashboards, a whole team. What it didn't have was anyone willing to load the real costs and sit with what the numbers actually said. So it steered by the comfortable average for years.

I rebuilt it from the ground up: real acquisition cost, real cohort lifetime value, the lot. Once we could see it, we leaned into what the average had buried. Returning customers went from 4% to 15%. Acquisition cost fell by more than 70%. None of it showed up in MER until we made it.

Knowing when the number says scale, when it says fix it first, and when it says stop, that isn't a formula. That's twenty years of watching these numbers move and being wrong enough times to know what they're saying. The model is the easy half. The read is the job.

← Back to Notes