Home / Performance
Tracking, analytics & AI-assisted optimization
How we evaluate campaign performance.
Good decisions need two things: the right metric for the offer, and tracking that reports it accurately. This page explains how we approach both, and where AI-assisted analysis fits in.
01 / Cost & return metrics
Choose the metric that matches the payout.
We agree on a primary metric at the start of every project, based on how the offer pays and what you are optimizing for.
- CPACost per acquisition
-
Ad spend ÷ conversions
The cost of each conversion, whether that conversion is a lead, a sale, an install, or another payable event.
How we use it: We compare CPA to the payout for the offer and break it down by placement, device, creative, and audience to find where cost is concentrated.
- CPLCost per lead
-
Ad spend ÷ leads
The cost of each lead submission. On lead-gen offers, the payout usually depends on lead quality, not just volume.
How we use it: We look at CPL alongside acceptance or qualification rates where the network provides them, because a low CPL on rejected leads is not a good result.
- CPSCost per sale
-
Ad spend ÷ sales
The cost of each completed purchase on offers that pay on a sale rather than a lead or click.
How we use it: We check that the sale event is reported back reliably, since purchases often happen on a merchant domain outside your direct control.
- tCPATarget cost per acquisition
-
Platform bid strategy
A bidding setting where the ad platform attempts to deliver conversions at or near a target cost you specify.
How we use it: tCPA only works as well as the conversion signal behind it. We verify the optimization event, volume, and consistency before relying on automated bidding.
- ROASReturn on ad spend
-
Revenue ÷ ad spend
How much revenue each unit of spend returns. For affiliates, revenue is usually commission or payout, not the merchant's order value.
How we use it: We make sure the value passed back to the platform matches what you are actually paid, so ROAS reporting and value-based bidding stay grounded.
- CVRConversion rate
-
Conversions ÷ clicks
The share of paid clicks that end in a conversion across the whole funnel.
How we use it: We track CVR over time and by segment, because a falling rate often signals creative fatigue, audience drift, or a tracking break.
- LP CVRLanding page conversion rate
-
Offer clicks ÷ landing page views
The share of landing page visitors who continue to the offer or complete the page's action. It isolates the page from the rest of the funnel.
How we use it: This is the clearest measure of whether the page itself is doing its job, and the primary metric for most page tests.
02 / Tracking & attribution
Following a conversion from click to postback.
Each step in this path is a place where data can be lost, duplicated, or misattributed. A tracking audit tests every hand-off.
Pixels
Browser-side tags such as the Meta Pixel and native platform pixels. We check placement, firing conditions, and whether they fire once per intended action.
Conversion events
The specific actions reported back, such as Lead, Purchase, or a custom event. We confirm the platform is optimizing to the event you are paid on, and that server and browser events are deduplicated.
Postbacks
Server-to-server calls from the affiliate network to your tracker, and from the tracker to the ad platform. We test that click IDs survive every redirect and that postbacks fire with the right values.
Attribution
Platforms, trackers, and networks use different windows and models, so their numbers rarely match exactly. We document the expected gap and investigate when it grows beyond it.
03 / Funnel drop-off
Find the step where visitors leave.
A high CPA is a symptom. Funnel drop-off analysis shows where the problem actually sits.
We measure how many visitors make it from each stage to the next: from ad click to page load, from page load to engaged reading, from reading to the offer click, and from the offer to the conversion. A sharp drop at one step tells us where to focus, whether that is page speed, the opening message, the call-to-action, a broken hand-off, or friction on the offer itself.
04 / AI-assisted analysis
More of the data, reviewed more often.
AI-assisted analysis may be used to identify trends, anomalies, testing opportunities, and potential areas for improvement. A person reviews every finding before it becomes a recommendation.
Trends
Gradual shifts in CPA, CVR, or ROAS that are easy to miss day to day, such as creative fatigue or audience saturation.
Anomalies
Sudden spikes or drops that may point to a tracking break, a policy issue, or a change on the offer side.
Testing opportunities
Segments, placements, or page sections where the data suggests a test is worth running next.
Areas for improvement
Weak funnel steps and underperforming page elements, prioritized for human review.
What AI does not do: it does not manage campaigns, move budgets, or publish changes on its own, and it does not guarantee better results. It helps us analyze; people decide.
05 / Creative testing
Tests designed to be read.
A test is only useful if its result can be trusted. We keep tests simple, isolated, and tied to one metric.
- Form a hypothesisStart from something observed in the data, such as a drop-off point or a weak segment, not a guess.
- Change one variableHeadline, opening, layout, CTA, or ad angle, one at a time, so the result points to a cause.
- Pick the metric in advanceUsually landing page CVR for page tests and CPA or ROAS for creative and funnel tests.
- Let it collect enough dataAvoid calling winners early on small samples or on a single day of traffic.
- Record the outcomeKeep a simple log of what was tested and what happened, so learnings carry forward across offers.
Next step
Want a second look at your numbers?
Share your current CPA, CPL, CPS, tCPA, or ROAS and how tracking is set up. We will tell you what we would check first.