Conversion rate optimization for affiliates

1. Performance is a fishbowl

Affiliate marketing is one of the most honest places you can work. Every click, every conversion, every payout is in the dashboard. If you're doing a great job, the chart goes up and the payouts get bigger. If something needs work, the chart stays flat or dips — and the same dashboard tells you exactly where to look. The numbers don't lie, and that's the good news: you never have to guess whether you're winning.

Our job is to make sure the fishbowl is clean and the rules are fair. We run the offer, we pay you per conversion, and we give you the tracking platform, the deeplink parameters, the click report, the postback infrastructure, and the creative library. Your job is the bit that's genuinely yours: which audience to reach, which channel to invest in, which landing experience to send them to, and what to change when the numbers point at something. This guide is the playbook for that work — how to read the chart, find the lever, and take your results to the next level.

2. How this guide applies to your traffic mode

Around 30% of our affiliates are paid-traffic media buyers. The other 70% are some mix of webmasters running content sites, SEO specialists targeting search intent, community moderators in Reddit / Discord / Telegram / forums, and people who own engaged email lists or content audiences. The principles in this guide apply to all of you. The levers differ.

Throughout this guide, where a tactic is mode-specific you'll see a colored block like the one below mapping the principle to each mode. The four modes:

If you're running more than one mode (a content site that also has an email list, for example), treat each as a separate campaign for reporting purposes. Aggregating across modes hides which one is actually working.

3. The four numbers that matter

Most affiliate dashboards — Everflow included — give you dozens of metrics. Four of them carry the weight; the rest are noise until you have meaningful volume.

Clicks

Recorded the moment your tracking domain receives the request. If clicks are recording, your link is structurally working. If not, you have a tracking issue, not a performance issue. (See clicks recording, no conversions for the troubleshooting tree.)

Conversions

New paying subscriptions attributed to your link, in one of three states: pending (in verification), approved (cleared, counts), or rejected (chargeback or refund inside the verification window). Optimization decisions should be made on approved conversions only — pending is just a forecast.

EPC — earnings per click

Total earnings divided by total clicks. The single most useful number in your reports because it folds together both your conversion rate and your country mix into one figure. EPC is universal: it tells everyone "what is each click worth?" What changes is what you compare it to.

Conversion rate (CR)

Conversions divided by clicks, as a percentage. Tells you how well your traffic matches the offer. EPC tells you whether the match is profitable (or — for unpaid modes — whether the traffic is worth the effort to source). Read together: high CR with low EPC usually means you're converting Tier 3 traffic at a great rate — real performance, just at lower per-conversion payouts. Low CR with high EPC usually means a small audience converting hard at high payout — also real, but harder to scale.

4. The sample size problem

Both EPC and CR swing wildly at small numbers, and most affiliates start drawing conclusions far too early. A single conversion on 50 clicks looks like a 2% conversion rate, but the next 50 clicks could yield zero — the "true" rate is somewhere wider than what 50 clicks will tell you.

Practical thresholds:

• Under 100 clicks per segment — you are looking at noise. Treat conclusions as suspicious.
• 100–300 clicks per segment — directional signal. Use it to choose what to test next, not what to scale.
• 300–1,000 clicks per segment — increasingly reliable. You can act on differences this size.
• Over 1,000 clicks per segment — you have a real read. If something is broken at this volume, it is broken for real reasons.

"Per segment" matters. If you have 1,000 clicks total but you're slicing them into ten countries, three landing experiences, and two creatives or pages, no single segment has enough data to draw a conclusion. Either send more traffic or aggregate up to a level that has volume.

Sample size hits SEO and community traffic harder than paid, because volume comes more slowly. If you're writing one new page per week or posting in one community per week, you may need months of data to read a segment confidently. That's a feature, not a bug — it forces you to make a few high-conviction bets rather than thrashing weekly.

5. Eight dimensions to slice your data by

Aggregate numbers hide everything. The same EPC of $0.30 can mean "every traffic source breaks even" or "one traffic source is making me $1.50 EPC and three are losing me money." You cannot make decisions from the aggregate. Slice by these dimensions, in roughly this order of impact:

1. Country (geo) — the single biggest driver of EPC variance. T1 vs T3 traffic at the same conversion rate produces wildly different EPC.
2. Sub-parameter / deeplink — different sub1–sub4 values send users to different landing experiences. Different experiences convert differently.
3. Channel / source — for paid, this is platform and ad set. For SEO, the search query and the page that ranked. For community, which sub / server / forum. For email, which segment or send.
4. Creative / content / placement — the ad, the article, the post, the email. The single highest-leverage variable inside any mode.
5. Device type — mobile and desktop convert differently because the landing experience renders differently.
6. Hour of day in the user's local timezone — adult AI demand follows a daily clock, not a weekly one. Late evening / overnight in the user's timezone is the consistent peak conversion window. Day-of-week effects exist but are dwarfed by hour-of-day; treat the daily rhythm as the dominant time signal.
7. Audience cohort / age of campaign — channels and audiences degrade as they saturate. Same setup that worked in week one stops working in week six.

Everflow can slice by all of these — most are available in the standard click and conversion reports without any custom setup. If a dimension is not visible by default, it is almost always available as a column you can add or a group-by you can apply.

6. Country geo split: the dominant variable

If you remember one thing from this guide: your country mix dominates your EPC. Two affiliates running the exact same content or ads at the exact same conversion rate can have EPCs that differ by 40% just because of which countries their audiences come from.

ourdream.ai pays by country tier. The published rates on the Levels page show what each tier pays per conversion. A summary view:

What this means in practice

Two affiliates on Pink VIP, both converting at 1.2% on 5,000 clicks — 60 conversions each. Same skill level, same offer. But:

• Affiliate A's audience is 80% Tier 1. Their 60 conversions = 48 × $40 + 12 × $35 = $2,340. EPC: $0.47.
• Affiliate B's audience is 80% Tier 3. Their 60 conversions = 12 × $40 + 48 × $30 = $1,920. EPC: $0.38.

Identical conversion rate, $420 difference per 5,000 clicks. The same gap exists whether those clicks came from Google Ads, organic search, a Reddit post, or a newsletter. Geo mix is upstream of channel choice — but it's not just about chasing T1. Some of the most profitable affiliates we work with built their playbook around a single non-T1 country they understand better than anyone else.

The country-by-country reality check

Tier groupings are payout buckets, not strategy. Within each tier countries behave very differently — both in conversion rate and in propensity to pay. Real patterns we see:

How to act on this — by mode

Step one is the same for everyone: pull a country breakdown of your last 30 days, group clicks/conversions/EPC/CR by country, and identify which three to five countries do most of the work. What you do next depends on how you drive traffic.

For deeper context on EPC and CR interpretation, see the EPC and CR help article.

7. Landing experience and deeplinks

After geo, the landing experience does more to move conversion rate than anything else inside the funnel. ourdream.ai supports deeplinking via sub1 through sub4 — these route the user to different experiences when they arrive. The defaults work, but the defaults are not optimal for every audience.

What to test in the landing experience

1. Deeplink choice — try different combinations of sub1–sub4 for the same audience. A chat-led entry point may convert better for users who came in from a chat-style hook than the default homepage. This applies equally whether the "hook" is an ad creative, a blog post intro, a community post, or an email subject line.
2. Page load speed on mobile — open your link from a mobile device on cellular, time it. If the user waits more than 3 seconds for visible content, your referrer-to-landing transition is leaking conversions you cannot see in any report.
3. Mobile vs desktop split — slice by device. Most adult traffic is mobile-heavy, but the ratio differs sharply by source. SEO traffic tends to skew desktop; community traffic skews mobile; paid varies by platform. Match your creative or content to the device your traffic actually arrives on.
4. Hook-to-landing match — if your hook (ad, headline, post, subject line) sets a specific expectation, the landing page should deliver on it immediately. Mismatch is a top cause of high-bounce, low-CR traffic. The user clicks because they were promised X; if they land on Y, they bounce.

How to set deeplinks at the link level is covered in the tracking links onboarding guide — sub1 through sub4 explained, with safe defaults.

8. Hour-of-day patterns and the daily clock

A common assumption from other affiliate categories is that adult-adjacent traffic has a strong weekend cycle. In our space it doesn't. Day-to-day performance is remarkably consistent— Tuesday looks a lot like Saturday at the same hour. The real time variance is the daily clock, in the user's local timezone: nighttime is peak, daytime is trough, every day of the week.

This sounds small but it's the single most actionable time signal in the data. It also has a sharp implication for anyone scaling internationally: time follows the geo. A US-targeted campaign and a Japan-targeted campaign both peak around 9pm to 1am local — but those are eight to thirteen hours apart on the wall clock. If you blend timezones in a single campaign without timezone-aware scheduling, you'll see a flatter, weaker pattern than the underlying signal actually contains.

What to look for

• Hour of day in the user's local timezone — the dominant signal. Late evening through early-morning (roughly 9pm to 2am local, with the deepest peak around 10pm to midnight) is consistently the highest-converting window. Midday is consistently the weakest. This pattern shows up regardless of how the traffic arrived.
• Day-of-week is a minor secondary effect. A small weekend lift exists in some segments but it's usually within the noise band of normal day-to-day variance — don't make scheduling decisions on day-of-week unless you have a very large sample size confirming the effect for your specific traffic. Pull your last 30 days by day-of-week before you optimize against it; in most cases the bars look about the same height.
• Cohort decay — pull a 7-day rolling EPC for each campaign or content piece. Most channels peak in the first one to two weeks and decay as the audience saturates. Knowing where on the curve you are tells you whether to refresh or retire.

What to do with the pattern — by mode

9. The optimization loop

Optimization is not a one-time event. It's a loop you run regularly, with the same four steps every time. The loop's cadence varies by mode (paid runs daily, SEO runs monthly), but the steps don't.

1. Hypothesis. Look at your latest report. Form a specific, falsifiable claim. Bad: "I think Reddit performs well." Good: "I think traffic from sub1=community to deeplink sub2=female has a 40% higher EPC than to sub2=mixed."
2. Test. Run two variants with everything held constant except the variable you're testing. Same audience, same time window, different deeplinks — or same content angle, different communities — or same email body, different subject lines. Don't change two things at once.
3. Measure. Wait for enough volume per segment. (See the sample size section above.) Compare EPC and CR on each variant.
4. Iterate. Adopt the winner. Form the next hypothesis. Repeat. Most of the time the winning variant becomes your baseline; sometimes the test reveals a third option you should test next.

Affiliates who run this loop scale. Affiliates who don't, don't.

10. Common patterns and what they tell you

A pattern-recognition cheatsheet. None of these is a hard rule — they are starting hypotheses to investigate.

11. Kill criteria and double-down criteria

The single cleanest decision rule for affiliate work is the 10% margin line: if your revenue from a segment is at least 10% above your cost, keep going and look for ways to scale. If it's below, fix it or kill it. Set this line before you launch, not after the results come in — otherwise you'll rationalize keeping underperforming work alive because you've already invested in it.

When to kill a test

Don't kill on a single data point. Wait for at least 5 conversions (or, if you've burned 3× your max-CPA in spend with zero conversions, that's also a fair signal). One conversion tells you almost nothing — the "true" CPA is wider than what 1–4 conversions can tell you.

When to double down

• Margin consistently above 10% across 5+ conversions per segment. Below 5, you might be reading noise. Aim for 1,000+ clicks per segment before treating a margin number as confidently scalable.
• Pattern holds across at least two distinct time periods. Single-week spikes are sometimes real, often coincidence — especially around weekends, paydays, and holidays.
• You can name the reason it's working. If you can't explain why a campaign / page / community / send is profitable, you can't defend it when it stops being profitable. "T1 traffic + tools angle + Wednesday morning" is a defensible reason. "Just lucky" is not.
• Aim higher than 10% before aggressive scaling. 10% margin is the keep-running line. For 5–10× scaling, look for segments running at 30%+ margin — they have room to absorb the EPC erosion that always comes with scale.