What Is a Good CPA in 2026? Benchmarks by Platform and Event

How CPA is calculated (and the two equivalent funnel forms)

The definitional formula is:

CPA = Total Ad Spend ÷ Conversions

$1,000 spent for 25 purchases is a $40 CPA. The conversions in the denominator must be the event the spend was optimized for. Mixing Lead into a purchase campaign's CPA is how a $28 "win" appears next to a $40 problem.

Two identities are the same number from different inputs — they are what the CPA calculator funnel mode runs:

• CPA = CPC ÷ conversion rate. A $0.50 CPC at 5% CVR is $10.
• CPA = CPM ÷ (CTR × CVR × 1,000). A $10 CPM at 2% CTR and 5% CVR is the same $10.

That second form is why a cheap CPC can still lose. A $0.90 TikTok CPC at 1% CVR is a $90 CPA. A $5.42 Google CPC at 8% CVR is a $67.75 CPA^[3]. Compare platforms at CPA, not at CPC.

Break-even CPA is the only universal "good"

Industry medians answer "am I in the same zip code as peers?" They do not answer "can I afford this?"

Break-even CPA = AOV × gross margin.

| AOV | Gross margin | Break-even CPA | Implied ROAS at break-even | | --- | ---: | ---: | ---: | | $40 | 60% | $24 | 1.7× | | $80 | 40% | $32 | 2.5× | | $80 | 25% | $20 | 4.0× | | $150 | 50% | $75 | 2.0× |

A $30 Meta purchase CPA is a gift on the $150 / 50% row and a slow leak on the $80 / 25% row. That is why the calculator overlays break-even and optional LTV / ROAS ceilings on top of realized CPA instead of declaring a global grade.

Two other ceilings are useful when AOV is the wrong unit:

• Target CAC from LTV = LTV ÷ desired LTV:CAC (a 3:1 target on $300 LTV is a $100 fully loaded CAC ceiling, not a media-only CPA). Pair with the LTV calculator.
• Target CPA from ROAS = AOV ÷ target ROAS (a 4× target on $80 AOV is a $20 media-CPA ceiling). Pair with the ROAS calculator.

2026 bands, kept on their own event

Honest, source-attributed figures. Low/high are the published industry-span extremes in that report — not a smoothed “typical” band. Triple Whale rows are ecommerce purchase CPA.

| Segment | Event | Typical range | Published figure | Statistic | Source | | --- | --- | --- | ---: | --- | --- | | Google Ads search | Cost per lead | $26.84 – $131.63 | $66.69 | Average | WordStream / LocaliQ Search Advertising Benchmarks 2026 | | Meta lead campaigns | Cost per lead | $3.16 – $76.71 | $27.66 | Average | WordStream Facebook Ads Benchmarks 2025 | | Meta ecommerce | Purchase | $29.99 – $49.48 | $38.19 | Median | Triple Whale 2025 cohort (updated Feb 2026) | | Google Ads ecommerce | Purchase | $25.15 – $35.37 | $23.74 | Median | Triple Whale 2025 cohort (updated Feb 2026) |

Two cautions before you screenshot the table into a Slack thread:

1. Ecommerce-weighted purchase CPA reads lower than all-industry search CPL because a purchase and a lead are different products, and because Triple Whale's cohort is ecommerce-heavy. The $23.74 Google purchase median and the $66.69 Google lead average can both be true in the same week.
2. TikTok "CPA" is the least trustworthy single number in this category. Recaps that treat TikTok as ecommerce purchase land near the high teens; recaps that treat TikTok as an upper-funnel action quote $260–$750+^[4]. Until a publisher states the event, skip the TikTok row.

For CPM / CPC / CTR — the inputs that produce CPA — use the Paid Social Benchmarks analyzer rather than this table. That tool stays at platform × objective grain so it does not duplicate the glossary's per-industry cost tables.

Why CPA rises (and what to decompose first)

When CPA inflates, the useful identity is CPA = CPC × (1 / CVR). Three usual causes, in the order we see them:

1. Creative fatigue. CTR falls, CPMs rise, conversions per dollar fall. Meta's official Creative fatigue status still fires only after cost per result doubles against your own history — by then the CPA chart has already turned. See the Andromeda playbook for why a refresh has to be a new concept, not a re-crop.
2. Landing-page or offer decay. CPC holds; CVR slips. Message match between ad and page is the cheapest CPA lever that is not a bid change.
3. Auction competition. Seasonal spend or new entrants raise clearing prices. This is the last diagnosis, not the first.

Do not cut spend first. Decompose, then decide. The calculator's funnel mode and CVR-sensitivity chart exist for that exact conversation: holding CPC constant, doubling CVR halves CPA.

How to use the bands without lying to yourself

1. Name the event on every CPA you report (Purchase, Lead, Trial start).
2. Compute break-even before you open a benchmark table.
3. Compare inside one platform and one event. A Meta lead CPL beating a Google purchase CPA is not a win.
4. Read CPA next to MER and incrementality. A $28 platform CPA can coexist with a blended CAC of $95 once agency, creative, and tools are loaded — the comparison page walks that failure mode.
5. Re-run the number in the CPA calculator and export the CSV into the weekly doc so the inputs travel with the verdict.

The glossary still owns "what CPA means." The comparison still owns "CPA vs CAC." This post and the calculator own the 2026 bands, the funnel identities, and the break-even test that turns a published median into a decision.