Your dashboard shows 3.4x ROAS and your bank account shows something else. That gap is not a tracking bug, and it is almost never a creative problem.

It is arithmetic. Platform ROAS divides revenue by spend, which quietly assumes you keep every dollar of that revenue. Operators who scale profitably build the break-even ROAS formula first and set targets from their own margin, not from a benchmark someone posted online.

That is a system. Picking 3x because it sounds healthy is a habit.

What you'll be able to do

  • Calculate your true break-even ROAS and a target ROAS that actually pays you.
  • Turn a CRM close rate into a maximum cost per lead you refuse to exceed.
  • Read marginal CPA at the next $4,000 of spend and know whether to scale, fix the offer, or fix follow-up.

What you need

  • Last quarter's Shopify or Stripe payout reports (revenue, fees, refunds).
  • A spreadsheet and your actual shipping, pick/pack, and COGS invoices.
  • Admin access to your Meta, Google, and TikTok ad accounts.
  • Your CRM's lead-to-closed-won rate on a completed cohort.

No coding anywhere in this. If you can write a SUM formula, you can build this.

Why a "Healthy" 3x ROAS Is Often a Slow Bleed

Take an $85 average order value. Now subtract what leaves the business: $30 COGS, $10 outbound shipping and packaging, $2.77 in card processing (2.9% plus $0.30), $3 pick and pack, and $11.90 in expected refunds and returns.

Total variable cost: $57.67, or 67.8% of revenue. Contribution margin, often called CM2, is 32.2%.

Break-even ROAS is 1 ÷ 0.322 = 3.11x. At a reported 3.0x, that brand spends $28.33 in ads to earn $27.33 of margin. It loses about $1.00 per order, and it loses faster the more it scales.

The forgotten lines are always the same three: card processing, free-shipping subsidy, returns. Card processing is small per order and easy to skip. The free-shipping subsidy is invisible because it never appears as a line item.

Returns are the biggest. NRF's returns research has put US retail returns around 14% to 17% of sales, with apparel running well above that.

One discipline makes everything after this trustworthy: express every cost as a percentage of the platform's reported revenue. Then the margin you plug in lives in the same number space as the ROAS you see on screen.

Break-Even ROAS Is 1 ÷ Margin. Your Target Is Another Number.

Break-even ROAS equals 1 ÷ CM2%. A 25% margin is 4.0x, 30% is 3.33x, and 40% is 2.5x.

That is arithmetic, not a benchmark, and it means anyone saying "3x is healthy" is implicitly claiming a 33.3% contribution margin. Ask for their build.

The target ROAS calculation: 1 ÷ (CM2% minus desired profit %). On the $85 order with a 12% profit goal, that is 1 ÷ (0.322 minus 0.12) = 4.95x. Per order the same identity reads: max CPA = $27.33 minus $10.20 = $17.13.

Do not enter break-even as your floor. At break-even you work for free and finance the platform's float. Do not set it 70% above target either: delivery starves, learning breaks, and CPA often gets worse, not better.

Volume caveat: Google's guidance has been roughly 30 conversions per 30 days for Target CPA and 50 for Target ROAS, while Meta's old rule of thumb is about 50 optimization events per ad set per week. Below those thresholds your target is a suggestion, not a control.

Lead Gen: Turn Your Close Rate Into a Maximum CPL

For lead gen, the maximum cost per lead formula replaces ROAS entirely:

Max CPL = gross profit per customer × close rate × allowable ad-spend share

Drop the share and you get break-even CPL. The allowable share is the judgment call, usually 30% to 60% of gross profit. Pick a number, write it down, defend it.

Home services: a $4,200 average job at 38% gross margin is $1,596 of gross profit per customer. At a 22% lead-to-sold close rate that is $351 of expected gross profit per lead. At a 50% share, max CPL is $175.55.

Now the SaaS version that breaks people. A $9,000 contract at 78% margin is $7,020 of gross profit. At a 4% lead-to-closed-won rate that is $281 per lead, so a 35% share gives a ~$98 ceiling.

A $150 CPL that "closes at 4%" loses roughly $600 per customer, while the platform reports 3x ROAS on lead value and tells you to scale.

Measure close rate lead-to-closed-won on a cohort that has finished the cycle. A 30-day snapshot of an 80-day cycle understates close rate, over-tightens CPL, and starves a working channel. Most offline conversion mistakes start exactly here, and weak top of funnel targeting just makes the close rate look worse than it is.

Where to Enter the Numbers: Meta, Google, TikTok

Meta: Ad set, then Optimization and delivery, then the performance goal "Maximize value of conversions", then a ROAS goal bid strategy with your floor. It requires value and currency sent on the Purchase event through the Pixel or Conversions API. Label wording shifts, so verify in product.

Google: Campaign, then Settings, then Bidding, then "Maximize conversion value" with Target ROAS for ecommerce, or "Maximize conversions" with Target CPA for lead gen. You need conversion value tracking, not just counts. Value rules let you correct for margin by location, audience, or device.

TikTok: Ad group, then Bidding and optimization, then value-based bidding with a minimum ROAS, or cost cap for lead gen. It needs Pixel or Events API value data. After the 2024 to 2025 ownership transition, budget TikTok as a channel with concentration risk.

Sequence changes, do not stack them: creative, then budget, then bid target, roughly one to two weeks apart. Changing a target mid-learning-phase resets learning.

Set floors in the platform's own number space, because Meta's default 7-day click, 1-day view window will never match Shopify or GA4. If those three numbers keep diverging, server-side tracking is the usual fix.

The Only Number That Tells You to Add Money: Marginal CPA

Blended CPA is an average over weeks where your best days drag your worst days up. Marginal CPA is Δspend ÷ Δconversions, computed between budget tiers, and it is the only number that describes the next dollar.

$8,000 of spend, 320 orders, +$746 profit. Push to $12,000 and you get 448 orders and +$244, because the last $4,000 bought 128 orders at a marginal CPA of $31.25, already above the $27.33 break-even.

Push to $16,000 and profit drops to a $2,335 loss, with marginal CPA at $76.92. Blended CPA still reads $26.79 at $12k while the increment has lost money for weeks (see how these numbers fit a creative test plan before you blame the ads).

Rule: scale while marginal CPA sits at or below max CPA, stop hard at break-even CPA. Raise budget 20% to 30% every 3 to 7 days and recompute the increment each time.

Cross-check with MER and aMER, meaning total revenue (or new-customer revenue) divided by total ad spend. Platform ROAS double-counts the same order across Meta and Google and often sums to more than actual revenue ÷ actual spend. Before killing a new campaign, de-bias the read with a maturity multiplier: day-7 ROAS ÷ day-45 ROAS from your own historical cohorts.

Scale, Fix the Offer, or Fix Follow-Up (and the Maintenance Bill)

Three decisions, no fourth. Scale when marginal CPA is at or below max CPA.

Fix the offer when marginal CPA lands between max and break-even at every spend level: ads are not the problem, margin is. Pull the levers: bundle or raise AOV, renegotiate COGS, kill the free-shipping subsidy, cut refund drivers.

Fix follow-up when CPL is under the ceiling but close rate is running at half of plan: time to first touch, contact rate, sequence length, rep capacity.

Diagnostic shortcut: if the ad account hits CPL targets and the P&L does not, stop briefing the media buyer and audit the CRM. Speed-to-lead research going back to the Lead Response Management studies found 5-minute responses qualified far better than 30-minute ones. A 20% CPL cut that doubles time to first touch is a net loss.

The build takes an afternoon. Keeping it true takes half a day to a full day per month: per-SKU COGS through supplier changes, refund rates through seasonality, carrier and processing fee changes, CRM close rates through sales-team turnover. A stale spreadsheet produces confident, wrong targets, which is worse than no target.

This is where DIY stops being the cheap option. Measurement tooling runs from the low hundreds a month (Triple Whale, Hyros) to four figures (Northbeam, Rockerbox), and a fractional CFO or ecommerce analyst bills roughly $150 to $300 an hour.

Someone has to own the maintenance. Decide whether that is your afternoon or someone else's.

Where to go next

  1. Build the CM2 line by line on your last 90 days, in percent of reported revenue.
  2. Set your ROAS goal and Target ROAS to the target number, never the break-even number.
  3. Log spend and orders at two budget tiers for 14 days and compute marginal CPA.
  4. Audit time to first touch in the CRM before touching another bid target.

The math here has not changed since the first direct-response merchant, and that is the point. The formula is stable. The platforms are not.

You now have the exact math, so you can run it yourself this week. If you would rather hand the margin build, the bid targets, and the monthly maintenance to people who do it every day, our managed Growth and Scale retainers cover ad accounts across Meta, Google, and TikTok with the profit math wired in from day one.

Cover photo by Jakub Zerdzicki on Pexels.