When ROAS suddenly drops, the hand moves before the head does. Cut the whole budget, or kill the worst-performing channel. But that often makes performance worse, not better. Improving ROAS means finding where it's leaking first, and only then touching anything. Today, let's walk through the order to check before you touch the budget. Diagnose first, improve second.
The average ROAS number doesn't tell you what to fix
Say your overall ROAS is 300%. Not bad on the surface. But that single number gives you nothing to act on — the only decision it supports is "cut budget or don't."
Break it down by channel, and the story changes completely.
Channel A is at 480% while Channel C sits at 150%. The 300% average flattens both of those into one number. The thing to fix is C — not "the account" as a whole. Looking at one blended average hides exactly this. That's why the first move in improving ROAS is always to split it apart.
The leak is usually one of three things
Once you've split it by channel, you need to figure out why the low one is low. The cause is usually one of three things.
One, allocation is off. A good channel still has room to grow but isn't getting the budget, while an already-saturated channel keeps getting more. It's not that the channel is bad — the money is just sitting in the wrong place.
Two, the channel's efficiency has genuinely declined. The creative has aged and response has cooled, or you've already captured the cheap users and only pricier ones remain. This needs to be solved inside the channel.
Three, the back end is leaking. The ads are bringing people in fine, but they're dropping off at the landing page or checkout. No amount of tweaking the ads will fix ROAS here — you need to look at conversion rate or average order value instead.
These three call for completely different fixes, which is why narrowing down the cause matters so much.
The fastest lever is reallocation
Of the three, fixing allocation is the quickest win. Building new creative or overhauling a landing page takes time — but budget can move today.
The key is to look at the next won, not the average. Not your ROAS so far, but the ROAS that the next ₩1,000,000 you add would generate. That's called marginal ROAS.
Channel A looks best at an average of 480%, right? But A is already saturated — you've poured in about as much as it can absorb. Add another ₩1,000,000 there, and its efficiency (marginal ROAS) is only 160%. Channel B, meanwhile, has a lower average of 320% but still has room, so its next ₩1,000,000 earns close to 300%.
So the answer is clear. Shift budget from A to B and overall ROAS goes up. If you go by average alone and think "A is best, pile it all there," you end up moving in exactly the wrong direction. We go deeper on this marginal-utility view in our marketing budget allocation post.
A ROAS bump afterward isn't automatically "thanks to" reallocation
One thing worth flagging here. If ROAS goes up the week after you reallocate, it's tempting to credit the reallocation entirely — but that's not a safe conclusion.
Maybe the season happened to be good that week, or a competitor pulled back, or a promotion overlapped. ROAS rising at the same time as reallocation is different from reallocation causing the rise. Conflate the two and your next decision goes sideways. If this distinction feels fuzzy, see our correlation vs. causation post.
If you want to actually confirm reallocation worked, the standard approach is a holdout experiment — move budget in one place and leave a comparable one untouched. Observation alone only gets you association; confirmation needs an experiment.
Try this today
Do just one thing. Line up channel-level ROAS and look at the spread, not the average. Once you've found the worst channel, ask one question: "Is this channel just bad, or is it low because we've poured in too much (saturation)?"
If it's genuinely bad, pull budget. If it's low from saturation, stop scaling it and route the money to a channel with room. That one question is enough to decide what to move this week.
Wrapping up
Improving ROAS isn't about cutting budget — it's about finding where it's leaking and reallocating. That means seeing past the average to the channel-level spread, and to each channel's "next won."
The catch is that calculating marginal ROAS and saturation by hand for every channel is a hassle. If you want this done automatically, try uploading your media report CSV to our free saturation diagnosis tool. It shows you, channel by channel, whether you're saturated or still have room, and where efficiency is leaking, using a response curve. Everything you upload is processed in your browser only — nothing gets sent to a server, so there's no risk in putting your media spend numbers anywhere external.
And don't forget: even if ROAS goes up after you shift budget, an experiment is the safe way to confirm it was really the reallocation.