CPA, CPL, CPM, ROAS: the acquisition metrics that decide budgets

CPA, CPL, CPM, and ROAS all answer the same question, where the money actually went, but each one answers it from a different point in the funnel. Mix them up and budgets end up optimized against the wrong signal - low CPM instead of low CPA, cheap leads instead of paying customers.

What CPA (cost per acquisition) actually measures

Spend $5,000 on a campaign, land 50 paying customers, and CPA comes out to $100 - the whole of ad spend divided by the count of people who actually paid.

It only works when tracking is clean - server-to-server postbacks, deduplicated conversions, a clear attribution window. Calculate it from click IDs sitting in a spreadsheet instead, and it's a guess dressed up as a number.

CPA varies wildly by business: a $50/month SaaS subscription might target $100-$150, a $2,000 high-ticket course runs $400-$600, and a depositing player in iGaming can cost $80-$250 depending on geo and source. None of those numbers mean anything except relative to customer lifetime value.

Two things hide inside CPA: time lag and quality. A user converting on day 30 looks expensive if you're only measuring a 7-day window, and a cheap CPA sourced from incentivized traffic tends to churn faster than anything organic. Check retention before scaling on the number alone.

CPA vs. ROAS: short-term vs. long-term

ROAS, revenue divided by spend, tells a related but different story - a $100 CPA generating $300 in revenue is a 3x ROAS, though one-time upsells or refunds can inflate that number. CPA strips the question down to one thing: whether the customer turned out to be worth more than they cost.

CPA fits fixed-payout actions - affiliate offers, app installs - while ROAS fits variable revenue, e-commerce, subscriptions. Both are only as good as the data feeding them.

CPL (cost per lead), and why it gets misused

A lead can be almost any action short of a sale - an email signup, a form fill, a demo request, a trial start - and CPL simply divides spend by however many of those you bought, though not all of them carry the same weight.

A $5 CPL from a pop-up campaign in Tier-3 traffic and a $50 CPL from a targeted LinkedIn form are not remotely comparable - the first might convert at 1%, the second at 10%. Effective CPL folds conversion rate into the number directly: at a 10% lead-to-customer rate, effective CPA is CPL / 0.1.

B2B SaaS demo leads typically run $40-$120 in Tier-1, e-commerce email captures $1-$5, real estate buyer leads $20-$60, dating app signups $0.50-$2.50 - and none of those ranges mean anything without knowing what happens after the lead converts.

The most common mistake is optimizing for the lowest CPL while never tracking what happens downstream - a low CPL is pure waste if lead quality is garbage. Set a quality threshold instead: time on site, pages visited, a short qualification question.

• Calculate effective CPA from CPL before trusting the number.
• Segment leads by source and quality score.
• Hold off scaling a low CPL until downstream conversion actually confirms it.

CPM (cost per mille): you're paying for attention

CPM buys eyeballs: cost per 1,000 impressions, and the default pricing model for brand awareness and top-of-funnel campaigns.

Typical CPMs run $5-$12 on TikTok in Tier-1, $8-$25 on Meta, $3-$10 on Google Display, $10-$30 on YouTube, $2-$8 on programmatic. A high CPM doesn't automatically mean bad value - viewability and audience targeting decide that.

Used alone for performance campaigns, CPM turns dangerous fast: a $10 CPM at 0.1% CTR works out to a $10 CPC, and at a 2% conversion rate, that's a $500 CPA. A $20 CPM at 0.5% CTR and 5% conversion, by contrast, lands at $80 CPA - the higher CPM wins outright.

CPM is genuinely useful for comparing inventory costs across platforms, as long as the downstream metrics get calculated before any decision gets made. It moves with ad quality, placement, and competition.

CPM vs. CPC: when to use each

CPM suits awareness or retargeting, anywhere frequency matters; CPC suits direct response. A $10 CPM at 1% CTR works out to a $1 CPC, and if CPC lands the same across two platforms while CPM differs, the platform with the lower CPM is delivering more reach per dollar.

CPM fits video views, brand campaigns, and reach objectives; CPC fits search and shopping, where intent is already explicit.

Making sense of ROAS (return on ad spend)

A 3x ROAS means three dollars back for every dollar spent - revenue from ads divided by ad spend. Both sides of that fraction are trickier than they look on paper.

Revenue itself is ambiguous because gross revenue, net revenue after returns, or contribution margin all give different answers. Most platforms, Meta and Google included, report gross revenue at click time, cancellations and all, so a 4x ROAS on Meta can settle at 2x once refunds are subtracted.

Healthy ranges run 2x-5x for e-commerce and 3x-7x for SaaS, where revenue arrives delayed. Lead gen is the hardest to pin down, since it depends on assigning a real value per lead - a $50 lead with a 10% close rate and a $5,000 deal size works out to 10x ROAS on paper, but only once the deal actually closes.

Optimizing ROAS without a defined attribution window backfires either way - a 7-day click window misses whatever converts on day 8, a 28-day window over-credits the last click. The window that works is whichever one matches the actual customer decision cycle.

• Use net revenue, after returns and chargebacks.
• Set the attribution window to the purchase itself: 7-day click for impulse buys, 28-day for considered ones.
• Never compare ROAS across channels unless the attribution is uniform across all of them.

How these metrics interact: a real budget example

Take a $10,000 campaign budget and buy by CPM first. At $10 CPM, that's 1M impressions; at 0.2% CTR, 2,000 clicks; at 3% conversion, 60 customers - CPA = $166.67.

Buy by CPC instead, at $1.50, and the same $10,000 buys 6,666 clicks - at 3% conversion, 200 customers, CPA = $50. Same budget, far lower CPA: buying by the click beat buying by impressions here, because every dollar went toward a click that mattered instead of an impression that might not.

A ROAS below target means either CPA needs to come down or revenue per customer needs to go up.

I've seen campaigns where a $50 CPM at 1% CTR and 8% conversion produced a $62.50 CPA, beating a $20 CPM at 0.1% CTR and 2% conversion that landed at $1,000 CPA. The higher CPM won purely because the traffic underneath it was better quality.

Where these four metrics get misread

CPA looked at in isolation is the easiest way to fool yourself - a $30 CPA on a $29 product is a straightforward loss, and pairing it against AOV or LTV is the only way to know whether a given CPA is actually fine.

Attribution models get compared as if they measure the same thing, and they don't - last-click, first-click, and linear all produce different answers for the identical campaign, so the only fix is picking one model and staying with it.

Time gets ignored too often: a campaign can show a 4x ROAS in week one and 1.5x by month three, once retargeting pools exhaust themselves, which means metrics need time to stabilize before they're worth judging at all.

Segmentation is where blended numbers do the most damage - a $100 blended CPA can be hiding a $50 CPA from search sitting right next to a $200 CPA from display, and without that split, optimization ends up targeting the wrong channel entirely.

Which metric to track, by business goal

FAQ

What is the difference between CPA and CPL?

CPA is cost per acquired customer, a sale or an install; CPL is cost per lead, any pre-sale action like a signup or form fill. CPL sits earlier in the funnel; CPA is the bottom line the funnel is built to reach.

Is a lower CPM always better?

No, a low CPM can just as easily mean low-quality placements or bad targeting. A $3 CPM producing zero conversions is worse than a $20 CPM that actually drives sales, which is why downstream metrics need tracking regardless of how cheap the CPM looks.

How do I calculate ROAS for a subscription product?

Use average customer lifetime value over a defined period, 6-12 months is typical: ROAS = LTV / CPA. For monthly subscriptions, annualize the revenue but account for churn along the way.

What is a good CPA for e-commerce?

It depends entirely on AOV, and a common benchmark is CPA ≤ 30% of AOV, which puts a $50 AOV under a $15 CPA and a $150 AOV under $45. Adjust them for your own margins.

Can I use CPM for performance campaigns?

Yes, as long as the creative drives a high CTR and the conversion rate is actually known. CPM campaigns can scale efficiently, but only with tight tracking to keep waste in check.

Ioann Putevoy

Head of Traffic & growth lead. I build products and take them to market - see the portfolio.