What “scaling” actually means
Scaling a campaign means putting more budget behind an ad that’s already running, on the theory that if ₹5,000 a day brought in good orders, ₹8,000 a day will bring in more of them at a similar rate. Stopping — or pausing — means doing the opposite: pulling budget off a campaign, either because it never worked or because it’s stopped working.
Both decisions get made constantly on a live ad account, and both get made badly by default, because the number everyone looks at first — ROAS, return on ad spend, meaning revenue divided by what you spent to get it — is the number least equipped to answer either question on its own.
Why ROAS alone gets this wrong
ROAS is measured at the moment of purchase. It has no way of knowing what happens next: whether the order gets delivered or comes back as an RTO (return to origin — a parcel the courier couldn’t deliver, refused at the door or address not found), whether the customer paid or ordered cash-on-delivery and never opens the door, or whether this is a buyer who’ll order again next month or never again.
Take a real shape of the problem. A campaign shows ₹4 in revenue for every ₹1 spent — a 4x ROAS, which reads as strong. But it’s running in a category and region with a 30% RTO rate. Roughly a third of that “revenue” never actually arrives as money: no payment, plus you’ve paid for shipping both ways. The campaign that looked like your best performer is quietly one of your worst, and the dashboard never says so, because the dashboard stops counting at the order.
This is why we don’t scale off ROAS. We scale off what a campaign’s orders turn into after returns, RTOs and refunds settle — net contribution, not gross revenue.
Why it happens — the trap is structural, not a mistake
Ad platforms optimise for the event you tell them to optimise for, and the easiest event to tell them is “purchase.” Meta and Google can both be pointed at a deeper event — a delivered order, a non-refunded sale — but that takes more setup, more data, and more patience for the numbers to stabilise than pointing at “purchase” and watching ROAS climb. Most accounts, run by an agency paid on ad spend or by a founder short on time, never make that switch. The platform keeps reporting a number that looks good and was never designed to answer the question you actually care about.
The second trap is diminishing returns disguised as growth. Every campaign has a ceiling — an audience size, a level of daily budget — past which the platform starts showing your ad to worse-fit people to spend the extra money. Revenue keeps climbing when you scale past that ceiling. Order quality doesn’t.
What we actually look at before scaling or stopping
Net contribution over the last 14-30 days, not the last 24 hours. A campaign’s true delivered/RTO ratio and repeat-purchase rate take at least that long to show up; anything shorter is noise.
Marginal, not average, performance. The question isn’t “is this campaign profitable overall” — it’s “would the next ₹10,000 of spend still be profitable.” A campaign can be net-positive in aggregate while its most recent spend is already underwater, because the easy, high-intent audience got exhausted first.
Where the extra spend is actually landing. If daily budget increases are pushing the same ad to a wider, colder audience, cost per order rises and RTO risk usually rises with it — a colder buyer is a less committed one.
What happens downstream, not just what the platform reports. We cross-check ad performance against actual delivered orders and real payment collection, not the platform’s own conversion pixel, because the pixel counts an order the moment someone clicks “place order” — before you know if you’ll ever see the money.
When those checks hold up, we scale — in 20-30% steps, watching the same numbers after each one, not in a single jump that resets the platform’s delivery algorithm and usually costs you the efficiency you scaled to protect. When they don’t hold up, we stop, even if the top-line ROAS is still telling a good story.
How we approach this for Partner Brands
We report net contribution alongside ROAS from week one, specifically so a founder never has to choose between the two numbers without seeing them side by side. When we recommend pausing a campaign that’s showing a “good” ROAS, we say exactly why — which downstream number changed — rather than asking you to trust a gut call. And because we’re paid a percentage of what you actually keep, not a percentage of ad spend, we have no structural reason to keep a campaign running past the point it stops making you money. Recommending a pause costs us the same way it costs you.
Sources
Ad-platform mechanics — audience saturation and the delivery-algorithm reset on large budget jumps — reflect Meta’s and Google’s own published advertiser guidance. The RTO/ROAS gap uses the same 2026 Indian logistics data cited in our RTO guide. Account figures here are illustrative, not client data.