The classic scene in management meetings: the report opens, ROAS shows 8, the decision is made — "it's working, let's double the budget." Two months later, at the same table: "We raised the budget but ROAS dropped to 4; the ads broke." The ads didn't break. The wrong number was being watched from the start.
Average ROAS vs marginal ROAS
The ROAS in your report is an average: total revenue divided by total spend. The budget question is different: "What will the next lira I spend bring back?" That's marginal ROAS, and it is always lower than the average — because every channel harvests the easiest conversions first. The first ₺1,000 goes to people already searching for you, who know the brand, with items sitting in carts; the 50,000th lira goes to people who've never heard of you and cost real money to convince. An account showing ROAS 8 often runs at 2–3 on its last slice of spend; when you double the budget, all the added money works in that band, and the average sinks. That's not a malfunction; it's the law of diminishing returns wearing an ads costume.
What further inflates the average: brand traffic
There's a layer that distorts the picture even more. Your average ROAS includes the audience that came to buy anyway: brand searches, remarketing, existing customers. These segments naturally show very high ROAS and pull the average up — but a budget increase doesn't grow them, because brand search volume is capped by your brand's awareness, not your ad budget. So a meaningful share of that "ROAS 8" may come from segments that cannot be scaled. The added budget, meanwhile, always flows to the least efficient end.
How I see the marginal in practice
Nobody has perfect measurement, but approximation is possible. First, incremental steps: instead of doubling at once, raise budget in ~20% steps and compute "added spend / added revenue" at each step — that ratio is that step's marginal ROAS, and after two or three rows in a table the curve reveals itself. Second, segment separation: keep brand/remarketing in their own campaigns and base growth decisions solely on cold-audience efficiency. Third, platform tools: Google's bid simulators and budget recommendations project "this spend → this many conversions" from their own data — remember they are sales tools, but as directional indicators they're valuable.
The right question: at which marginal do you stop?
Once you see marginal ROAS, the question changes from "what should ROAS be?" to "down to which marginal ROAS is spending still worth it?" The answer isn't in the ad account; it's in the business's economics: gross margin, return rate, repeat-purchase value. A business with 60% margin grows profitably down to a marginal ROAS of ~1.7; for a 25%-margin business the same number means losses. That's why the same "ROAS 4" is a brake for one business and an accelerator for another. On the agency side, the most valuable conversation is computing that threshold together with the client — because once it exists, budget debates end and arithmetic takes their place.
In short: average ROAS looks backwards, marginal ROAS looks forwards. A budget decision is a forward-looking decision and deserves a forward-looking number. Doubling budget because the report says 8 is hitting the gas while staring into the rear-view mirror.