Paid Media ROI: How to Measure What Your Ad Spend Returns

Sooner or later someone asks whether the paid media budget is paying back. It is a fair question, and a harder one than it looks. Google and Meta each report a healthy return on ad spend, the finance team sees a different number, and nobody can say which campaigns would be missed if they stopped.

This guide explains how to measure paid media ROI in a way the finance team can check. It covers the formulas and where each one misleads, how to set a target from your own margin and which metrics to watch. It then explains the conversion data the platforms need, what attribution can and cannot tell you, and how to decide where the next dollar should go.

DisclosureinBeat, which is mentioned in this article, is part of Fieldtrip, a group of marketing agencies in which Jake Jorgovan is a partner.

ROI and ROAS: the formulas and where they mislead

Return on investment compares the gross profit a campaign produced with what it cost. Gross profit here means the sales minus the cost of the goods or services sold, before any campaign costs:

ROI (%) = (gross profit from the campaign minus the campaign's full cost) ÷ full cost × 100

Return on ad spend compares the revenue the ads are credited with to the ad spend alone:

ROAS = revenue credited to the ads ÷ ad spend

The difference matters more than it seems. ROAS uses revenue, and it leaves out what you paid for the goods, the agency and the creative. A hypothetical example shows how far apart the two can be. You spend $2,000 on ads, and the platform credits them with $10,000 in sales:

• ROAS is $10,000 ÷ $2,000 = 5, which Google Ads expresses as 500% when you set a Target ROAS.
• If you treat revenue as the return, ROI looks like ($10,000 minus $2,000) ÷ $2,000 = 400%. That figure ignores what the goods cost, so it is not a profit-based ROI.
• With a 40% gross margin, those sales earned $4,000 in gross profit. After the $2,000 of ad spend, ROI is $2,000 ÷ $2,000 = 100%.
• If creative and campaign management cost another $1,000, the full cost is $3,000 and ROI falls to about 33%.

Image

Zoom

Illustrative scenario from the article; figures are assumptions, not market benchmarks.

All four numbers come from the same sales and spend, and they answer different questions. Report ROAS beside ROI, and say which costs the ROI includes. Both are returns on the sales the platform credited; a lift test, covered below, checks how much of that the ads caused.

Set the target from your margin

There is no good ROAS in general. A ROAS of 3 can be very profitable for a business with an 80% gross margin and lose money for one with a 25% margin, because the same revenue leaves very different gross profit behind.

The break-even point follows from your margin: break-even ROAS = 1 ÷ gross margin. At a 40% margin you need 2.5 just to cover the ad spend. At 25% you need 4. At 80% you need 1.25. Your target sits above that line, by enough to cover the agency, creative and tools, plus the profit you expect. In the example above, adding the $1,000 of creative and management means sales must reach $7,500 to break even (40% of $7,500 is $3,000), a ROAS of 3.75 on $2,000 of media.

For customers who buy more than once, set the ceiling with lifetime value. Use the margin a customer brings over a period you are willing to wait for, and decide how many months of payback the business can fund. The guide to lifetime value and cost per acquisition covers that trade-off in more detail.

For B2B, a form fill is not revenue. Work back from the deal: the share of leads that become qualified opportunities, the share of those that close and the margin on the average deal. That gives a value per lead you can bid against, and a reason to send the platforms better data (covered below).

Match the metrics to the campaign's job

Every campaign needs one primary measure tied to its goal. The others are diagnostics: they explain why the primary measure moved.

Click-through rate and impressions are easy to raise and say little about profit. An ad with fewer clicks and a higher conversion rate can be worth more than one that draws curious clicks.

For video, two ratios built from Meta's own metrics show whether the opening and the message are working. Meta counts a 3-second video play (opens in a new tab) when a video plays for at least three seconds (or nearly its full length if it is shorter), and a ThruPlay when it plays to completion or for at least 15 seconds. This guide uses two ratios built from them: the hook rate (3-second plays ÷ impressions) and the hold rate (ThruPlays ÷ 3-second plays). Neither ratio appears among the video metrics Meta lists, so write down the definition your team uses and keep it. For comparable videos, a high hook rate with a weak hold rate can suggest an opening that draws attention and a message that loses it.

Image

Zoom

The approach described in this article, shown as a visual guide.

Google's Quality Score (opens in a new tab) belongs with the diagnostics too. Google describes it as a diagnostic of ad and landing page relevance, scored from 1 to 10 for each keyword. It also says Quality Score "is not a key performance indicator and should not be optimized or aggregated with the rest of your data."

Get the conversion data right first

Every formula above depends on counting the right conversions, and the platforms optimize toward whatever you tell them a conversion is.

1. 1

   Define the outcomes in Google Analytics

   Google Analytics 4 measures key events (opens in a new tab): events you mark as important, such as a purchase, a qualified lead form or a booked call. A standard property can mark up to 30. Marking an event does not change historic data, so set them up before a campaign starts. To bid on a key event in Google Ads, create a Google Ads conversion from it and check its settings there. Older instructions to create "Goals" refer to Universal Analytics, whose standard properties stopped processing new data from July 1, 2023 (opens in a new tab).

2. 2

   Tag every paid link the same way

   Google says you should always use utm_source, utm_medium and utm_campaign (opens in a new tab) on campaign links, and recommends one medium value per channel, such as paid_social. Agree the naming before launch. Inconsistent tags split one channel across several rows of the report.

3. 3

   Send real business outcomes back to the platforms

   For sales that close by phone, in a meeting or after a long cycle, import them. Google Ads can import offline conversions (opens in a new tab) and recommends starting with enhanced conversions for leads, which match hashed customer details to the ad interaction. Meta's Conversions API (opens in a new tab) connects events from your server or CRM, including offline conversions, to Meta's ad systems. For B2B, send qualified opportunities and closed revenue, so the platform can learn which ads bring buyers.

4. 4

   Check what your bidding optimizes toward

   Start with the conversion goals each campaign uses. Google bids on the primary conversion actions (opens in a new tab) in those goals, and on every action in a custom goal, even one marked secondary. Target ROAS (opens in a new tab) needs conversion values. For Search and Shopping campaigns, Google requires at least 15 conversions in the past 30 days at the conversion-tracking level. Google also warns that a Target CPA (opens in a new tab) set too low "may cause you to forgo clicks that could result in conversions, resulting in fewer total conversions."

Attribution: how Google and Meta assign credit

Attribution assigns credit for a conversion to the ads and clicks before it. You may still see first-click, linear and time-decay models recommended, but Google's tools no longer offer them. Google Analytics notes that the first click, linear, time decay and position-based models "are no longer available as of November 2023" (opens in a new tab), leaving data-driven attribution and two last-click models. Google Ads moved conversion actions on those models to data-driven attribution, its default for most conversion actions, and still supports last click.

Google Ads' data-driven attribution works with any volume, but Google recommends at least 200 conversions and 2,000 ad interactions (opens in a new tab) in supported networks within 30 days for the model to assign credit well. Google Ads' Model comparison report (opens in a new tab) puts cost per conversion and conversion value per cost side by side under two models, which helps you find campaigns that last click undervalues.

Meta lets you choose an attribution model (opens in a new tab) for each ad set. Standard attribution credits conversions inside a time window: within 1 or 7 days of a link click, within 1 day of an ad view or within 1 day of other engagement, for website and in-store conversions. Incremental attribution uses models that predict whether an ad caused the conversion; it is an estimate, which a lift test can check. Meta warns that results in its Campaign Overview table "cannot be compared" across ad sets that use different attribution models.

Two consequences follow for anyone reading the reports:

• Each platform credits its own ads under its own rules, so the same sale can appear in both Google's and Meta's numbers. Adding platform conversions together can overstate the total. Reconcile them with orders or closed revenue in your own systems.
• Changing the attribution model changes what automated bidding chases, because Google's bid strategies use the conversions column the model fills. Test a change before rolling it out.

Test what the ads actually cause

Attribution assigns credit under a model. On its own, it does not show whether the sale would have happened anyway, which is the question behind every budget increase. A controlled lift test estimates the ads' effect by comparing a group that can see your ads with a control group held out from them. Meta reports its lift results with intervals showing the range of most probable outcomes; check that range before you use a result to raise the budget.

For a campaign running 180 days, Meta's guide doubles the thresholds to $10,000 and 1,000 conversions, and the alternative data-source route is unavailable to advertisers mainly targeting countries covered by the EU ePrivacy directive.

Both mix models are open source. Robyn's documentation describes calibrating the model with "ground-truth" methods such as geo tests and lift studies, and Google offers geo experiments to calibrate Meridian. Treat a mix model as a planning tool and check it against experiments.

You do not need all of these. A sensible start is one lift test a year on your largest channel, timed before a decision to scale it, with platform reports used for week-to-week management in between.

In B2B the numbers will always be partial. On the Leaders of B2B Podcast (opens in a new tab) in December 2024, I pointed out that it is not uncommon for a B2B company to close 20 deals a year at half a million dollars each. The few interactions that move one of those deals, such as a piece of content or a customer testimonial, can't be tracked in the data. My view was that "there's intuition plus data as a guiding light, but it can't be the full piece there." Ask new customers what brought them in, and read the answers beside the reports.

Raise the return before you raise the budget

Once the measurement is sound, four levers are worth working on before you add budget. Each one shows up in a measure you already track.

• The landing page. At the same cost per click, doubling a page's conversion rate from 1% to 2% halves the cost per acquisition. Check that the page matches the ad's message, loads quickly on a phone and gives the visitor one clear next step.
• The creative. Test one variable at a time, such as the hook, the offer or the format, against the current best ad. People tire of a creative they have seen too often, or of a whole format; the guide to user fatigue in social ads covers how to tell which is happening.
• The audience. Exclude existing customers and recent converters from prospecting campaigns. Remarketing to recent visitors can report a high ROAS even though some of those people may have bought anyway, which makes it a good candidate for a holdout test.
• The bid target. Google Ads recommends a starting Target ROAS from your actual ROAS over the last few weeks. Check that figure against your break-even ROAS before you accept it.

Decide where the next dollar goes

The measures above exist for one decision: where to put more money, and where to take it away.

The most useful lesson I have on this is about focus. In a 2021 episode of my podcast, I described the goal as finding a channel that produces results, building a system around it and scaling it "until it reaches a point of saturation." Saturation, I said, is the point "where there's kind of like diminishing returns." My example was Google Ads: once your best keywords have all the investment they can use, it is time to test something else. Describing the first channel that worked at Content Allies, I put the test simply: "I know that I can put money into channel one and positive money will come out of it." I suggested giving a new channel about three months to show its first results.

Saturation is where average numbers mislead. As a channel approaches it, the quarter's ROAS can still look healthy while the last few thousand dollars added little, because the average blends the cheap early conversions with the expensive later ones.

1. Raise budgets in steps and compare equal periods, allowing time for conversions to arrive.
2. Work out marginal CPA: the extra spend divided by the extra qualified conversions. Work out marginal ROAS: the extra revenue, reconciled with your own records, divided by the extra spend.
3. Compare marginal ROAS with your target ROAS. Compare the cost per qualified lead or opportunity with a ceiling for that same stage, based on its close rate and the expected gross profit per customer. For cost per new customer, include the other acquisition costs before comparing it with your CAC ceiling.

Image

Zoom

The approach described in this article, shown as a visual guide.

1. Where other changes could explain the difference, use a holdout. When the added return misses your target, stop adding spend there and test the next channel. More total conversions alone do not justify the increase.

Questions for your next budget review

These questions cover the whole page, from costs to lift tests. Use them with your team or agency before the next budget change.

Common questions

What is a good ROI for paid media?

An ROI above 0% means the program covered every cost included in the calculation. Set your target higher, for the profit and payback the business needs. On the ROAS side, media-only break-even is 1 ÷ gross margin, and including agency, creative and tools raises it. A published average from other companies does not reflect your margin, prices or sales cycle.

How often should we review paid media ROI?

Review platform metrics weekly and reconcile them with revenue every month. For Google Ads campaigns using Target CPA, Google recommends (opens in a new tab) evaluating the last 30 days with at least 30 conversions. For other campaigns, choose a window that fits the conversion volume and sales cycle, and allow for conversions that arrive late. Give a new channel about a quarter to show first results, and let the sales cycle finish before judging payback.

Which tools do we need to track paid media ROI?

A short list covers it: Google Analytics with key events and consistent UTM tags, each ad platform's own reporting, and your CRM or order system as the record of revenue. Reconcile the platforms against that record each month. A lift test or mix model comes later, for the largest budgets.

What is the difference between ROAS and ROI?

ROAS divides the revenue credited to ads by the media spend. ROI divides the gross profit left after the full program cost by that cost. Use ROAS to compare campaigns' revenue efficiency and ROI to judge whether the program makes money.