Your ROAS Target Should Drop as You Scale: How to Budget for Efficiency Decay

Everyone repeats the same comforting idea: find your winning ad, find your 3x, then just keep pouring money in and the number holds. The ROAS stays put, the revenue scales with the spend, and life is good. Set it and feed it.

In reality, spend doesn't behave like that. The first dollars into Meta find the people already halfway to buying from you. The two-hundred-thousandth dollar is chasing people who need a lot more convincing. Efficiency decays as you scale, and it decays on a curve, not a cliff. If your target ROAS doesn't come down to meet that curve, you do one of two things. You choke the account by refusing to spend at a "worse" number, or you panic at a softer ROAS that was always going to show up.

Here's my take after taking a fair few accounts up the spend ladder: the ROAS you can run at $25k a month is not the ROAS you should expect at $180k a month. Pretending otherwise is how good brands stall at a ceiling they built themselves.

Let me show you what this actually looks like, then how to budget for it on purpose.

A scale-up, told in the numbers

I'll use a composite here, numbers invented and anonymised, but built from the shape of accounts I've genuinely run. Picture a supplements brand sitting at a ~$72 AOV with a solid repeat habit, the kind of product people reorder every couple of months.

When they came on, they were spending roughly $24k a month on Meta at a blended ROAS of about 3.6. Comfortable. Profitable. And completely stuck. Every time they nudged the budget up, the ROAS wobbled by morning and they yanked it back. They'd been hovering at that same spend for the better part of a year, treating 3.6 as the line that must not be crossed.

We took them from ~$24k to ~$190k a month over about nine months. Here's roughly how blended efficiency moved as we climbed, with the revenue it threw off at each tier:

• ~$24k/mo at a ~3.6 ROAS: ~$86k revenue from Meta
• ~$55k/mo at a ~3.1 ROAS: ~$170k revenue
• ~$95k/mo at a ~2.7 ROAS: ~$256k revenue
• ~$140k/mo at a ~2.4 ROAS: ~$336k revenue
• ~$190k/mo at a ~2.2 ROAS: ~$418k revenue

Look at that ROAS column on its own and your gut says the account fell apart. It didn't. To put this into perspective: the ROAS dropped by roughly 40% from top to bottom, and the revenue went up close to five times over. The 3.6 they were protecting was generating ~$86k a month. The 2.2 they were scared of was generating ~$418k.

That's the whole trade. A lower ROAS on a much bigger number is not a downgrade. It's exactly what you signed up for the day you decided you wanted to grow instead of just stay tidy.

The floor that decides how low you can go

The reason a falling ROAS isn't a crisis is that there's a floor underneath it, and that floor usually sits a long way below where founders start sweating.

Your floor is your breakeven ROAS, and your cost of delivery sets it. Take the same supplements brand. On a $72 order, say cost of goods is ~$26, card fees are ~2.5% (~$1.80), shipping is ~$4, and pick-and-pack is ~$3. That's a cost of delivery of about $35, or roughly 48% of the order. The other 52% is contribution margin, the bit you have to buy customers out of.

To break even on Meta on a first order, the ad spend has to come back out of that 52%. That lands first-order breakeven at a ROAS of about 1.9. Below 1.9 they lose money on the first purchase. Above it, they make money on it.

Now the part that changes the whole picture. That 1.9 is breakeven on the first order alone. This brand reorders. When I fold a conservative slice of 60 to 90 day repeat value back in, the point where a customer pays for themselves over the relationship rather than the single transaction, their true breakeven sits closer to a ROAS of 1.6.

So at ~$190k a month and a blended 2.2, they weren't scraping the bottom of anything. They were sitting comfortably above first-order breakeven and well clear of relationship breakeven. There was still room beneath them. The brand that refuses to dip below 3.6 "because that's our number" is leaving the entire 3.6-to-1.6 band of profitable spend untouched.

That band is where the growth lives. Most founders never spend a dollar in it because they've mistaken a habit for a limit.

What we actually changed at each step

Knowing the floor exists is one thing. Climbing toward it without flinching is another. The mistake isn't scaling. It's scaling the spend while holding the target frozen, then treating the inevitable softening as a fire to put out. The fix is to decide the step-downs in advance so a lower number is a plan, not an ambush.

Here's roughly how we ran each tier on that account.

We found the real breakeven before touching a single budget. Not the first-order floor, the relationship floor. Cost of delivery as a share of AOV, then the first-order breakeven, then a careful portion of repeat value folded in. That gave us the ~1.6 number. Everything above it was fair territory. You can't scale calmly into a band you've never measured.

We set a target band instead of a target line. Rather than "we run at 3.6", we agreed on a band: strong at low spend, easing as we climbed, and never dipping below a hard floor we set above breakeven. The band told the brand and the media buyer that a 2.4 at $140k was on plan, not off the rails. A single number gives you nothing to lean on when the dashboard moves. A band does.

We pre-wrote the expected ROAS against spend tiers. We mapped the whole thing out before we climbed, basically the table above. So when we hit ~$95k and saw a 2.7, we checked it against a plan that said "expect roughly 2.7 here" and kept going. That pre-written map is your defence against your own 7am panic. The number you predicted doesn't frighten you.

We moved budget in bites the account could digest. I tend to step spend up by something like 20 to 30%, then hold long enough to read a real signal, usually a week or two depending on volume, before the next step. Shove it up 80% in one jump and you'll get a temporary efficiency dip that's a function of the change itself, not the ceiling of the account, and you'll misread it as the account topping out. Patience here is the difference between scaling and stalling.

We judged the climb on blended, not in-platform. Meta's reported ROAS and your true blended ROAS drift apart as you grow, because more of your spend starts assisting sales that surface elsewhere. I read the climb on blended ROAS, or MER if you prefer that lens, measured against the breakeven floor. That's the figure actually wired to the bank account. The in-platform number flatters you on the way up and panics you on the way down.

Healthy decay versus a stall in disguise

One honest warning, because "drop your target as you scale" gets twisted into "let ROAS slide forever". It isn't that.

There's a real difference between efficiency decaying because you're reaching more of the market, which is expected and fine, and efficiency decaying because your creative has gone stale or you've exhausted the audiences that actually convert. Both look identical on the dashboard: a line drifting down. They are not the same problem, and they don't have the same fix.

The tell is the floor. If you're stepping down through your planned band and still sitting comfortably above relationship breakeven, you're scaling, keep going. If you've blown through the band and you're pressed against breakeven with no fresh winning creative in the pipeline, you're not scaling anymore. You're buying worse traffic at the same price. That's the moment to stop adding budget and go fix the creative, not the moment to drop your target again.

So the discipline cuts both ways. Step the target down to let genuine growth happen. Hold the floor so a stall can't dress itself up as a strategy.

Before your next budget review

If you've been guarding one ROAS number like it's sacred, I'd gently push back on it. Pull your cost of delivery, work out your real breakeven with repeat value included, and you'll very likely find a whole band of profitable spend sitting below the number you've been defending. That band is the single most likely place your next chunk of growth is hiding.

Then map your step-downs before you climb, not after. A 2.4 at scale that you forecast feels completely different from a 2.4 that jumped out at you, even though it's the identical number on the screen.

If you'd find it useful to see where your own floor actually sits, a Signal/Noise Audit takes your unit economics and spend history apart and lays out the band of ROAS you can genuinely scale into, plus whether a softening number is healthy decay or a creative problem in a costume. No pitch in it, just the maths set out so you can see your real ceiling rather than the one you've been assuming.

So where's your line drawn right now, and have you ever actually checked whether it's a floor or just a habit you never questioned?