How to Split an Ecommerce Marketing Budget

Here is the symptom I actually get called about: every channel’s own dashboard looks acceptable on its own terms. Google Ads shows a decent return on ad spend, revenue divided by what was paid for it. Meta shows a decent cost per result. Email open rates are fine. And total growth has stalled anyway, because the mix between them was built by whoever pitched hardest this quarter, not by any plan.

Every guide to fixing this hands you a percentage-of-revenue table: spend this share overall, split it this way by channel. I read three of them while writing this, and their numbers disagree with each other, and in two cases disagree with themselves on the same page. A borrowed percentage was never going to diagnose your account. What follows is a check that uses your own numbers instead.

The symptom: every channel looks fine and growth still stalled

Individually healthy channels and stalled overall growth is the single most common pattern I see opening an account for the first time. It happens because a dashboard measures a channel against itself, not against what else the same dollar could have done. A channel can be genuinely efficient and still be the wrong place for the next dollar, and nothing in its own report will ever say so.

Why percentage-of-revenue tables don’t actually tell you anything

Because the published tables don’t agree with each other, and that disagreement is the tell. One widely shared framework gives a headline share of revenue for overall marketing spend, then a different share for early-stage brands, then a third share again in its own FAQ, for overlapping revenue bands. Another lists six spending categories with suggested shares that, added together, exceed the entire budget more than once over. If the people publishing these frameworks cannot make their own numbers agree on a single page, the number was never the point of the exercise.

A percentage of revenue also ignores the thing that actually decides whether spend on a channel is working: margin. Two stores at the same revenue, spending the same share on ads, can have completely different amounts of room to pay for a customer, because gross margin sets the ceiling a channel can spend and still be worth it. A table that only knows your revenue cannot know that.

What should actually decide the split?

What each channel returns against what it costs to acquire a customer there, compared across channels rather than against a published average. That comparison is the customer acquisition cost by channel, and it is the one number a percentage table cannot give you because it depends entirely on your own margin, your own catalog and your own customers.

It is also worth knowing before you compare that most ad platforms already reallocate budget for you inside a single channel, which is part of why a channel’s own total can look healthy while still being pointed at the wrong ad sets. Meta’s own documentation on Advantage campaign budget describes a budget being moved automatically and continuously across a campaign’s ad sets to chase the best available results. Google Ads works on a similar principle at the campaign level; its help page on average daily budgets describes the daily figure as what you are roughly comfortable spending, not a hard cap the platform follows exactly. Both tell you that the budget you set at the top is not the budget that actually lands where you typed it.

The ten-minute check: rank channels by what a dollar actually returns

Pull the last 90 days of spend and revenue for every channel into one sheet, including email and SMS even though they rarely carry a cost per click. For each one, divide total spend by new customers acquired, not total orders, since a channel that only sells to existing customers is doing a different job and should not be judged against one that finds new ones.

That last step matters more than it looks. Multiple channels frequently claim credit for the same order, because most analytics tools default to crediting the last channel a visitor touched before buying. Google Analytics 4’s own page on attribution settings describes an attribution model as a set of rules that decides how credit is divided among the touchpoints on a path to a sale, and the model you have selected changes every number this check produces. Read your mix against one consistent model, and expect a channel like email, which usually touches an order late, to look worse under a model that credits it less generously.

How do I know a channel deserves more budget?

A channel deserves more budget when its cost per new customer sits comfortably under your margin floor and it has not yet run out of audience. Both conditions matter. A channel can be cheap and still be nearly saturated, in which case the next dollar buys a worse customer than the last one did, not a proportionally cheap one.

The signal that a channel still has room is simple: cost per new customer holding steady or falling as spend rises. The moment it starts climbing while nothing else about the account changed, that channel has found its ceiling for now, whatever the percentage table said it should be getting.

How do I know a channel deserves less, or none?

A channel deserves less when its cost per new customer sits above your margin floor with no clear path down, or when its sales turn out to mostly overlap with a cheaper channel’s. Cutting a channel is not a failure of that channel. It is often the correct response to what the numbers say, and it can raise your overall return even while total revenue falls.

At Petticoat Lane, spend was down 32% year over year at a better return, $238,587 managed over 24 months at a 3.9x blended return. That is what a deliberate cut looks like when it works: less money moving, pointed more accurately, with the freed budget not forced into a channel that could not absorb it efficiently either.

What if my budget is too small to split at all?

Below roughly $5,000 a month in total spend, splitting across more than two channels usually hurts you, because every channel needs enough volume to tell you anything, and a thin split starves all of them of the data each one needs to improve. None of the published frameworks addresses this directly; they assume a budget large enough to divide six ways.

The practical answer at a small budget is sequencing, not splitting. Pick the one or two channels your own numbers say are cheapest to acquire on, fund those properly, and add a third only once the first two are both past the volume where their own numbers are stable month to month. A thin spread across five channels at $1,000 each teaches you less than a real test of two at $2,500.

When the problem is bigger than any one channel

When the split keeps drifting back to whoever pitched last, that is not a channel problem anymore. It is a decision that needs to sit with someone whose job is the whole mix, not any one platform’s performance. A paid-media retainer manages what happens inside a channel; it does not referee which channels get funded in the first place, and a split that nobody owns ends up being decided by whoever showed up with a pitch deck this month.

That ownership, deciding the mix and the budget across every channel rather than running one of them well, is the actual job description behind a fractional CMO. It is also most of what separates it from a single-channel retainer: the fractional CMO tier exists specifically to own channel mix and budget allocation as an ongoing decision, not a one-time table. More on the channels themselves is in Google Ads vs. Meta Ads for ecommerce, and the rest of what I write about running the whole function is in ecommerce strategy on the blog.

Questions people ask