Your agency’s quarterly review opens on a slide that celebrates how TACoS improved again. Heads nod. The number is real, the trend line is genuinely down, and nobody in the room asks the only question that matters: who chose that metric?
Almost always, the agency did. And the metric you grade your agency on isn’t really a reporting preference — it’s an instruction. ROAS vs cac? It tells the agency, and the algorithms working underneath them, exactly what kind of customer to go out and buy with your money. So if you choose it carelessly, you can end up with an agency that hits its targets every quarter while your business gets worse underneath them.
ROAS, CAC, and TACoS each have a legitimate job, and so do their cousins CPA, ACoS, and MER. Things go sideways when the team being measured gets to decide on the measurement definition. So here’s how I’d sort them, first as a founder holding an agency accountable, then through the lens I’d use as an investor reading a target company’s paid program.
The four metrics
| Metric | What it measures | Formula | When it’s the right yardstick |
|---|---|---|---|
| CAC (Customer Acquisition Cost) | What it costs you to win one new customer | Total marketing spend ÷ new customers acquired | High-LTV businesses that need volume: repeat-purchase brands, anything where the customer’s value compounds after the first order |
| CPA (Cost per Acquisition) | What it costs you to generate one order or action | Total ad spend ÷ orders (define the denominator carefully) | Subscription businesses, where every first order starts a subscription — CPA and CAC converge. Elsewhere, only with “new customer” written into the definition |
| ROAS (Return on Ad Spend) | Revenue generated per ad dollar, campaign by campaign | Ad-attributed revenue ÷ ad spend | Businesses where the first order is most of the value or AOV needs improving: big considered purchases, gift-heavy categories, low cost products where cart size affects overall margin |
| TACoS (Total Advertising Cost of Sales) | Ad spend as a share of your total revenue, organic included | Total ad spend ÷ total revenue (paid + organic) | Judging a partner who genuinely owns the whole channel — paid and organic — like a full-service Amazon or Walmart marketplace agency |
A strategy and goals targeting each of these metrics is defensible. The problem is matching the wrong metric to the wrong business, and the worst mismatch is the one agencies propose most enthusiastically when your business needs a different strategy.
TACoS: the metric that lets organic pay for paid
Watch what TACoS does mechanically. Because the denominator is all your revenue, every organic dollar your brand earns makes the paid program look more efficient. Your email list grows, TACoS improves. A creator mentions you, TACoS improves. Your brand gets stronger for reasons that have nothing to do with ads, and the ads take the bow.
There’s a well-documented illustration of how much organic can hide. When eBay ran large-scale experiments turning off its branded search ads, traffic barely moved; buyers simply arrived through the free organic listings instead, and the study put the true return on that spend at roughly negative 63% (Blake, Nosko & Tadelis, Econometrica). Under a blended metric, all of that substitution had been invisible. The dashboard said the ads worked. The experiment said the organic demand was carrying them.
So when is TACoS legitimate? When the same partner is responsible for both sides of the ledger. A full-service marketplace agency running your Amazon or Walmart business owns your listings, content, SEO, and reviews along with the ad console; paid and organic are one interlocking system there, and a blended metric is a fair grade. Same logic if you’ve handed an agency a microsite where they control the whole funnel.
Understand what you’re really doing in that arrangement, though: grading an agency on TACoS is handing them a mini P&L. If you’re going to do that, run it like one, which means they own and control the budget for the organic activities too. When your Amazon agency decides the business would benefit from Vine reviews or influencer posts, that activity gets charged to the Amazon channel budget. Both are great additions to an Amazon media plan. Neither is a zero-cost line item just because no ad console sends an invoice.
The example founders miss most often is their own email list. If the agency wants a send pointing your customers to Amazon instead of your own site, that email has a price. Your average send generates a known amount of revenue at your own-site margins, and redirecting it means giving up that revenue to make the marketplace number look better. That opportunity cost belongs in the TACoS math. A blended metric is only honest when every source of revenue it benefits from shows up somewhere in the math.
But a paid-only agency driving traffic to your primary website should never be measured on TACoS. They don’t run your email, your SEO, your PR, or your social, yet a blended metric pays them for all of it. You’ve given them credit for rent they don’t pay. Hold paid agencies to paid metrics: ROAS, CAC, or CPA, on the traffic they actually touch.
ROAS, CAC, and CPA are targeting instructions in disguise
Founders often treat the choice among these like a reporting preference, but it’s really an operating decision. Each metric sends the algorithm hunting for different humans.
Optimize to ROAS and you’ve told the machine that revenue per dollar is all that matters. A $400-basket customer who cost $100 to acquire and a $40-basket customer who cost $10 look identical: both are a 4x. The algorithm will happily chase the expensive big spender, because acquisition cost per person doesn’t enter the equation. That’s exactly what you want when most of a customer’s lifetime value lands in the first order — low repeat rates, considered one-time purchases, gift-driven categories. Let the machine go find the whales.
Optimize to CAC and you’ve issued the opposite instruction: hold the line on what we pay per new customer. The algorithm stops splurging on glamorous baskets and starts forcing volume, because cheap-to-acquire customers are the only way to hit the number. That’s the right pressure when lifetime value runs well past the first order. A strong-repeat brand mostly needs people in the door at a controlled price; your retention engine takes it from there. Grade that business on first-order ROAS and you’ll under-spend on exactly the customers who would have compounded.
CPA is the cost-side metric agencies quote most, and the one with the loosest definition, so pin it down. CPA counts cost per order; CAC counts cost per new customer. In a subscription business the two converge, which is why CPA often makes sense there: every first order starts a subscription, so cost per order is cost per new subscriber, and optimizing to CPA produces exactly the volume-at-a-controlled-price behavior a high-LTV model needs. Outside subscription, the gap between those two definitions is where an agency can hide. Graded on cost per order, they can hit the target by pointing spend at customers you already own — retargeting your list produces wonderfully cheap orders while acquiring nobody new. You paid to harvest revenue you already had. If you grade on a cost metric, write the denominator into the contract: new customers, with returning customers reported separately.
The decision rule fits in one line: the closer your customer’s lifetime value sits to their first order, the more ROAS makes sense; the further LTV runs past the first order, the more CAC — or, in subscription, CPA — should govern. If you don’t know your LTV-to-first-order ratio, that’s the homework before any agency conversation, because without it you’re letting someone else guess, and the guess will favor whichever metric flatters them.
One caution in both directions: none of these metrics knows your margin. A 4x ROAS on a 20-point-margin product loses money; a “cheap” CPA that attracts discount-hunting one-timers is expensive at any price. Whichever you choose, someone on your side of the table has to keep contribution margin in view.
If you live on Amazon: ACoS
ACoS is ROAS flipped upside down — ad spend ÷ ad revenue, so lower is better. Everything above about ROAS applies unchanged: it grades the campaign, ignores what each customer cost to acquire, and chases big baskets when optimized to. TACoS is its blended cousin, and the same ownership test decides which one is fair.
MER: the grade for your marketing department
MER (total revenue ÷ total marketing spend) looks like TACoS’s twin, and mechanically it nearly is. The difference is who you point it at.
Hand MER to a single agency and you’ve recreated the TACoS problem: credit for revenue they don’t influence. Your CMO or VP of Marketing is a different story. They own the whole blend — paid, email, SEO, social, and the agencies themselves — so blended efficiency is exactly what they should answer for. MER is how you hold your marketing department accountable; ROAS, CAC, and CPA are how your marketing department holds its vendors accountable. I covered how to set MER as a guardrail in my piece on ecommerce metrics. Keep the blended metrics inside the building, and grade each outside partner on what they actually touch.
The investor read: the metric regime is hidden value
If you evaluate consumer companies for a living, the diligence habit this suggests is cheap and revealing: before you trust any paid-efficiency number in the deck, ask three questions. Which metric is the agency contractually graded on? Who selected it? And does it match the LTV shape of the business?
Misalignment shows up constantly, and it cuts both ways. A subscription brand proudly reporting 5x first-order ROAS is probably under-acquiring; the founder is leaving compounding customers unbought to protect a ratio that was never the right governor. A one-and-done category run on aggressive CAC targets is probably stuffing the funnel with small-basket customers who will never pay back. And any pitch where paid performance is quoted in TACoS while the agency only runs paid deserves a second look, because some of that “efficiency” is organic demand getting credited to paid.
For investors, that’s a useful discovery. A metric mismatch is one of the cheapest value-creation levers you’ll find in diligence. Re-aiming the spend (same budget, same channels, new governing metric matched to the LTV math) can change a growth trajectory without a dollar of new investment. In diligence, that kind of sloppiness can be useful: it tells you where underexploited value may be hiding.
The practical takeaway
Pick the metric before you pick the agency, and pick it off your LTV math rather than the agency’s pitch deck. Paid-only partners get paid-only metrics, with new customers in the denominator. Reserve TACoS for partners who own organic outcomes too, budget and all, opportunity costs included. MER stays inside the building, as the grade for the leader who owns the whole blend. And whoever you hire, keep one rule intact: the party being graded doesn’t get to choose the grading scale.
If you’re not sure which metric should govern your spend:
That’s a question your own numbers can answer, and it’s a good place for us to start together. We could look at your LTV-to-first-order math and how you are currently grading your agencies. Then we’d see whether the two line up. You’d come away with a short list of metrics worth running the business on, plus guardrails for everything else.. It’s the kind of fractional CMO work I like to begin with: low-stakes, quick, and useful whether or not we ever take it further. I promise it’ll be a working session, not a lecture on formulas. If your agency picked its own report card, let’s talk →.
