Affiliate Marketing KPIs Explained: EPC, CR, LTV, ROI

Two offers land in your inbox on the same day. The first converts at 3% and pays $40. The second converts at 0.9% and pays $180. The first has more than three times the conversion rate, and it is the worse offer by a wide margin.

Ten thousand clicks on offer one gives you 300 conversions and $12,000. Ten thousand clicks on offer two gives you 90 conversions and $16,200. The affiliate who ranks offers by conversion rate leaves $4,200 on the table and never knows it.

That is the whole problem with affiliate marketing KPIs. Each one answers a specific question accurately and every other question badly, and most bad decisions come from using a metric outside the question it was built for. This piece goes through EPC, CR, CPA, ROI, ROAS, RPM, payout ratio and LTV — the formula for each, a worked calculation, and the situations where each one will mislead you.

The metrics at a glance

EPC: the number that makes offers comparable

EPC = total revenue ÷ total clicks.

Take the opening example. Offer two: 10,000 clicks × 0.9% = 90 conversions, 90 × $180 = $16,200, so EPC is $16,200 ÷ 10,000 = $1.62. Offer one comes out at $12,000 ÷ 10,000 = $1.20.

EPC collapses conversion rate and payout into a single figure. That is exactly why it works for cross-offer comparison: it does not care how the money got made, only how much each click was worth.

Two things to be careful about.

First, EPC is a revenue metric, not a profit metric. An offer with a $1.62 EPC is a disaster if the traffic that converts on it costs $2.10 a click. Always read EPC against the CPC for the specific traffic that produced it — not your blended average CPC, which hides everything interesting.

Second, EPC only compares fairly when the traffic is comparable. An EPC calculated on your own email list will be several times an EPC calculated on cold push traffic, and comparing them tells you about your traffic rather than about the offers. Compare EPC within a source.

Network EPC is a hint, not a fact

Networks publish an average EPC for each offer. Treat it as directional. It is averaged across every affiliate running every kind of traffic in every geo, and it is frequently dominated by one large affiliate with a channel nothing like yours. A high published EPC means the offer can work. It does not mean it will work for you.

Conversion rate: useful for diagnosis, useless for ranking

CR = conversions ÷ clicks × 100.

Conversion rate is a funnel health metric. It tells you whether the traffic and the page are matched, and it is the fastest way to spot a problem: a CR that halves overnight usually means a broken landing page, a geo that started serving a different offer, or a traffic source that changed its inventory.

Where it fails is comparison. A 6% conversion rate on a free trial signup and a 0.4% conversion rate on a $2,000 purchase are not on the same scale, and neither number tells you what the click was worth.

Also be precise about the denominator. Click-to-conversion, landing-page-view-to-conversion, and registration-to-deposit are three different rates and people call all three "conversion rate" in the same conversation. Registration-to-deposit is the one that exposes traffic quality most clearly in finance and iGaming — a source producing plenty of registrations and almost no deposits is sending you curiosity, not intent.

CPA and the payout ratio

CPA = total spend ÷ conversions. This is your cost per acquisition, which is not the same as the CPA payout an advertiser offers you. Keep the two words separate in your head or you will confuse yourself in every conversation about margins.

Spend $8,000 to get 120 conversions and your CPA is $66.67. If the offer pays $110, your margin per conversion is $43.33 and your ROI is 65%.

Payout ratio = your commission ÷ the revenue the user actually generates for the advertiser.

This one is harder to calculate because you need the advertiser's side of the equation, but it is the number that tells you whether a deal is fair. If a $180 CPA is attached to users who generate $600 of net revenue over twelve months, your payout ratio is 30%. If the same user profile is worth $250, a $180 CPA is generous and probably will not survive the next renegotiation.

You will rarely be handed this number. You can infer it over time by running a revshare deal alongside a CPA deal on similar traffic and comparing what the same cohort produces each way — which is the practical core of choosing between revenue share and a one-time CPA.

ROI and ROAS: same information, different framing

ROAS = revenue ÷ spend. ROI = (revenue − spend) ÷ spend.

Spend $8,000, earn $12,400.

• ROAS = 12,400 ÷ 8,000 = 1.55x, or 155%
• ROI = (12,400 − 8,000) ÷ 8,000 = 55%

Both describe the same campaign. ROAS is a multiple and reads fast; ROI is the honest one because it treats the money you spent as a cost rather than counting it inside the return. The failure mode is mixing them in the same conversation. Someone reports 155% and someone else hears ROI, and a campaign making 55% gets treated as a campaign making 155%.

Neither figure includes your actual costs beyond media. Tracker fees, hosting, tools, VAs, chargebacks, and the time you spent are all outside the calculation. A campaign running at 15% ROI on paper can be losing money once real costs land.

Why ROI on revshare needs a time horizon

On a fixed CPA deal, ROI settles quickly. Media spend and revenue both land within days, and the number you compute in week two is close to final.

On revenue share, ROI is a moving target, and quoting it without a horizon is meaningless. Consider a cohort of 100 first-time depositors acquired in January for $9,500 in media spend — a $95 acquisition cost per depositor — on a 40% revenue share.

The same cohort is a catastrophe at month one and a solid result at month twelve. Both readings are arithmetically correct. Only one of them is a decision.

This is the trap that empties bank accounts. An affiliate reads month-one ROI, panics, and cuts a campaign that was six months from a 24% return. Or the reverse: an affiliate scales aggressively on a revshare deal without modelling the cash gap, and runs out of working capital in month three while owed money that has not been earned yet.

Two things follow from the table. First, know your break-even month before you launch — here it falls in month seven. Second, check whether the deal carries negative carryover, because a losing month that rolls forward changes this entire model and can push break-even out indefinitely.

The comparison people skip

That same cohort on a flat $110 CPA offer would return $11,000 against $9,500 in spend — a 15.8% ROI, settled within about 30 days.

Revshare wins on the headline number: 24.2% against 15.8%. But the CPA money comes back in a month and can be redeployed. If you could genuinely recycle capital monthly at 15.8%, twelve turns compounds to roughly 5.8x your starting capital, against 1.24x for the revshare cohort.

You cannot actually do that — caps, saturation, rising CPCs, and seasonality all bite long before month twelve — but the direction matters. Fast-settling deals let you compound. Slow-settling deals give you a higher ceiling and a heavier cash requirement. Which one suits you depends on your working capital, not on which number is bigger.

RPM and where it belongs

RPM = (revenue ÷ impressions) × 1,000. Revenue per thousand impressions.

420,000 impressions producing $3,780 gives an RPM of $9.00.

RPM is the right unit when you buy or sell on impressions — display, native, most site monetisation. It quietly incorporates click-through rate, which EPC does not. An offer with a strong EPC and a creative nobody clicks will have a poor RPM, and if you are paying per impression, RPM is the number that reflects your reality.

If you buy on clicks, RPM mostly adds confusion. Use EPC against CPC.

LTV and cohort thinking

LTV = total commission from a cohort ÷ number of users in that cohort, over a stated window.

From the revshare table: $11,800 ÷ 100 depositors = $118 per depositor at a twelve-month horizon. Against a $95 acquisition cost, that is an LTV to CAC ratio of 1.24.

The word doing the work there is cohort. Blended LTV — total commission divided by total users across all time — is worse than useless, because it mixes users who have had a year to generate revenue with users who signed up last week. It sags whenever you scale and looks great whenever you stop spending. Neither movement means anything.

Group users by acquisition month, track each group separately, and you get answers that hold up:

• Does a January cohort behave like a June cohort? If it decays faster, either traffic quality slipped or the advertiser changed something on their side.
• Does traffic from one placement retain better than another? Push your placement-level sub-IDs through to the advertiser and you can see this directly instead of guessing.
• How much of your total LTV lands in the first 30 days? That ratio is what tells you how much working capital scaling will require.

For high-ticket and SaaS, the cohort question shifts from decay to renewal — how many are still paying at month 12 — but the method is identical, and it is the whole basis of building recurring commission income rather than one-off spikes.

The reporting lag problem

Conversions are attributed to the date of the click. A click on Monday that converts on Thursday improves Monday's numbers retroactively, three days after you already looked at them.

This makes recent data systematically pessimistic. As a rough rule of thumb on offers with a same-session conversion, most of the eventual total is visible within a day. On offers with a genuine consideration period — brokers, high-ticket, anything requiring a deposit or a demo — a meaningful share of conversions lands days later, and the first day's reading can understate the final result substantially.

Add to that the approval cycle. Pending conversions become approved or rejected. Chargebacks and reversals arrive weeks after the fact. If you treat pending revenue as final, your reported profit runs ahead of your actual bank balance, and the correction usually lands right after you have scaled spend on the strength of it.

Three habits handle this:

1. Compare like-aged data. Yesterday against last month is not a comparison. Yesterday against the same weekday last week, both at the same age, is.
2. Track a maturation curve for each offer. Record what percentage of the final total was visible at 24 hours, 72 hours, and 7 days. After a few weeks you can read day-one data with a correction factor instead of a guess.
3. Report on approved revenue, forecast on pending. Two columns, never one.

Which metric for which decision

• Choosing between two offers: EPC, read against the CPC of the traffic you would send.
• Setting a bid: target CPA derived from payout and margin target.
• Deciding whether to keep a campaign alive: ROI, with a horizon appropriate to the deal type.
• Diagnosing a sudden drop: CR first, then CR by step, then by sub-ID.
• Judging a placement: profit and EPC. Never conversion count.
• Deciding how hard to scale: LTV to CAC, plus the payback window.
• Negotiating a payout: payout ratio, if you can estimate it.

Mistakes that show up repeatedly

Averaging averages. The mean of your daily ROI figures is not your period ROI. Sum revenue and sum spend, then divide once.

Judging on revenue instead of margin. A campaign doing $60,000 at 4% margin is more fragile than one doing $15,000 at 40%. One CPC increase kills the first.

Ignoring the denominator's quality. ROI on 40 conversions is a real signal. ROI on 3 conversions is a coin flip with a percentage sign on it.

Counting unapproved revenue. Pending is not paid.

Comparing EPC across traffic types. Your email EPC and your push EPC are different universes.

Letting the tracker assume a payout. If your postbacks pass a static payout rather than the dynamic value from the advertiser, every downstream metric is built on a number you invented. That single misconfiguration corrupts EPC, ROI, and LTV at once, which is why postback setup is a reporting concern and not just a plumbing one.

Working with Shazam on measurement

Getting these numbers right is mostly an infrastructure problem, and it is one we take off our partners' hands. Tracking is built and maintained for you, with dynamic payouts passed properly so EPC and ROI reflect what you actually earned, and placement-level data pushed through to advertisers where possible so cohort and retention questions have real answers rather than estimates. We build the landing pages as well, which keeps the click IDs and sub-IDs intact end to end.

Partners run on a 50% revenue share across crypto, forex, iGaming, gaming, high-ticket and tech, with a dedicated manager who can pull advertiser-side cohort data when your numbers and theirs disagree. Onboarding runs through our Telegram bot, and the community chat is a reasonable place to start if you want to ask first.