Intro
Your blended ROAS is 4x. The monthly report looks fine. The CEO nods at the slide. Six months later, growth has stalled and nobody can quite explain why.
Here's what usually happened. One campaign was running at 9x and three campaigns were running at 1.8x. The blended number averaged them into one healthy-looking 4x. The winning campaign carried the account. The three bleeders quietly drained it. Until they couldn't.
This is the problem with blended ROAS as a primary decision metric. It hides the bleed.
Why blended ROAS still gets reported
Because it's convenient.
The ad platforms serve a single blended number by default. Monthly business reviews fit one number per row on the slide. Stakeholders who don't operate paid media want one chart to look at. The simplicity of "one ROAS" is the entire reason agencies still lead with it. Segmented reporting requires more spreadsheet hygiene, more dashboard configuration, more time in the weekly review, and harder conversations about reallocation.
Most agencies skip those conversations. The blended number lets them.
When campaigns "look fine" but revenue doesn't move, the blended number is almost always the place the story breaks down. Real performance lives one layer beneath. Pull that layer up, and the picture changes.
The math, made explicit
A blended average is a weighted mean. It tells you nothing about distribution.
Consider an e-commerce account with $40,000 in monthly ad spend across four campaigns:
Campaign A: $10,000 spend · $90,000 revenue · ROAS 9.0x Campaign B: $10,000 spend · $18,000 revenue · ROAS 1.8x Campaign C: $10,000 spend · $18,000 revenue · ROAS 1.8x Campaign D: $10,000 spend · $18,000 revenue · ROAS 1.8x Total: $40,000 spend · $144,000 revenue · Blended ROAS 3.6x
A monthly report shows: "ROAS 3.6x. Healthy."
A segmented report shows: one campaign is doing all the work. Three campaigns are barely returning the cost of media. The right move is to cut C and D (or restructure them), scale A, and free budget for a new test.
Same data. Two completely different decisions.
The blended number is the headline. The segmented number is the decision.
What segmentation reveals — Audrey (e-commerce, 5 years)
Audrey is a Brazilian e-commerce selling professional massage tables and aesthetic equipment. When they came to Loocro in May 2021, they were running Google Ads only. The blended ROAS was "okay." Every month felt like a gamble.
We didn't fix the ROAS by buying smarter clicks. We fixed it by changing what we reported.
Year 1 (2021): Blended ROAS 7.0x. We rebuilt the Google account and launched Meta cold. By end of month one, both channels were segmented in the reporting layer: by campaign, by audience, by creative.
Year 3 (2023): Blended ROAS 9.5x. Three years of segmented reading had revealed which audiences fed Google's branded search, which Google keywords signaled high-LTV cohorts worth retargeting on Meta, and which creatives consistently underperformed against the baseline. We didn't run more campaigns. We ran more disciplined campaigns.
Year 5 (2025): Blended ROAS 11.3x.
| Year | Blended ROAS |
|---|---|
| 2021 (year 1) | 7.0x |
| 2022 | 7.0x |
| 2023 | 9.5x |
| 2024 | 8.9x (supply-chain dip) |
| 2025 | 11.3x |
2024 dip was caused by an external supply-chain crisis. Stock shortages with no clear replenishment timeline. ROAS held above 8x throughout.
The blended number climbed every year. But the climb wasn't a function of any single optimization. It was a function of the reading. Five years of disciplined segmentation gave us the granularity to scale winners and cut bleeders before they accumulated.
What segmentation reveals — Ponce (B2B high-ticket, 10 months)
Ponce sells aesthetic and podiatry equipment to clinics. Average ticket: roughly R$2,000. WhatsApp closes. After a 2.5-year break, Ponce came back to Loocro in June 2025. The return produced six monthly all-time records in ten months, with a 21x average ROAS and a 39x peak.
The peak number is the headline. The 21x average is more revealing. The reading is what made both possible.
Month Revenue (BRL) Blended ROAS
Jun/2025 R$159,899 23.1x
Jul/2025 R$273,448 24.8x
Aug/2025 R$341,438 39.0x ← peak (Meta-specific)
Sep/2025 R$327,980 22.8x
Oct/2025 R$363,392 23.1x
Nov/2025 R$507,393 22.4x
Dec/2025 R$181,173 12.7x
Jan/2026 R$242,990 18.9x
Feb/2026 R$254,243 15.0x
Mar/2026 R$364,599 22.8x
The blended ROAS swings between 12.7x and 39x across the ten months. If we only reported blended, the monthly business reviews would have been a series of "why was December lower" conversations. Useful, but reactive.
Segmented, the same data tells a different story. The Meta audience that drove Aug 2025's 39x stayed strong through Nov. December's dip was concentrated in two specific campaigns hitting seasonal saturation; the rest of the account held. Knowing which segments were responsible meant we could pull budget back from saturated audiences in December and front-load January into the segments that were still warm.
The records weren't a campaign miracle. They were a reading discipline applied to a strong product with a competent sales team.
How to read ROAS the right way
Four dimensions cover most e-commerce and B2B operations:
1. ROAS by campaign. The first cut. Surfaces which campaigns are carrying the account and which are leaning on the others. Below 70 conversions per campaign per month, the number gets noisy; above that, it's a clean signal.
2. ROAS by audience. The second cut. Two campaigns with the same blended ROAS often have very different audience compositions underneath. A campaign at 4x driven by lookalikes performs differently from a campaign at 4x driven by retargeting; the lookalike-driven 4x is scalable, the retargeting-driven 4x has a ceiling.
3. ROAS by product line. The third cut, especially critical for e-commerce. A 5x ROAS aggregated across the catalog can hide that bestsellers run at 12x while a long tail drags at 1.5x. Often the fix isn't ad-side; it's product-side (pull the underperformers off the feed).
4. ROAS by creative. The fourth cut. Within a single campaign, creative performance can vary 3-5x across active variants. The blended ROAS at the campaign level absorbs that variance and makes the campaign look stable. Creative-level reporting reveals which variants deserve scale and which should be retired.
All four cuts in parallel give you the operational picture. None of them is meant to replace the blended summary. They're meant to explain it.
How Loocro reports ROAS, every month
Blended ROAS appears on every monthly report. We don't apologize for it. It's the headline. It's what the CEO looks at first.
Behind the headline, the segmented layer is what drives the weekly reallocation decisions. By Monday morning, we know which campaigns earned more budget this week, which audiences saturated, which creatives are ready for scale, and which product lines need a feed audit. The decisions get made on segmented data. The reporting closes with the blended number for the stakeholder summary.
That's the discipline. It compounds because the inputs to it compound: every week of segmented reading adds another data point to the long-running picture of what works and what doesn't. After five years of that discipline, the blended number climbs because the segmented decisions get sharper.
That's what happened at Audrey. That's what's happening at Ponce right now.
The 30-minute ROAS layer audit
Run this on your current account:
- Pull blended ROAS for the last 90 days. Note the number.
- Pull ROAS by campaign for the same 90-day window. Sort descending. High variance (top 3 far above the average, bottom 3 well below) means segmentation will reveal real signal.
- Identify your top 3 campaigns by spend. For each, pull ROAS by audience. If audience-level variance is wide, you have reallocation opportunities the blended number is hiding.
- Run the same exercise at the creative level for your top campaign. Variance above 50% between top and bottom creatives means the campaign is undermanaged.
- If three or more of these checks reveal variance the monthly report doesn't mention, your current reporting is hiding decisions you need to be making.