Google Ads ROI: how to calculate the real return, not just ROAS

For marketing ROI as this guide models it, subtract direct and variable fulfilment costs from attributed revenue. Then subtract the full campaign investment, media and all, and divide what remains by that investment.

ROAS only divides attributed revenue by media spend. A return with no stated cost, attribution and time basis is a number, not a result.

What is ROAS, and how is ROI different?

ROAS has one numerator and one denominator, and both are narrow. Marketing ROI, as this article uses the term, is wider on both sides.

The denominator is a choice. Other people make a different one.

Google's own ROI page works an example with six units costing $100 each. They sell for $200 each from $200 of ads. The page calculates the return as $1,200 less $800 of costs, divided by the $800. That's 50%. So Google's denominator is production costs plus advertising.

This article divides by the marketing investment alone and says so. The question a business owner asks is what the campaign returned on the money put into the campaign. Neither convention is wrong. Presenting them as the same formula is.

Four words also need separating:

• Revenue is what was sold.

• Gross profit is revenue less cost of sales.

• Contribution is revenue less the direct and variable costs this model allocates. It may or may not match the accountant's gross profit line.

• Company net profit is what is left after every overhead and tax. Nothing in this article calculates it.

The margin used throughout is contribution margin, m, calculated as C divided by R. It's a share of revenue, not a markup on cost. A business that marks up by 50% has a 33% margin. Getting that wrong flatters every result that follows.

Ask the business for its actual cost classifications rather than borrowing a margin from its industry.

Gather revenue, margin and campaign costs

The ledger has three parts. Each item goes in one of them exactly once.

Revenue, R, is attributed net revenue. That means sales after discounts, cancellations, refunds and bad debt adjustments. Use the accounting basis the business uses, for the cohort or period being measured. Deduct each reduction once. A refund taken off net revenue is not taken off again inside the margin.

Fulfilment costs, D, are the direct and variable costs of delivering that revenue. Product cost, materials, direct labour and subcontractors all go here. So do payment fees, shipping and fulfilment, and variable sales handling or commissions. Each goes into one declared bucket only, never into both delivery and acquisition.

Campaign investment, I, is media charges, A, plus everything else the campaign required, F. That covers:

• management fees

• setup

• tracking and call tracking

• creative

• the landing page work that exists because of the campaign

• software

• feed work

• an allocation of shared support

Our guide to how much Google Ads costs covers the costs that are easy to leave out. What agencies charge for management is covered in our guide to Google Ads management pricing.

Record how a one off build or a shared tool is spread across the reporting horizon. There are two errors to avoid. One is leaving setup out because it won't recur next month. The other is charging the office rent or an entire old website to one campaign.

For a future spend decision, model the costs that will change. That includes step increases in staff or tools. Keep that prospective model separate from the historical fully loaded report.

Use one currency and one tax treatment. Google's tax page says Google Ads sales carry 10% GST for accounts with an Australian business address.

The ATO's page on claiming GST credits sets the conditions. You must be registered for GST. The purchase must be for business use, with GST in the price. Purchases over $82.50 also need a tax invoice.

Where the credit is recoverable, exclude the GST from the cost and the GST collected from the revenue. Keep any irrecoverable GST in the cost basis. Registration alone does not make every purchase creditable. So the bookkeeper confirms the treatment rather than this article prescribing one.

Every step reconciles. $100,000 less $68,000 of product and variable fulfilment costs leaves $32,000 of contribution. The chart shows them as product costs. Then $20,000 of media, $3,000 of management and $2,000 of other campaign costs leave $7,000.

That final figure is contribution after the included costs. It is not company net profit. Product costs, management and other costs are the steps a ROAS figure never sees.

Calculate the break even return

Once m and I are known, the threshold falls out on the same basis.

Take A at $10,000, F at $2,500 and m at 0.40. Media only break even is 2.50x. Full campaign break even is 3.13x and needs $31,250 of revenue.

That is zero marketing ROI on the included costs. It is not a sufficient business target and not proof that company overhead is covered. The target formula assumes the same cost allocation and the same margin hold at the new volume. They often don't.

Three cases return no number. A report should say so rather than print a zero. If media spend is zero, ROAS is undefined. If investment is zero, this ROI ratio is undefined.

If m is zero or negative, no positive revenue multiple covers a positive investment under the unchanged model. The problem is then the pricing or the cost of delivery, not the campaign.

The calculator returns revenue ROAS, a basic margin threshold and an additional contribution scenario. Its margin field is labelled gross margin. To make its threshold match the contribution basis here, enter the margin after the variable costs in D. Treat the result as a campaign check, not a company net ROI.

Follow leads through to customers

For a lead generation business, the revenue arrives weeks after the click. It comes through a sequence of stages. Each one has to be measured on its own denominator.

Two arithmetic traps live in that table. Multiplying an overall lead to customer rate by the qualification share again counts qualification twice. And a zero denominator is not a zero cost. No customers yet means CAC is undefined, not nil.

Here is the example electrician cohort from the worked models below, one stage at a time.

Each drop has a different cause. Spam and duplicates take 40 raw leads down to 30 genuine ones. Fit with the offer decides how many qualify. Follow up decides how many book. Cancellations or unpaid invoices take 16 booked jobs down to 12 paying ones.

The denominator changes the answer as well. $4,500 of media is $112.50 per raw lead but $150 per genuine lead. So name the denominator every time.

Distinguish completed calls, form submissions, qualified opportunities and customers at every step. Treat rates from immature or tiny cohorts as uncertain. Set the target lead cost from contribution rather than from a benchmark. Our guide to a good cost per lead for Google Ads explains how. Industry figures, such as our Google Ads cost per lead by industry table, are sanity checks, not targets.

Calls are easy to miss in the count. Our guide to call tracking for Google Ads covers measuring them.

Three worked Google Ads return models

All three are example Australian dollar models on a consistent ex GST basis where GST is recoverable. Each lead cohort has matured. Each customer outcome is attributable under the stated model. The margin and costs correspond to that cohort. They are not monthly market benchmarks and not Elev8d client results.

Electrician

Qualification is 80% and the close rate among qualified leads is 50%. The overall close rate from genuine leads is 40%. Twelve jobs at $1,200 is $14,400. Every CPL in the ledger divides by the 30 genuine leads.

Contribution is $14,400 x 0.55 = $7,920. Subtracting the $6,000 investment leaves $1,920, or 32% of $6,000. Break even is $6,000 divided by ($4,500 x 0.55). That comes to 2.42x, against the 3.20x achieved.

The spend in the diagram is the full $6,000 investment, not the $4,500 of media. So the costs it generates are fully loaded. The 1.32x multiple it shows is contribution against investment before the investment is subtracted. The ledger does the subtraction.

Professional services

Twenty qualified enquiries from 32 is 62.5%, not a rounded 63%. The denominator stays visible. Six customers is a 30% close rate among qualified enquiries. The $45,000 belongs to those six customers after the stated maturity period. Nothing from matters still in progress is pulled forward into a realised column.

Contribution is $22,500 and the remaining campaign contribution is $12,000. Marketing ROI is 114.29%, a return on the investment and not a profit margin on revenue. This model has both the highest ROAS and the highest ROI of the three.

Ecommerce

Net revenue is $100,000 after the refund adjustment. Product and variable fulfilment costs are $68,000. The $32,000 that remains is contribution before acquisition. Deducting $20,000 of media, $3,000 of management and $2,000 of other costs once leaves $7,000. That is 28% on $25,000, despite a 5x ROAS.

The model does not supply a new customer count. So it cannot also establish a customer acquisition cost. Inventing one would make the ledger look more complete than the data is.

Those are arithmetic outputs from the assumptions above, not sector benchmarks. The order they fall in says nothing about which kind of business should advertise.

Include lifetime value without hiding losses

Start with the losses, because a good looking ratio hides them best.

Take a business that earns $50,000 of revenue at 5x ROAS on $10,000 of media. Its contribution margin is 15%. So the business has $50,000 x 0.15 - $10,000 = -$2,500 before a single fee. That is a loss, however the ROAS looks.

Real accounts do the same. We ran 180 Google Ads account audits from August 2021 to September 2026.

In one, an online store showed 6.1x ROAS over one quarter at an 18% contribution margin. Once fees and other campaign costs were counted, its marketing ROI was -11%. The ROAS was real. So was the loss.

A customer acquired for $500 whose first purchase contributes $400 leaves $400 - $500 = -$100 after that purchase.

Both sides of that comparison use contribution before acquisition. Say the customer's cumulative contribution reaches $400, $800, $1,200 and $1,600 across four purchases. The position after the $500 acquisition cost is then -$100, +$300, +$700 and +$1,100.

The later purchases are a scenario until cohort evidence shows they happen. The scenario has to carry retention and servicing costs, refunds and a stated horizon. Revenue lifetime value cannot be set against contribution CAC.

Discounting belongs where the timing of future contribution materially changes the decision. No ratio of lifetime value to acquisition cost is universally ideal, 3 to 1 included.

Keep three things apart in the report: realised contribution, expected future contribution and cash recovered. Payback depends on when the contribution arrives and not on the lifetime total.

Our customer lifetime value calculator gives a profit adjusted value, CAC analysis and what if scenarios. It works from the gross profit margin you enter. That margin isn't automatically contribution after servicing costs you left out.

Reconcile attribution, cohorts and actual revenue

Before trusting any ratio, read what the account's conversion value means. It could be net sale revenue, gross order value or a profit figure. It could also be an estimated lead value or a bidding adjustment.

Google's Target ROAS page says the strategy predicts future conversions and their values. It uses the conversion values you report. You set values for the conversions you track. Conversion value rules can apply a multiplication factor for higher value customers, devices or locations.

A value shaped for bidding is the right input for bidding. But it's the wrong one for a revenue claim. Report value against cost as what it is. Deduplicate the same order or lead, and reconcile refunds and cancellations before calling anything revenue ROAS. Our guide to Smart Bidding strategies covers the Target ROAS settings.

In the same audits, we reconcile platform attributed revenue for new clients. It has typically been 10% to 30% away from the deduplicated business figure, in either direction.

Attribution allocates credit. Google Analytics' attribution page defines it as crediting key events to ads, clicks and factors along the path. Its data driven and last click models distribute the same conversion differently. All of them exclude direct visits unless the path is entirely direct.

None of that proves the sale would have vanished without the ad. Brand and returning customer activity can carry a strong reported ROAS without creating new demand. Non brand traffic is not automatically incremental either.

Google's Conversion Lift runs a controlled comparison between people who saw ads and people who did not. It is available by contacting a Google account representative. It isn't open to every account. A small campaign should not count on having it.

The blended row is where the arithmetic is easiest to get wrong. So rebuild it from totals.

The combined row is $4,000 over 900 clicks, 104 purchases over 900 clicks and $23,900 over $4,000. All three are calculated from the unrounded totals. Averaging the two displayed ratios instead gives $4.85, 10.75% and 11.75x, a ROAS nearly double the real one.

Then match the reporting clock to the question. Track acquisition cohorts from the lead through qualification, customer status, recognised revenue and collection. Do not divide this month's revenue by this month's media when the leads were acquired three months ago.

State the attribution window, the report dates, the currency and the conversion delay. Keep operational monthly reporting apart from matured cohort reporting.

Business.gov.au's page on accounting methods explains that cash accounting records income when the money arrives. But accrual accounting records the sale when it takes place. So in an accrual business, the accounts do not simply reflect revenue collected. A signed proposal, a recognised sale and a payment received are three different facts.

Use the finance approved revenue basis for the ROI model and a separate cash payback view. Reconcile channel claims against deduplicated business totals so one sale is not full revenue for three channels. We review incoming reports that cover both Google and Meta. Of those, about 1 in 2 reports add the two platforms' attributed sales into one total.

The implementation path is short to describe and slow to build:

• a verified website or call action

• qualification recorded in the CRM

• the customer outcome and its revenue

• an offline connection back to the click

Google lists four ways to track conversions: website actions, app conversions, phone calls and offline conversions. Our guide to Google Ads conversion tracking setup covers the CRM import, its time limits and consent rules.

Improve return and assess the next dollar

Hold the electrician's revenue and investment still and change only the margin. The loss shows as a coral bar below the line.

At 55% contribution margin, C is $7,920, P is $1,920 and marketing ROI is 32%. At 40%, C is $5,760, P is $14,400 x 0.40 - $6,000 = -$240 and marketing ROI is -4%. ROAS is 3.20x in both cases. That's the whole argument of this article in one chart.

The next dollar has its own economics. Take an increase of $1,000 in media with $200 of associated costs. Say it produces $2,000 of additional revenue at the same 55% margin. Its contribution is $1,100, $100 short of the $1,200 incremental investment.

A campaign with a healthy average return does not make that increase profitable. The increment also has to carry any capacity cost and survive the question of whether the demand exists. Our guide to how to scale Google Ads covers marginal returns, capacity and the growth plan itself.

Improving return means improving the verified constraint, and only that. If unsuitable searches are the problem, start with your Google Ads negative keywords. If people click without enquiring, check whether the ad should land on a landing page or your homepage.

Weak qualification, poor follow up, low customer contribution and measurement errors are the other usual suspects. None of those fixes is guaranteed to lift ROI. Can't tell which constraint is binding? Our guide to why Google Ads is not working works through the diagnosis. Where the close rate, the retention or the margin is uncertain, report a range rather than a point.

"Good" return depends on the costs included, the contribution the business needs and the payback period. It also depends on the risk and the capacity. There is no universal average and no 5x target that answers the question. Whether the channel deserves the next investment at all is another question. Our guide to whether Google Ads is worth it for a small business answers it.

Build an honest return report

One report, auditable by the business, every month.

If data is missing, report the highest valid stage and stop there. Don't estimate the missing inputs just to complete the table. Work down from the top until the answer is yes.

We review new clients' incoming agency reports. Of those showing a return figure, about 2 in 3 reports lead with ROAS. They take no margin, product cost or fees into account. That is why the template asks for the basis before the number. Our guide to the Google Ads metrics that matter covers the rest of the dashboard.

Return figures in advertising, case studies or sales pitches carry a legal standard as well. ACCC guidance on false or misleading claims says claims should be true, accurate and based on reasonable grounds. The guidance adds that it makes no difference whether a business meant to mislead.

One bridge to the other channel. An SEO comparison is only fair on the same cost, value, attribution and time definitions. It also has to allow for a return horizon measured in quarters rather than weeks. Our guide to how to measure SEO ROI sets up that side of the ledger.

What we recommend at Elev8d

Next steps

Written by Ajay K.Co-founder, Elev8d

Ajay is the co-founder of Elev8d and writes most of what you will read here. Psychology grad turned marketer, which is a fancy way of saying he cares about why people click, not just where things rank. Every guide is written in plain English for Australian businesses, tested on real campaigns before it earns a spot on this site.