Why a good ROAS can still mean bad growth: a check for Inner West businesses

What ROAS is, and what it was never meant to say

ROAS means return on ad spend: the revenue an ad platform attributes to a campaign divided by what the campaign cost. Spend $1,000, get credited with $4,000 of revenue, and the campaign has a ROAS of 4.0. It is simple, it appears in every dashboard, and it is useful for comparing campaigns inside one platform. It was never designed to answer the question owners actually ask, which is whether the business is better off.

Three things sit between ROAS and that answer. ROAS counts revenue, not profit. It counts credit given by the platform, not sales the ad caused. And it counts a window of time, not the full cash picture. Each is covered below. First, a way to think about the gap.

ROAS tells you what the platform claims. Profit tells you what you kept.

Revenue-distance: how far a number sits from money in the bank

Every metric in a dashboard sits some distance from cash in the bank. We call that revenue-distance. A click is far away. An order in your system is closer. Gross profit received and banked, from a customer who would not otherwise have bought, is the closest.

ROAS lives in the second row. Most of the damage in small-business advertising comes from making decisions in row two as though they had been made in row six.

Seven things a good ROAS hides

1. Revenue the ad did not cause

Platforms report revenue from people who clicked an ad before buying. They cannot tell whether the person would have bought anyway. Brand search is the clearest case. If someone types your business name into Google and clicks your ad, the ad gets the credit, and the sale was probably coming. A large field experiment run by economists on eBay, published as an NBER working paper, found that brand-keyword ads had no measurable short-term benefit, and that frequent users who bought regardless accounted for most of the advertising expense, so average returns were negative. It is one company's marketplace, not a Marrickville cafe, so treat it as a warning about where to look and not as a rule.

2. Cannibalised organic and word of mouth

Ads can sit on top of customers who would have arrived through your Business Profile, a friend's recommendation or your existing email list. Google provides brand exclusions for Performance Max and Search campaigns, which keep campaigns from serving for branded queries you want to avoid. If you run Performance Max without them, brand searches can flow into a campaign you believe is finding new people. The Search versus Performance Max guide goes into this.

3. Margin and GST

ROAS is revenue over spend. Profit is margin on that revenue over spend. A ROAS of 3.0 on a product with a 25% margin loses money. The next section gives the formula.

4. New versus repeat customers

A $120 first order from a new customer and a $120 order from someone who has bought four times look identical in ROAS. They are not. The first may become a repeat customer, which the dashboard does not value. The second would likely have happened without the ad.

5. Cash-flow timing

You pay for ads now. Margin arrives over weeks or months, and repeat purchases over a year. A campaign can be profitable over twelve months and still drain the account this quarter. That is not a flaw in the campaign, but you need to know it before you scale.

6. Attribution windows and view-through credit

Platforms decide how long after an ad a purchase can still be credited to it. In Google Ads, Google's help page says the click-through conversion window defaults to 30 days and can be set to between 1 and 30 days, or 60 or 90 days, for Search and Display campaigns, and that the view-through window defaults to 1 day and can be set between 1 and 30 days. View-through means a conversion credited after someone saw an ad and did not click it. Meta's default, per a third-party guide updated in March 2025 (Meta's own help pages were not available to us), is 7-day click and 1-day view. Google also states that data-driven attribution is the default model for most conversion actions, so credit is distributed by a model you do not control.

Put two platforms together and the problem is visible. A customer sees a Meta ad, clicks a Google ad three days later and buys. Both platforms claim the sale. The sum of platform-reported revenue can exceed what the till recorded.

7. Small-sample noise

On small budgets, conversion counts are small, and small counts move about by chance. If a campaign records 40 orders, the typical swing from luck alone is about the square root of 40, roughly 6, or 16% either way (simple arithmetic, illustrative). With an unchanged average order, a campaign showing a ROAS of 2.0 could as easily have shown 1.7 or 2.3 from randomness. Do not change a budget because of a week's data, and do not call a test a winner on a handful of conversions.

The related trap is optimising toward the cheapest customer. If the bidding is told to minimise cost per conversion, it finds the people easiest to convert: past buyers, brand searchers, bargain hunters. Cost per conversion falls, and the customer base gets no bigger.

Break-even ROAS: the number every campaign needs

Break-even ROAS is the ROAS at which an ad pays for itself and no more. The formula is break-even ROAS = 1 / margin, where margin is the share of each dollar of revenue left after the direct costs of making and delivering the sale (product cost, labour on the job, payment fees, delivery), before ad spend.

• Margin 50%: break-even ROAS is 1 / 0.50 = 2.0.
• Margin 40%: break-even is 1 / 0.40 = 2.5.
• Margin 25%: break-even is 1 / 0.25 = 4.0.

Now the GST adjustment, which trips up many shopfront owners. Registered businesses include GST in the price they charge, and GST is 10% of most goods and services sold in Australia (ATO). The 10% is not yours: it goes to the ATO. If the revenue your platform reports includes GST, your margin must be calculated against the GST-inclusive figure, which means dividing the ex-GST margin by 1.1. A 40% margin on ex-GST sales is a 36.4% margin on GST-inclusive revenue, so break-even ROAS is 1.1 / 0.40 = 2.75, not 2.5. If your tracking sends revenue ex-GST, the plain formula applies. Check which version your account uses. If you are unsure of your own tax treatment, ask your accountant.

A worked example: an Inner West homewares store (illustrative)

Take an illustrative independent homewares and furniture store on a main strip in the Inner West. It has a small online shop and a showroom. It spends $3,000 in a month on two campaigns, a branded Search campaign and a prospecting campaign aimed at people who have never heard of the store. The figures are invented to show the sums.

Assumptions: gross margin of 40% on ex-GST sales; platform revenue includes GST, so break-even ROAS is 1.1 / 0.40 = 2.75.

On paper, the business is roughly break-even, with a star campaign carrying a weak one. Now apply two questions the dashboard does not ask.

How much of the branded revenue was incremental? Suppose, as an illustrative assumption, that only 15% of those branded buyers would not have found the store without the ad. Incremental revenue is 15% of $3,600 = $540. Incremental gross profit is $540 / 1.1 x 40% = $196. Against $600 spend, that is -$404. Incremental ROAS is $540 / $600 = 0.9, well below break-even, even though the dashboard says 6.0.

What are the prospecting customers worth over a year? Suppose the $4,800 came from 40 new customers at an average order of $120, and that 16 of them (40%) buy again within a year at $120. Repeat revenue is 16 x $120 = $1,920. Repeat gross profit is $1,920 / 1.1 x 40% = $698. First-order gross profit was $1,745, so twelve-month gross profit is $2,443 against $2,400 spent: about +$44. That is roughly break-even over a year, with $655 of cash tied up in the meantime.

The decision this changes: the 6.0 campaign is the loss-maker and the 2.0 campaign is the one building the customer base. A dashboard sorted by ROAS would cut the second and scale the first. Neither conclusion rests on certainty: the 15% and the 40% are guesses that you would replace with your own data. But the direction survives a wide range of guesses.

The four-question growth check

Instead of asking whether ROAS is good, run this check on any campaign before scaling, cutting or celebrating it.

1. Who was it? Was the customer new to you, or someone already in your orbit: a past buyer, a regular, a person who searched your name? Sort your own orders by first-time and returning customers, and compare them with what the platform reports.
2. Would it have happened anyway? What happens to enquiries and sales if the campaign is paused for two weeks, or if brand is excluded from the campaign claiming it? A pause test is imperfect, and seasonality and events muddy it, but it is the nearest thing a small business has to a control group.
3. Does it clear break-even after margin and GST? Calculate break-even ROAS properly, using the same revenue basis (inclusive or exclusive of GST) as the reported figure. Compare incremental ROAS, not reported ROAS.
4. When does the cash arrive? Work out the payback period. How many weeks until gross profit from this month's customers covers this month's spend? Check you can carry that gap.

Four honest answers will tell you more than any dashboard. The numbers do not need to be precise. They need to be written down, so that you can see which of your guesses is carrying the conclusion.

What we would do first this week

• Work out your margin per sale, ex-GST, after direct costs. Ask your bookkeeper if you are unsure which costs count.
• Calculate break-even ROAS using the formula above, adjusted for GST if your reported revenue includes it.
• Look at your search terms report (it shows how ads performed when triggered by actual searches) and find the share of spend and revenue that comes from your own business name.
• Check whether Performance Max, if you run it, has brand exclusions applied. See the Search versus Performance Max guide.
• Split reported conversions by new and returning customers using your own order or booking records.
• Make sure the numbers are real before arguing about them: the tracking and reconcile guide shows how to check them against your phone log and bookings.
• Decide the next test and what result would change your mind, before you run it.

An AI agent can do the repetitive part of this: pulling a search terms export of several thousand rows, tagging branded and non-branded queries, and matching conversions to customer records. A person decides which assumptions are fair and what to do about the result.

When ROAS is the right metric

ROAS is not a bad metric. It is a badly used one. It works well when these conditions hold:

• The revenue is real and measured. An online shop where every order passes through your tracking, with the value attached, is the ideal case.
• Margins are stable and known. Break-even ROAS then becomes a reliable target, and value-based bidding can use it.
• The campaign targets new people. Prospecting campaigns measured on ROAS are far less exposed to the brand problem than branded campaigns.
• The numbers are large enough. There is enough volume that chance does not dominate.
• You compare like with like. ROAS is a good way to rank campaigns inside one account that share margins and customer type.

It fits poorly for dentists, tradies and architects, where the phone call is the conversion and the value of a lead is a guess. For those businesses, cost per booked customer, set against what a customer is worth over time, is a sharper tool, and the budgeting guide shows how to build it.

Limits of this advice

The incrementality question is the hardest one here, and we do not have a clean answer for small budgets. A pause test is noisy, a geo-split test needs more volume than most owner-led businesses have, and the eBay study is a large marketplace and not a local shop. Brand search can also be defensive: if a competitor bids on your name, pausing your own brand campaign could hand them the click. In that case a modest branded spend may be worth it, even with a poor incremental ROAS.

The margin, repeat-rate and incremental-share figures in our example are illustrations. Yours will differ, and so will the conclusions. We have not seen your accounts, and platform interfaces and defaults change, so check current help pages for attribution windows before relying on them. None of this is financial or tax advice. Margin, GST and cash-flow questions belong with your accountant.