Deciding Whether a Paid Campaign Is Worth Scaling

Mira runs paid search for a $89 online course. The ad account shows 4.1x ROAS on $3,200 of spend. Her business partner wants to triple the budget tomorrow. Mira's landing page converts at 2.4%, refunds are 8%, and fulfillment plus tax-adjacent fees leave about 62% of revenue as contribution before ads. The decision is not is 4.1x good. The decision is whether the next $6,400 will buy the same customers at the same economics.

The ROAS Calculator divides attributed revenue by ad spend. $13,120 ÷ $3,200 = 4.1x. That number ignores course refunds, payment fees, and the contractor who answers student email. If 62% of revenue remains after those non-ad variable costs, advertising has to produce at least 1 ÷ 0.62 ≈ 1.61x just to cover ad spend. Mira's 4.1x clears that bar easily — on last week's auctions.

A campaign at 1.9x with the same 62% contribution is still profitable on ads, barely. A campaign at 5x that attributes revenue the platform invented from view-through conversions may not be. ROAS is not profit. It is a revenue-to-spend ratio inside whoever's attribution window you accepted.

Mira's $3,200 bought 1,600 clicks. The CPC Calculator shows $2.00. Of 1,600 clicks, 1,420 became sessions and 34 purchases. The Conversion Rate Calculator on 34 ÷ 1,420 sessions is 2.4%. CTR on the ads was 3.1% — fine for her account history, and not a reason to scale by itself. CPM on the same campaign was $8.40, which only matters if she is comparing placements, not deciding a budget triple.

If she triples spend inside the same exact-match cluster, CPC often rises before conversion rate does. A $2.80 CPC at the same 2.4% conversion rate means each purchase costs $2.80 ÷ 0.024 ≈ $117 in ad click cost, versus $83 before. Revenue per purchase is $89. She would be buying $89 customers for $117 in clicks before refunds. Scaling on last week's $2.00 CPC is how a healthy ROAS becomes a hole.

Thirty-four purchasers from $3,200 of media plus $400 of landing-page tools and a contractor slice is $3,600 of acquisition cost. The Customer Acquisition Cost Calculator gives $3,600 ÷ 34 ≈ $105.88. After an 8% refund rate she has about 31 kept customers, so fully kept CAC is closer to $116. Contribution after non-ad costs on a kept $89 sale is about $55. She does not recover CAC on the first purchase.

That does not automatically kill the campaign. It kills a scale decision that assumes one purchase pays for itself. If Mira has evidence that 40% of buyers take a $49 add-on within 60 days, contribution changes. If she does not have that evidence, tripling spend is buying more first purchases that do not cover themselves.

Raise spend 20–30% for one full auction cycle, not 200%. Hold creative and landing page constant so you can see CPC and conversion rate move. Recalculate ROAS on the incremental spend only — last week's leftover conversions credited to this week's budget will flatter the result. Recalculate CAC on new customers in that window, not on the blended month.

Do not scale a campaign whose attributed revenue is mostly branded search you would have received anyway. Do not scale in the week a discount code is inflating conversion rate. Do not scale when refunds have not had time to land — course refunds in particular arrive after the ROAS screenshot.

The calculators will not detect branded-query contamination or a coupon. They will show, quickly, that a $2.80 CPC at 2.4% conversion on an $89 product cannot be rescued by a prettier ROAS tile. Scale the campaign that still clears contribution on the next dollar. Pause the one that only looked cheap at last week's price of clicks.

Write the scale rule down before you raise the budget: incremental ROAS must stay above 1.61x on click-attributed sales, CPC may rise to $2.30 but not $2.80, and conversion rate may not fall below 2.0% on the same landing page. If any one of those three breaks, the extra spend goes back. That is a duller meeting than tripling spend on a green tile. It is also how you avoid buying a week of $89 customers at $117 a click.

CTR and CPM belong in the diagnosis, not the go/no-go. A falling CTR with a stable conversion rate usually means the ad is tired. A stable CTR with a falling conversion rate means the page or the audience quality changed. Mira should not triple spend to fix a tired ad, and she should not kill a converting page because CPM rose on a broader placement.

A 30% step on $3,200 is $960. Hold the ad and the page. In that week she spends $960, gets 420 clicks at a $2.29 CPC, 380 sessions, and 8 purchases. Incremental ROAS is (8 × $89) ÷ $960 = 0.74x — below the 1.61x contribution break-even. Conversion rate is 8 ÷ 380 = 2.1%. The Conversion Rate Calculator and CPC Calculator already say the extra dollars are buying a thinner query slice. Tripling the budget now would scale a failing increment, not last week's 4.1x.

Split the same $960 by query type and the picture changes again. $310 of branded search produced 5 of the 8 purchases and a 4.0x incremental ROAS. $650 of non-brand produced 3 purchases and 0.41x. Scaling the campaign as one object would pour money into the half that does not clear contribution. The ROAS Calculator on the blended week still looks almost acceptable if leftover view-through from last week is included. Incremental, click-based, non-brand only is the cut that should get the next dollar.

The edge case is a weekend CPC of $1.55 and a weekday CPC of $2.45 at the same conversion rate. Average CPC of $2.00 hid that weekdays were already near the $2.30 tripwire. Raising budget evenly across the week pushes more spend into the expensive days. Cap weekday bids first, or shift the extra $960 to the hours that still buy an $89 sale for less than ~$55 of contribution. CPM rising from $8.40 to $11 on a broader placement is a placement diagnosis. It is not, by itself, a reason to raise or kill the budget.