Executive summary: The "best" bid model depends on where friction lives in your funnel. CPC shifts conversion risk to the advertiser; CPA shifts it to the network; eCPM prices eyeballs. In 2026, advertisers who lock into one model leave 20–40% of yield on the table. ReTarget.gg's auction normalizes all three into a single ranking while letting each advertiser pay on their chosen terms.
The CPC vs CPA vs eCPM debate has run since the 2000s. The answer has not changed: it depends on where friction lives in your funnel.
What has changed is that running all three simultaneously is now standard infrastructure — and advertisers who pick a single model for every geo and overlay type are systematically underbidding.
This post is the cheat-sheet we wish existed when we sized our own auctions: definitions, risk allocation, 2026 pricing realism for iGaming, a decision matrix, and how ReTarget.gg runs all three without fragmenting publisher yield.
The three models in one paragraph each
| Model | You pay when… | Who holds conversion risk |
|---|---|---|
| CPC (cost per click) | A user clicks to your landing page | Advertiser (post-click funnel is yours) |
| CPA (cost per acquisition) | A defined event fires (reg, FTD, lead) | Network + publisher (pre-event funnel is theirs) |
| eCPM (effective cost per mille) | 1,000 impressions are served | Advertiser (brand/exposure goal; weak post-impression guarantee) |
Everything else is nuance.
CPC — cost per click
You pay when a user lands on your page. The publisher and network take risk on whether that click converts.
When CPC wins
- New offers with unknown conversion rates. You need data before committing to a CPA.
- Strong landing page teams. You can A/B the LP without renegotiating deal terms.
- Cold-traffic discovery. Blocked and declined traffic is high-intent but not pre-qualified — CPC lets you test without overpaying for conversions that may not materialize.
- Geo expansion. Entering a new market where historical CVR is zero? CPC is the honest entry bid.
When CPC loses
- Long multi-step funnels (registration → KYC → first deposit). Drop-off after the click is yours to absorb.
- Thin LP optimization resources. If you cannot iterate on landing pages, you are paying for clicks into a leaky bucket.
- Markets with click-quality concerns. ReTarget filters bots and invalid traffic — but CPC always has more exposure to quality variance than CPA.
2026 pricing realism (iGaming)
| Geo tier | Typical CPC range | Notes |
|---|---|---|
| Tier-1 EU (DE, UK, NL) | $0.80–$2.50 | High competition, strong LP required |
| Tier-2 EU / CA / AU | $0.35–$1.20 | Moderate competition |
| LATAM | $0.15–$0.65 | Volume play; test CPC before CPA |
| Eastern Europe | $0.12–$0.45 | Price-sensitive; overlay CTR often high |
| Blocked / declined overlay | $0.40–$3.00 | Premium vs. display — intent is higher |
Blocked-traffic CPC often clears above open-web display because the visitor already demonstrated category intent.
CPA — cost per acquisition
You pay when a defined event fires. Registration, first deposit, qualified lead — you pick the event. The network and publisher take all upstream risk.
When CPA wins
- Mature offers with stable conversion rates. You have 90+ days of CVR data per geo.
- Regulated funnels where the conversion is meaningful. FTD and KYC-pass are hard events — worth paying for directly.
- Advertisers who do not want to think about LP optimization. You set a CPA cap; the network optimizes delivery.
- Budget predictability. Finance teams prefer "we pay $X per depositor" over variable CPC spend.
When CPA loses
- Test campaigns. Publishers will not route volume to an unproven CPA — the network needs confidence in conversion.
- Server-side measurement gaps. If you cannot fire the conversion event reliably (pixel blocked, app-only tracking), CPA accounting breaks.
- Low-volume geos. Not enough conversions to optimize; CPC or eCPM may deliver more total volume.
2026 pricing realism (iGaming)
| Event | Typical CPA range | Notes |
|---|---|---|
| Qualified registration | $25–$300 | Varies wildly by geo and vertical |
| First-time deposit (FTD) | $80–$600 | Tier-1 EU at top of range |
| KYC-pass (no deposit) | $15–$80 | Harder to price; less common |
| Lead (non-gaming) | $5–$40 | Fintech / crypto adjacencies |
Publisher note: CPA deals require conversion tracking integration. Read Advertiser guide for postback setup.
eCPM — effective cost per mille
You pay per thousand impressions. Not clicks, not actions — eyeballs on your offer.
When eCPM wins
- Brand campaigns where impression count is the goal, not immediate conversion.
- Inventory-rich placements — Geo Popup and full-screen overlays bank many impressions per blocked session.
- Benchmarking channel quality — eCPM is the lingua franca for comparing overlay vs display vs native.
- Markets where you need presence before performance data exists.
When eCPM loses
- Direct-response advertisers without strong creative. Paying for impressions with weak CTR is expensive attention.
- Low-value geos where an impression does not equate to meaningful brand exposure.
- Performance-only finance teams — eCPM requires comfort with top-of-funnel spend.
2026 pricing realism (iGaming, blocked traffic)
| Placement type | Typical eCPM | Why |
|---|---|---|
| Geo Popup (blocked) | $0.40–$8.00 | High intent; few impressions per session but strong CTR |
| Decline Popup (KYC fail) | $1.50–$12.00 | Fewer impressions, much higher engagement |
| Open-web display (comparison) | $0.10–$2.00 | Lower intent baseline |
Blocked-traffic eCPM often beats display-network averages because the visitor is mid-intent, not browsing passively.
How risk flows through the funnel
Understanding bid models is understanding who eats the drop-off:
Impression → Click → Registration → KYC → FTD
│ │ │ │ │
eCPM CPC ────────────────────────────┐
CPA ───────────┘
- eCPM covers only the impression. Everything after is free to the advertiser (but may never happen).
- CPC covers through the click. Registration onward is advertiser risk.
- CPA covers through the defined event. Everything before is network/publisher risk.
Pick the model where you have the most information and optimization leverage.
The 2026 reality: run all three
In a classical affiliate network you pick one model per offer. The publisher gets stuck with whichever the advertiser chose. That made sense in 2010.
In 2026 it is a bug, because:
-
Conversion latency varies by geo. Tier-1 markets have fast funnels (CPA works). Tier-3 markets often need brand exposure first (eCPM works).
-
Inventory varies by overlay type. Geo Popup yields high impression counts; Decline Popup yields fewer but higher-intent sessions. Different models suit each.
-
Advertisers test on CPC, scale on CPA. Forcing a single model burns test budget or starves publishers during the learning phase.
-
Auction efficiency requires normalization. You cannot rank a $1.50 CPC bid against a $120 CPA bid without converting to a common currency. That currency is eCPM — internally.
How ReTarget.gg handles this
ReTarget.gg's auction lets advertisers fund a single prepaid balance and select CPC, CPA, or eCPM per campaign. The network normalizes everything to eCPM internally for ranking. The advertiser pays under whichever model they chose. Publishers see one unified yield number.
Setup: Advertiser deposits → Advertiser guide.
Decision matrix
| Your situation | Recommended model | Why |
|---|---|---|
| New offer, unknown CVR | CPC | Limits downside while collecting data |
| Mature offer, stable funnel, 90d+ data | CPA | Predictable unit economics |
| Brand awareness in regulated geos | eCPM | Presence without conversion pressure |
| Mix of geos at different maturity | All three | Right model per campaign, one auction |
| Blocked / declined overlay (testing) | CPC | High intent but variable downstream |
| Blocked / declined overlay (scaled) | CPA or eCPM | Once CVR or impression value is known |
| Budget-capped quarterly spend | CPA | Finance-friendly cap |
| Aggressive market share push | eCPM + CPC | Volume + learning |
Worked example: same traffic, three outcomes
Scenario: 10,000 blocked sessions in Brazil. Overlay CTR: 12%. CPC bid: $0.45. CPA bid: $55 FTD. eCPM bid: $2.80.
| Model | Math | Advertiser cost | Publisher revenue |
|---|---|---|---|
| CPC | 10,000 × 12% × $0.45 | $540 | $540 × revshare |
| CPA | 10,000 × 12% × 8% CVR × $55 | $5,280 | $5,280 × revshare |
| eCPM | 10,000 × $2.80 / 1,000 | $28 | $28 × revshare |
CPC is cheapest for the advertiser when CVR is uncertain. CPA wins when CVR is proven and high. eCPM is cheap but does not guarantee clicks.
The auction picks the highest normalized eCPM — so a strong CPC bid can beat a weak CPA bid and vice versa. That is the point of unification.
Common mistakes
Choosing CPA before you have 500+ conversions per geo. You will overpay or under-deliver.
Staying on CPC forever. Once CVR is stable, CPA often unlocks more publisher volume.
Ignoring eCPM for benchmarking. Even if you buy on CPC, eCPM equivalent tells you if the channel beats display.
Different networks per model. Fragmented publisher relationships reduce fill rate. One auction with three models beats three separate integrations.
FAQ
Can publishers choose which model they accept? Publishers see unified yield. The network handles model mixing — you do not negotiate per deal.
How is eCPM calculated internally for CPC bids? Expected value: CPC × predicted CTR × 1,000. Prediction uses historical overlay performance per geo and placement.
What conversion events does CPA support? Registration, FTD, KYC-pass, and custom server-side events via postback. See Advertiser guide.
Is prepaid balance required? Yes — advertisers fund ahead. This eliminates publisher payment risk common in net-60 affiliate networks.
Which model do blocked-traffic advertisers prefer? CPC for testing (60%), CPA for scale (30%), eCPM for brand (10%) — based on our campaign mix. Your mileage varies.
Related reading
- How to monetize geo-blocked traffic in iGaming
- What to show after a failed KYC check
- ReTarget.gg vs classic networks
- Advertiser overview
Ship your first campaign: Advertiser deposits — 10 minutes to live.