Conversion rate optimization for Secrets AI traffic

Most affiliates optimize the wrong step. They see a weak result, change the banner, and hope. A better approach is to find where in the funnel users drop, fix that step, and leave the rest alone. This guide walks through how to read your Everflow data for the Secrets AI offer and turn it into a weekly routine that improves earnings without guesswork.

Know the funnel you are measuring

Traffic to Secrets AI passes through a few distinct steps:

1. Impression or view on your site, ad or post.
2. Click on your tracking link, recorded in Everflow.
3. Landing on the offer page (or your pre-lander first).
4. Signup, where the visitor creates an account.
5. First purchase, where they become a paying user. This is the qualifying event for CPA.
6. Repeat purchases, such as renewals, upgrades and token purchases. These drive RevShare.

Which of these events show up as separate conversions in your dashboard depends on how your offer is set up. Ask your affiliate manager which events you can see, and confirm what counts as a qualifying conversion for your deal.

Click-to-signup vs signup-to-paid

These two ratios point to different problems, so look at them separately.

Click-to-signup is about the front end

A low click-to-signup ratio means visitors arrive and leave. Likely causes:

• The traffic does not match the offer (wrong audience, wrong age group, wrong intent).
• The ad or post promised something the landing page does not show.
• The landing page does not fit the device or language.
• Tracking or redirects add delay and visitors bounce.

Fixes live in your creative, your pre-lander, your landing page choice and your targeting.

Signup-to-paid is about expectation and audience

A healthy signup rate with few paying users means visitors are curious but not committed. Likely causes:

• The angle attracted browsers rather than buyers, for example a "free" message that does not match the product.
• The audience is in geos or on devices with lower purchasing behavior.
• Incentive-like messaging pulled in low-quality signups.

Fixes live in your angle and your traffic quality, not in the landing page.

If you only see the paid conversion in your dashboard, use your manager. They can often tell you whether the gap is at signup or at purchase.

Landing page selection

Landing page variants live in the Creatives tab of your offer in Everflow, and you pick one when you generate a tracking link. Your affiliate manager can point you to the ones that fit your traffic. Do not assume the default is best.

Match the landing page to the source:

• Character-led traffic (persona accounts, anime communities): a landing page that leads with characters or the matching style.
• Feature-led traffic (reviews, comparisons): a landing page that leads with features such as memory or custom characters.
• Language-specific traffic: the localized version for that language. Secrets runs in English, Spanish, French, German and Japanese.
• Mobile-heavy traffic: confirm the page loads fast and the signup action is visible without scrolling.

Run two landing pages at a time and split traffic evenly. To compare them, open the Flex Report (Analytics → Flex), add Offer URL as a column, add a Clicks > 300 filter so you only compare pages with enough traffic to trust, and sort by CVR. Top partners can request custom landers; ask once you have data to justify one.

Pre-lander tests

A pre-lander is a page you control between your traffic and the offer. Test one variable at a time:

• Headline (feature-led vs character-led)
• Format (character picker, short quiz, feature explainer)
• Length (one screen vs a short scroll)
• Button copy

Rules for clean tests:

• Pass the full query string through to the outbound link, every time.
• Split traffic evenly and simultaneously. Testing variant A this week and B next week mixes in timing effects.
• Wait for enough paying conversions on each side before calling a winner. A handful of conversions is noise.
• Keep a test log: date, variant, hypothesis, result.

Also test no pre-lander at all. On some sources a direct link to the offer converts better.

Geo and device cuts

Aggregate numbers hide the story. Cut every report by:

• Country. CPA is geo-tiered, so a country with a lower conversion rate can still earn more if it is Tier 1. Look at earnings per click by country, not conversion rate alone.
• Device type. Mobile and desktop often behave very differently.
• Operating system and browser. In-app browsers (Instagram, TikTok, Telegram) can break flows. A sudden gap on one browser is often a tracking or rendering problem.
• Language. If you send multilingual traffic, check that each language lands on the matching localized page.

The Flex Report (Analytics → Flex) breaks clicks and conversions down by Country, Device, Offer, Offer URL, Source ID, sub-IDs and Hour of Day. Set the timezone to your audience's before reading hour-of-day data.

If a dimension you need is not available, pass it in sub2 to sub5.

Sub-ID based pruning

Source ID and sub2 to sub5 let you see performance by source, account, creative and placement. Pruning means cutting what does not earn so budget and effort go to what does.

sub1 is reserved for your click ID, so do not prune on it.

A simple pruning process:

1. In the Flex Report, set the last 14 or 30 days and add the field you are pruning as a column (Source ID, or sub2 to sub4 for account, creative or placement). Export to CSV if you want to work outside Everflow.
2. Sort by clicks and ignore rows with too little data to judge. As a rule of thumb, give a channel 200 to 500 clicks before you kill it.
3. For rows with enough data, compare earnings per click to your average.
4. Cut the bottom rows that have had a fair test.
5. Investigate rows with many clicks and zero conversions. Zero is often a tracking problem, not a traffic problem.
6. Scale the top rows.

Decide your thresholds before you look at the data. Deciding after leads to keeping favorites.

A weekly optimization routine

Set aside the same slot every week.

Monday: review

• Pull last week's Flex Report by country, device and your main sub-ID.
• Use the Variance Report (Analytics → Variance) to compare click-to-conversion and earnings per click against the prior week.
• Flag anything that moved sharply in either direction.

Monday: check tracking

• Click through each live link from a clean browser and confirm the click registers in Everflow with sub-IDs intact.
• In the Conversion Report (Reporting → Conversion), check that recent conversions carry a real sub1 value.
• If postbacks are set up, check that conversions appear on the traffic source side.

Tuesday: prune and scale

• Cut placements, creatives or accounts that failed their test.
• Shift budget or posting effort to the top performers.

Wednesday: launch one test

• One new pre-lander, landing page, creative angle or geo. Only one major variable per test.

Friday: talk to your manager

• Send a short note: what you tested, what you saw, what you plan next.
• Ask about new creatives, new landers, or a commission model change if your traffic profile has changed.

When to change commission model

Your data can tell you which model suits you. If your users keep buying after the first purchase, the 40% lifetime RevShare grows with them. If your traffic converts once and churns, CPA (up to $70 per qualifying paying user, geo-tiered) may pay more. A hybrid is available on request. Review this every few months with your manager. See /commissions.

Next steps

• Fix data gaps first with Affiliate clicks not tracking.
• Scale what works with Scaling paid traffic for AI companion offers.
• Common questions are in /help.
• Ready to start: apply to the Secrets AI affiliate program.