ROAS：比率变了，业务却没变

Return on ad spend (ROAS) is conversion value divided by ad cost, the ratio Google Ads reports as conversion value per cost. A 4x ROAS means each dollar of ad cost was credited with four dollars of conversion value, which may be revenue, an assigned lead value, or gross profit.

That last clause is where most ROAS confusion starts. The arithmetic is simple. The inputs are not, and three of them can move the number while sales, margins, and customers stay exactly where they were: the value basis in the numerator, the attribution window a platform applies, and the habit of reading attributed value as value the ads caused.

How ROAS Is Calculated

Divide total conversion value by total ad cost for the same campaigns and the same period. Google Ads documents the metric as "Conv. value / cost" and describes it as an estimate of return on investment (VERIFIED: Google Ads Help). The ratio only behaves like a return on investment if the value you feed it already reflects profit.

An illustrative example: a campaign spends $2,000 and is credited with $9,000 in conversion value. ROAS is $9,000 divided by $2,000, or 4.5.

Multiple, Ratio, and Percentage

The same result can be written as 4.5x, 4.5:1, or 450%. Pick one and state it in every table header.

For click-based campaigns, ROAS equals conversion rate multiplied by average value per conversion, divided by cost per click. So a falling ROAS can come from fewer conversions per click, smaller orders, or pricier clicks, and each calls for a different fix.

Rolling Up Campaigns: Total Value Over Total Spend

Across several campaigns, divide total value by total cost. Do not average the individual ratios.

Take two illustrative campaigns. Campaign A spends $1,000 and is credited with $6,000, a 6x ROAS. Campaign B spends $9,000 and is credited with $18,000, a 2x ROAS. The simple average is 4x, but the real combined figure is $24,000 divided by $10,000, or 2.4x. Averaging lets the small campaign inflate the whole account.

What Counts in the Numerator

The platform formula says conversion value, not revenue. Conversion value is whatever number the advertiser sends or configures for each conversion, which is why two accounts reporting "4x" can describe completely different economics.

Revenue, Assigned Value, or Gross Profit

Ecommerce accounts usually pass order revenue. Lead generation accounts often assign a fixed value to each form fill or call, based on how often leads become customers and what those customers are worth. Some advertisers report profit instead. Google Ads can produce gross profit metrics by combining cart data with cost of goods sold supplied through Merchant Center, for supported campaigns (VERIFIED: Google Ads Help, cart data documentation).

Each basis answers a different question. Assigned-value ROAS is only as good as the assumptions behind the lead value. Gross profit ROAS gets closest to whether the ads paid for themselves, though it still leaves out shipping, returns, and overhead unless those are built into the value. Whichever you use, write down the basis and whether tax, shipping, discounts, and refunds are included.

When the Value Basis Changes, the Trend Line Breaks

This is the first way ROAS moves without the business moving. Suppose an account spends $3,000 and is credited with $12,000 in revenue, a 4x ROAS. The team switches the reported value to gross profit, and the average gross margin is 35%. The same sales now carry $4,200 in value, so ROAS reads 1.4x.

Customers, prices, and spend are unchanged. Any chart that runs across the switch date compares two different metrics under one name. Annotate the break, restate history on the new basis if you can, and revisit any bidding target set under the old one.

ROAS vs ROI, CPA, and CAC

Return on investment (ROI) compares profit with the full investment that produced it. ROAS compares conversion value with ad cost alone. A campaign can show a comfortable revenue ROAS and still lose money once product costs, fulfillment, platform and agency fees, and creative production are counted. What belongs in ROI's cost base is a topic for its own entry.

Cost per acquisition (CPA) divides ad cost by the number of conversions. It ignores value, which makes it a good guide when every conversion is worth about the same and a poor one when order values vary widely. For a fixed average value per conversion, a lower CPA means a higher ROAS.

Customer acquisition cost (CAC) usually covers more than ad spend and counts only new customers. A campaign that mostly reaches existing buyers can post a strong ROAS while doing little to lower CAC.

In short: ROAS measures the value ad spend was credited with, CPA what each conversion cost, CAC what each new customer cost, and ROI whether the whole investment earned a profit.

Break-Even ROAS and Why There Is No Universal "Good" Number

A "good" ROAS depends on margin, on which costs the value already reflects, and on how long the business can wait to earn back acquisition cost. A single benchmark ignores all three.

The 1 Divided by Margin Rule

For revenue-based ROAS, break-even is 1 divided by contribution margin, where contribution margin is the share of revenue left after variable costs other than advertising. This is algebra, not a published standard.

With an illustrative 30% contribution margin, break-even ROAS is 1 divided by 0.30, about 3.33x or 333%. Above that, the ads cover their own cost on the attributed figures, though not fixed costs, and not necessarily on an incremental basis. If the reported value is already gross profit, break-even sits at 1x.

Why One Account-Wide Target Misprices a Mixed Catalog

Consider two product lines. One carries a 50% margin, so its break-even ROAS is 2x. The other carries 20%, so its break-even is 5x. An account-wide target of 3x underinvests in the first line, where 2.5x would still be profitable, and loses money on the second. Separate targets, or profit-based values, fit the economics far better.

Why ROAS Differs Between Reports

An ad platform, an analytics tool, and a finance spreadsheet can each report a different ROAS for the same week without any of them being broken.

Attribution Models, Windows, and Channels

The marketing attribution rules applied to a conversion decide how its value is split across touchpoints. A platform report usually credits its own ads. A web analytics tool may spread the same sale across search, email, and direct visits, so each channel's ROAS comes out lower there. Add up platform-reported values from several ad platforms and the total can exceed the sales the business actually recorded.

The window matters as much. A seven-day click window can credit the same or fewer sales than a 28-day one, and a view-through window credits sales to people who saw an ad but never clicked. Before comparing two figures, check that model, window, conversion actions, and currency match.

Platform Window Changes (as of October 2026)

Windows are not always the advertiser's choice. Meta announced that from January 12, 2026, its Ads Insights API would stop returning the 7-day view and 28-day view attribution windows. The remaining windows are 1-day click, 7-day click, 28-day click, 1-day engaged view, and 1-day view (VERIFIED: Meta for Developers blog, October 16, 2025; Meta Ads Insights API only).

This is the second way ROAS moves without the business moving. A report that previously pulled a longer view window through the API can now credit fewer sales to the same ads.

Attributed ROAS vs Incremental ROAS

Attribution assigns credit for a sale that happened. It does not say whether the sale would have happened without the ad. That second question belongs to incrementality, and reading attributed ROAS as an answer to it is the third way the number misleads.

What Attribution Cannot Prove

Gordon, Moakler, and Zettelmeyer compared observational measurement methods with randomized experiments across 663 advertising experiments run on Facebook. The observational methods could not reliably recover the causal effects the experiments measured (VERIFIED: Marketing Science, 2023). Two of the authors were part-time Facebook employees while the research was conducted.

Earlier field experiments on eBay's paid search by Blake, Nosko, and Tadelis found that brand keyword ads had no measurable short-term benefit, with traffic shifting to unpaid links when the ads were paused (VERIFIED: NBER Working Paper 20171, 2014, published in Econometrica, 2015; US). It was one very well-known advertiser, so read it as a warning about brand search ROAS, not a general rule.

A credited sale is not a caused sale. As analysis, the campaigns most exposed to that gap are the ones that reach people already close to buying, such as brand search, because the sale they are credited with was likely to happen anyway.

Lift Tests as the Check

Google Ads separates attributed conversions from incremental ones through lift studies. Its Conversion Lift uses a Bayesian method that draws on historical campaign data as priors. As of October 2026, Google Ads Help lists result thresholds of 150 conversions in the treatment group and 65 in the control group, with a minimum spend of US$5,000 (VERIFIED: Google Ads Help, Conversion Lift Bayesian methodology). One point of analysis: those priors come from the same platform whose ads are being measured.

Marketing mix models offer another lens. Google's Meridian documentation defines ROI as the incremental outcome a channel drives relative to that channel's media cost, and marginal ROI as the return on one additional unit of spend (VERIFIED: Google Meridian documentation, vendor documentation). The marginal version matters because a campaign's average ROAS says little about what the next dollar will return.

Target ROAS in Automated Bidding

With Target ROAS bidding, Google Ads tries to achieve "an average return on ad spend (ROAS) equal to your target." The help text adds that some conversions may come in above the target and some below, while the system aims to keep overall conversion value per cost at the level set (VERIFIED: Google Ads Help, Target ROAS documentation).

Target vs Achieved ROAS

"Average" is the operative word. A target does not set a floor for every auction. It steers the mix.

The consequence, offered as analysis rather than a platform statement, is that a higher target pushes the system toward the conversions it expects to be most valuable for the cost, and there are usually fewer of those. A higher target can therefore mean a higher ratio on lower total value.

Budget interacts with targets too. Google announced that from August 17, 2026, Limited by budget campaigns using target CPA or target ROAS would perform more consistently toward the bid target, including after budget changes. Google recommended updating targets to match recent performance before that date and does not adjust them automatically (VERIFIED: Google Ads Help, as of October 2026). A target set deliberately low, in the hope of beating it, now pulls results toward that lower figure.

Frequently Asked Questions

What is a good ROAS?

There is no universal figure. A good ROAS is one above your break-even point, which for revenue-based ROAS is 1 divided by contribution margin.

What is the difference between ROAS and ROI?

ROAS compares conversion value with ad cost. ROI compares profit with the full investment, including product costs, fees, and other expenses ROAS leaves out. A campaign can have a healthy revenue ROAS and a negative ROI.

What is break-even ROAS?

It is the ROAS at which attributed contribution exactly covers ad cost. For revenue-based ROAS it equals 1 divided by contribution margin, so a 25% margin gives a break-even of 4x. If conversion value is already gross profit, break-even is 1x.

Why is my ROAS different in Google Ads and Google Analytics?

The two tools can count different conversions, credit different channels, and apply different attribution models and windows. Google Ads tends to credit its own ads, while an analytics tool may spread the same sale across several channels.

Is a higher ROAS always better?

No. A higher ratio can come with lower total sales and fewer contribution dollars, especially when it is reached by raising a bidding target. On its own it also says nothing about incrementality.

Related Terms

• Return on Investment (ROI): profit relative to the full investment.
• Customer Acquisition Cost (CAC): the cost of winning a new customer.
• Cost per Acquisition (CPA): ad cost divided by conversions.
• Incrementality: the outcome advertising caused, measured against a holdout.
• Marketing Attribution: the rules that split credit across touchpoints.
• Conversion Rate: the share of clicks or visits that convert.
• Web Analytics: site-side measurement that may credit a sale differently.
• Cost per Mille (CPM): the price per thousand impressions, one way ad cost is bought.
• Pay-Per-Click (PPC): a pricing model where the click is the billable event.

Sources

• Google Ads Help, answer 13405059, conversion value per cost ("Conv. value / cost"). Platform documentation. VERIFIED. https://support.google.com/google-ads/answer/13405059.
• Google Ads Help, answer 6268637, Target ROAS bidding. Platform documentation. VERIFIED for the average-target wording; the volume trade-off is analysis. https://support.google.com/google-ads/answer/6268637.
• Google Ads Help, answer 9028254, "About conversions with cart data." Platform documentation, supported campaigns only. VERIFIED. https://support.google.com/google-ads/answer/9028254.
• Google Ads Help, answer 16627074, "About Bayesian methodology in Conversion Lift." Platform documentation, as of October 2026. VERIFIED. https://support.google.com/google-ads/answer/16627074.
• Google Ads Help, answer 17061251, "Changes to target based bid strategies." Platform documentation, effective August 17, 2026. VERIFIED. https://support.google.com/google-ads/answer/17061251.
• Meta for Developers, "Ads Insights API metric availability updates," October 16, 2025. Platform developer announcement, Meta Ads Insights API only. VERIFIED. https://developers.facebook.com/blog/post/2025/10/16/ads-insights-api-metric-availability-updates/.
• Google Meridian documentation, default prior parameterizations (ROI and marginal ROI). Vendor documentation. VERIFIED. https://developers.google.com/meridian/docs/advanced-modeling/default-prior-parameterizations.
• Gordon, B., Moakler, R., and Zettelmeyer, F. (2023), "Close Enough? A Large-Scale Exploration of Non-Experimental Approaches to Advertising Measurement," Marketing Science 42(4), 768 to 793. Peer-reviewed journal; two authors were part-time Facebook employees during the research. VERIFIED. https://doi.org/10.1287/mksc.2022.1413.
• Blake, T., Nosko, C., and Tadelis, S., "Consumer Heterogeneity and Paid Search Effectiveness: A Large Scale Field Experiment," NBER Working Paper 20171, May 2014; published in Econometrica 83(1), 2015. Peer-reviewed journal, United States. VERIFIED. https://doi.org/10.3982/ECTA12423.