Blended ROAS is total revenue divided by total ad spend across every channel; creative ROAS is the revenue attributed to one creative asset, across every ad it runs in, divided by that creative's spend. Blended ROAS tells you whether advertising pays for the business overall. Creative ROAS tells you which assets earn that result, and it routinely tells a different story: a healthy blended 3.8:1 can sit on top of a portfolio where one hero carries the account and most creatives lose money.
This guide puts the two side by side with platform ROAS, works one month through all three, and covers how to calculate ROAS per creative and what to do with the answer. For the formula itself and the ACoS conversion, start with the ROAS glossary.
What is the difference between blended ROAS and creative ROAS?
They answer different questions at different scopes. Blended ROAS is the business-level number, creative ROAS is the asset-level number, and platform ROAS (what most people mean by plain "ROAS") sits between them:
| Blended ROAS | Platform ROAS | Creative ROAS | |
|---|---|---|---|
| Formula | Total revenue / total ad spend | Platform-attributed revenue / platform spend | One creative's attributed revenue / its spend, summed across every ad it runs in |
| Revenue counted | Every sale, ad-driven or not | Sales the platform credits under its attribution setting | Platform-credited sales of one asset |
| Scope | The whole business, all channels | One ad account or platform | One creative, across ads, campaigns and platforms |
| Question it answers | Is advertising paying for itself overall? | Is this channel pulling its weight? | Which assets earn the efficiency? |
| Who reads it | Finance and leadership | Media buyers | Creative strategists and buyers |
| Blind spot | Cannot say which channel or ad did it | Platforms overlap in credit; cannot say which creative | Thin data on small creatives; needs grouping |
| In the example below | 3.8:1 | Meta 4:1, TikTok 2.4:1 | Hero A 5.7:1, bottom nine 0.96:1 |
Blended ROAS vs ROAS. Blended ROAS uses your store's total revenue, so it catches sales the pixel misses and is not inflated when two platforms both claim one order. It also counts sales no ad caused, such as repeat buyers and email. Many teams use blended ROAS and MER (marketing efficiency ratio) as the same number; some put all marketing cost, not only ad spend, under MER. Say which one your report means. The rule that keeps all three honest is scope: divide revenue and spend from the same scope, never one of each.
A worked example: one month, three answers
A brand closes the month with $570,000 in total revenue against $150,000 of ad spend: $100,000 on Meta, $30,000 on TikTok and $20,000 on Amazon DSP. Blended ROAS is 3.8:1. Meta attributes $400,000 (4:1) and TikTok $72,000 (2.4:1). The DSP spend is awareness and streaming TV, judged on delivery rather than ROAS. Every number so far looks fine.
Now break Meta out by creative:
| Creative | Spend | Attributed revenue | ROAS | Share of spend |
|---|---|---|---|---|
| Hero A | $42,000 | $256,000 | 6.1:1 | 42% |
| Hero B | $18,000 | $79,000 | 4.4:1 | 18% |
| Mid C | $12,000 | $38,000 | 3.2:1 | 12% |
| Bottom 9 combined | $28,000 | $27,000 | 0.96:1 | 28% |
| Meta account | $100,000 | $400,000 | 4:1 | 100% |
Hero A also runs on TikTok. Counted as one creative across both platforms:
| Hero A | Spend | Attributed revenue | ROAS |
|---|---|---|---|
| On Meta | $42,000 | $256,000 | 6.1:1 |
| On TikTok | $8,000 | $30,000 | 3.8:1 |
| Creative total | $50,000 | $286,000 | 5.7:1 |
Three facts fall out that neither the blended nor the platform number can express. The bottom nine creatives, together, return less than they cost, on 28 percent of Meta's budget, before margin is considered. One asset produces 64 percent of Meta's attributed revenue and 42 percent of TikTok's. And without Hero A, TikTok's other creatives sit at 1.9:1. Do not add the platform figures to the blended one: platform credit overlaps and misses sales, which is exactly why the scopes stay separate.
Why does the blended number mislead?
Because an average reports the outcome and hides the inputs. ROAS at any level above the creative is spend-weighted, so a few heroes carrying most of the budget hold it up while most creatives sit below breakeven. Remove Hero A and the same Meta account reads 2.5:1. Nothing about the strategy, the audience or the offer changed. The account was never a 4:1 business; it was one creative's performance wearing an account's clothing.
Blended ROAS adds a second layer of cover, since repeat customers and organic sales sit in the numerator whatever the ads do. It is the right number for the board and the wrong one for deciding which creative to make next.
What does the spread tell you?
The distribution of creative ROAS is more informative than any average, and it resolves into three shapes.
Concentrated. A few creatives carry most of the spend and the return, as above. Healthy today and fragile on a schedule, because heroes fatigue. The binding risk is succession: when the hero decays, the average follows it down, and the replacement has to exist already.
Flat and acceptable. Most creatives cluster near target with no runaway winner. Scaling is a budget decision rather than a creative one, the most comfortable position in paid social and the rarest.
Flat and poor. Everything clusters below target. There is nothing to reallocate toward, so this points at the offer, the landing page, the audience or tracking, and shuffling budget between equally weak assets fixes none of them.
Track the shape month over month. An account moving from flat and poor to concentrated is making progress even if blended ROAS barely moved: it found something that works and now needs more of it.
How do you calculate creative ROAS?
The arithmetic is simple; the grouping is the work.
- Export ad-level performance with spend and attributed revenue for the window, under one attribution setting for the whole pull.
- Group ads by the underlying creative asset, not by ad ID. The same video typically runs in several ads, ad sets and campaigns, often on more than one platform. Every one of those rows belongs to one creative.
- Sum spend and revenue within each group, then divide once. Averaging per-ad ROAS values gives a different, wrong answer, because it weights a $50 ad the same as a $40,000 one.
- Apply a readability floor and mark everything under it undecided rather than failing.
The floor prevents the most common misreading. ROAS stabilizes with conversion count, not spend: a creative with two purchases has a ratio that will swing on the third. Set the floor in conversions where your reporting allows, something like 15 to 25 attributed conversions, and treat creatives still inside their test window the same way, as in creative hit rate measurement.
One scope caveat: this assumes conversion-objective spend. Awareness and delivery buying, including most Amazon DSP and streaming TV, is not trying to produce attributed revenue in the window, so a low ROAS there is a category error. Separate those buckets first and judge delivery on CPM, reach and pacing.
In Peachblue, steps two and three are done for you. Creatives are grouped by perceptual fingerprint, so the same asset is recognized across ads, campaigns and platforms even after re-crops and re-compression, and identical Amazon DSP assets collapse into one. The Low-data preset in the Creative Library keeps creatives below a readable volume apart from the losers.
What do you do with the answer?
Act on the shape, not on individual ratios.
Cut the confirmed bottom, not the undecided. Creatives past the floor and well below breakeven fund nothing. Creatives under the floor are unfinished experiments, and killing them is how teams stop testing by accident.
Do not simply move budget to the hero. It is often already at the spend where its efficiency decays, and more budget speeds up saturation. Concentration is the risk to reduce.
Iterate from the top. New hooks on a proven body, new formats of a proven concept, new angles on the message that worked. Iterating from a winner has a better hit rate than inventing cold.
Watch the hero for decay. It shows in CTR against the creative's own best week before it shows in ROAS; the diagnosis is in the creative fatigue guide.
Fund the bench. Succession is the whole game in a concentrated account. How much to spend finding the next hero is in the creative testing budget guide.
In Peachblue, ROAS is one of four inputs to a percentile composite score, with CTR, CPA and spend, so a high ratio on thin volume does not crown a winner by itself. The Creative Economics dashboard shows wasted spend on confirmed losers, bench depth and hit rate, and its fatigue board reads each creative against its own best week. The Next Creative Brief turns what your top creatives share into the next iteration batch.
Why not just use Ads Manager or a spreadsheet?
Because both count ads, not creatives. Ads Manager reports per ad, inside one platform, so Hero A's TikTok half never meets its Meta half. A spreadsheet can group them, but only through naming conventions maintained by hand, and conventions drift: assets get re-cropped for placements, re-cut shorter, re-compressed on upload and re-uploaded under new names. Each becomes a row that looks like a new creative.
The failure is specific and expensive. The hero fragments into four half-heroes, none clears the readability floor, and the asset carrying the account never appears as a winner. You then optimize against a picture in which your best creative does not exist.
That grouping problem is what Peachblue exists to solve, and it is why creative ROAS can be kept current from your daily data across every platform you connect, instead of rebuilt once a quarter. You can also ask Agent Peach, or Claude through the MCP server, which creatives carry the account's ROAS and get the ranking from your own data. The honest boundary: Peachblue reads the revenue each platform attributes and does no attribution modeling of its own, so blended ROAS stays a number for your store and finance tools.
Start with the export either way. Pull one month, group it by hand, and look at the shape. Most teams find the blended average was the least interesting number in the file.