CPS, CPL, or CPI: which affiliate payout model actually fits your programme
Every performance programme eventually asks the same question: what are we actually paying for? Get the answer wrong and you either overpay for traffic that never turns into revenue, or you underpay publishers for the work required to convert it — and they leave for a network that pays fairly. This is a practical guide to the three payout models that cover most of affiliate marketing, and how to tell which one your funnel actually needs.
CPS pays on a completed sale and carries the lowest advertiser risk. CPL pays on a qualified lead and suits longer sales cycles. CPI pays on a verified app install and works best paired with a retention bonus. Many programmes end up running more than one at once.
The one question that decides your model
Before comparing formulas, answer this: how far down the funnel does the publisher's traffic need to travel before you're willing to pay for it? A form fill is a different job than a closed sale, and a closed sale is a different job than an app install. Payout models exist because "drive traffic" is not one job — it's three, and each has its own risk profile, its own commission economics, and its own type of publisher who's good at it.
CPS — Cost Per Sale
CPS pays a commission — usually a percentage of order value, sometimes a flat fee — only when a referred visitor completes a purchase. It's the model most people picture when they hear "affiliate marketing", because it's the one with the tightest link between spend and revenue.
How CPS is calculated
Typically a percentage of the sale value (5%–30% depending on margin and vertical) or a fixed amount per order. A publisher sending a $200 order on a 15% CPS rate earns $30, full stop — no partial credit for visitors who browsed and left.
Where CPS works best
E-commerce, fashion, consumer electronics, digital products, and any programme where the purchase decision happens in a single session or a short, trackable window. It rewards publishers who can move someone from interest to checkout — content that ranks for buying-intent keywords, comparison pages, coupon and deal sites.
Where CPS falls short
Long or multi-device buying journeys break attribution unless tracking is solid (more on that in our S2S vs pixel tracking guide). CPS also asks a lot of new publishers — driving a completed sale is harder than driving a click, so it's not the easiest entry point.
CPL — Cost Per Lead
CPL pays when a visitor completes a defined action short of a purchase: a form submission, an account sign-up, a quote request, a demo booking. The advertiser is paying for a qualified prospect, not a closed deal.
How CPL is calculated
Almost always a flat fee per qualified lead, commonly anywhere from a few dollars to well over $100 depending on how valuable the eventual customer is — insurance and B2B software leads sit at the high end because the lifetime value downstream justifies it.
Where CPL works best
Insurance, financial services, real estate, B2B software, education — categories with sales cycles too long or too high-touch to expect a same-session purchase. It's also the friendliest model for newer publishers, since the visitor action required (fill a form) is a smaller ask than "buy now".
Where CPL falls short
Lead quality varies enormously. Without strict qualification criteria (real phone number, real intent, not a bot or a duplicate), advertisers end up paying for leads that were never going to convert — which is why most serious CPL campaigns include lead-scrubbing rules publishers have to meet before a lead counts.
A CPL campaign with no duplicate-detection or intent verification is an open door for low-quality leads. Set qualification rules before launch, not after you've paid for the first batch of bad ones.
CPI — Cost Per Install
CPI pays for a verified app install, most common in mobile user-acquisition campaigns. It's the simplest event to define — the app either installed on a device or it didn't — but the simplest event to define isn't always the most valuable one to pay for.
How CPI is calculated
A flat fee per verified install, tracked through an attribution SDK or postback. Rates vary hugely by platform, GEO, and app category — a finance app install in a tier-1 market can be worth many times a casual game install in a lower-ARPU GEO.
Where CPI works best
Mobile games, utility apps, and any product where a large volume of installs is the growth priority and post-install engagement is measured and optimised separately.
Where CPI falls short
An install is not a user. CPI alone rewards volume, not retention, which is why more advertisers now layer in event-based bonuses (CPI + a payout for the first in-app purchase or day-7 retention) rather than running CPI in isolation.
Side-by-side comparison
| Model | Paid when | Typical advertiser risk | Best-fit vertical |
|---|---|---|---|
| CPS | Sale completes | Lowest — pay only on revenue | E-commerce, retail, digital products |
| CPL | Qualified lead submitted | Medium — depends on lead quality controls | Insurance, finance, real estate, B2B |
| CPI | App installs | Medium-high without engagement bonuses | Mobile apps and games |
How advertisers should choose
Start from your sales cycle, not from what competitors run. If your purchase happens in one session, CPS keeps your cost tied directly to revenue. If your real value shows up weeks after the first touch, CPL lets you pay for interest now and qualify it later. If growth is measured in active users on a specific platform, CPI is the entry event — but pair it with a retention or in-app-event bonus so publishers are optimising for the same outcome you are.
How publishers should choose
Match the model to what your traffic is actually good at. If your audience is bottom-funnel and ready to buy — deal sites, comparison content, retargeting — CPS usually pays the most per visitor because you're capturing the highest-value event. If your traffic is earlier-funnel — informational content, broad reach, social — CPL rewards you for interest you can generate at volume without needing to close the sale yourself.
Check your EPC under each model before committing traffic at scale — it reflects what a campaign actually pays per click sent, and tells you more than the headline payout rate does.
A hybrid approach: stacking models by funnel stage
Some of the most durable programmes don't pick one model — they stack them. A small CPL for the qualified lead, plus a CPS kicker if that lead becomes a paying customer within 60 or 90 days, splits the risk fairly: the advertiser pays a little upfront for the qualification work, and pays the larger amount only when it's justified by actual revenue. This structure is common in insurance, subscription software, and any category where the lead and the sale are genuinely separate events worth rewarding separately.
Whichever model you run, none of it works without accurate conversion data. A CPS programme with leaky attribution underpays honest publishers; a CPL programme without duplicate detection overpays for leads that were never real. Tracking method is the foundation every payout model sits on — see our breakdown of S2S postback vs pixel tracking.
What Kapvexa tracks for each model
Kapvexa's platform supports CPS, CPL, and CPI campaigns side by side, with server-to-server postback tracking as the default so conversions are checked and approved before payouts go out — not just counted on a click. Advertisers set the model and the terms; publishers see approval and payout status for every conversion in real time.
Mistakes to avoid
| ❌ Costly habit | ✅ Better approach |
|---|---|
| Picking CPS for a long, high-touch sales cycle | Use CPL, or a CPL-then-CPS hybrid, so you pay for interest before the sale closes weeks later |
| Running CPL with no lead-quality rules | Set duplicate-detection and intent verification before launch, not after paying for bad leads |
| Running CPI in isolation | Layer in a retention or first-event bonus so publishers optimise for real users, not just installs |
| Choosing a model based on what competitors run | Choose based on your own sales cycle and how far down the funnel you're willing to pay |
| Ignoring tracking quality | Confirm conversions server-to-server — see our S2S vs pixel guide |
FAQ
Can one campaign use more than one payout model?
Yes. Hybrid models — a small CPL for the lead plus a smaller CPS on the eventual sale — are common in insurance, real estate, and subscription businesses where the lead and the sale happen weeks apart.
Which model is easiest for a new publisher to start with?
CPL, generally. The action required from the visitor is smaller (a form, not a purchase), so conversion rates are higher and a new publisher gets feedback on their traffic quality faster.
Why do advertisers prefer CPS over CPL?
CPS ties every dollar spent to actual revenue, so there's no risk of paying for leads that never convert. It shifts more of the qualification work onto the publisher, which advertisers see as lower risk.
Is CPI still relevant outside of mobile apps?
Its core use is still app installs, but the same install-then-activate structure shows up in software trials and browser extensions, so the underlying logic — pay for the install, then optimise for activation — applies more broadly.
How do I know if my payout model is wrong for my funnel?
The clearest signal is a mismatch between effort and payout — if publishers are doing sale-level persuasion work but getting paid lead-level rates, they'll drop the campaign, or if you're paying for leads that rarely become revenue, the model is exposing you to cost without matching upside.
Ready to run CPS, CPL, or CPI on one platform?
Set your model, approve every conversion, pay on schedule.
Talk to Kapvexa