1. Choose an offer only at a high rate
CPA and RevShare cannot be compared by a single digit. You need to understand confirmation, NGR, hold, cap and your own expenses. A high rate with poor quality or strict criteria can give the worst result.
2. Mix all sites in one subID
If one placement degrades the quality, you won’t be able to find it quickly. The markup should stop a particular segment, not the entire source.
3. Optimize for CTR
The CTR only answers the question of whether the message is clicked on. The economy starts further. For paid traffic, cost per confirmed result and what happens to quality after scaling are more important.
4. Change several variables at the same time
New creativity, landing, rateting and GEO create beautiful activity in one day, but kill causality. If the result changes, the team will not understand why.
5. Do not check postback before launch
Lost events turn a normal campaign into an “unrecoverable” one on paper. Test transition and click ID reconciliation are cheaper than a few days of incorrect statistics.
6. Scaling up a small sample
The first successful events may be accidental. Before volume increases, at least a basic period of stability and predetermined stopping rules are needed.
7. Compare cohorts of different ages
The month-long RevShare of an old audience cannot be honestly compared to seven days of a new one. Use the same observation window.
8. Ignore GEO and device
The average figure hides the problem. A single market or browser can pull down the entire campaign. The minimum cut is source × GEO × device.
9. Publishing Unconfirmed Terms
An old bonus, an irrelevant rate, or an incorrect payment method creates a discrepancy rateween the ad and the product. This is both a content quality and conversion issue.
10. Don't fix solutions
The team repeats the same tests unless it records the hypothesis, the period, the result, and the next step. A simple table of experiments saves more time than a new dashboard.
Rapid diagnostic table
| Symptom | Where to look first |
|---|---|
| There are clicks, there are few registrations | Creativity ⁇ Landing, Mobile, Speed |
| raw FTD a lot, confirmed little | KPI, quality, SubID |
| tracker and partner diverge | timezone, click ID, postback |
| After the budget growth, the economy is worse | placement mix, frequency, cap |
| SEO drops in just one section | template, intent, thin content |
Conclusion
Most problems become easier when the data is separated and each change is linked to a hypothesis. Use it to launch. 20-point checklist and partner's monthly report.
