Designing Commission Split Tests
Commission Testing Fundamentals
Commission structure is the single highest-leverage variable in your affiliate program. A 10% shift in commission model can change partner behavior, traffic quality, and long-term program economics. Yet most programs set commission rates once and only revisit them when forced to by competition or margin pressure.
Commission split tests compare two or more commission structures across matched groups of affiliates over a fixed period. The goal is not just which model drives more conversions -- it is which model produces profitable conversions that retain.
Setting Up a Commission Experiment
- Define the hypothesis: "Switching tier-2 iGaming affiliates from flat $150 CPA to 25% RevShare will increase 90-day affiliate LTV by 15%"
- Select the test variable: Only change ONE element (model type, rate, or tier threshold)
- Match control and test groups: Group affiliates by traffic volume, conversion rate, and tenure -- then split evenly
- Set the measurement window: Minimum 30 days for CPA tests, 60-90 days for RevShare (revenue needs time to accumulate)
- Define success metrics: Primary (revenue per affiliate) and secondary (conversion rate, traffic quality score, partner satisfaction)
Common Commission Experiments
| Test Type | Control | Variant | Measurement Window | Key Metric |
|---|---|---|---|---|
| CPA vs RevShare | Flat $200 CPA | 30% RevShare | 90 days | Revenue per affiliate per month |
| Tiered vs Flat | Flat $180 CPA | $150/$200/$250 by volume | 60 days | Conversion volume + quality |
| Hybrid test | 25% RevShare | $100 CPA + 15% RevShare | 60 days | Partner cash flow satisfaction + LTV |
| Payout threshold | $100 minimum | $50 minimum | 45 days | Affiliate activation rate |
| Accelerator | Flat rate | +20% bonus above 50 FTDs/month | 60 days | Top-tier volume lift |
Never test commission changes on your top 10% of affiliates without direct communication. High-performing partners expect stability and will interpret unannounced changes as a downgrade -- even if the variant is objectively better.
Matching and Randomization
Random assignment works when you have 100+ affiliates. Below that, use stratified matching. Rank affiliates by monthly revenue, split them into quartiles, then assign alternating affiliates from each quartile to control and test. This ensures both groups have similar distributions of high, medium, and low performers.
For Forex IB programs with multi-level hierarchies, test at the top-level IB and include all sub-IBs in the same group. Splitting a hierarchy across test groups contaminates your results because sub-IBs may shift volume between parent IBs.
Avoiding Contamination
- Communication isolation: Do not announce the test publicly -- brief each group separately on "their" deal structure
- Tracking separation: Use distinct deal IDs or campaign tags so reporting cleanly separates control from variant
- Time alignment: Start both groups on the same date to avoid seasonal bias
- Spillover check: Monitor whether test-group affiliates recruit new sub-affiliates at different rates
Use your platform's deal management system to create parallel commission plans. Assign affiliates to plans programmatically rather than manually editing individual deals -- this prevents errors and makes rollback instant.
Key Takeaways
- Commission is the highest-leverage variable -- test it systematically, not reactively
- Every experiment needs a specific hypothesis, matched groups, and a defined measurement window
- CPA tests need 30+ days; RevShare tests need 60-90 days for meaningful revenue data
- Use stratified matching when you have fewer than 100 affiliates to ensure balanced groups
- Never test silently on top performers -- communicate directly and frame changes as opportunities
Explore Further
Key terms and tools related to this lesson. Deepen your understanding and see how Track360 supports these concepts.
Affiliate Program
A structured partnership where a business rewards external partners (affiliates) for driving traffic, leads, or conversions through tracked referral activity.
Commission Split
A commission split is the division of earned commission between multiple parties, such as a master affiliate and their sub-affiliates, or a master IB and their sub-IBs.
Revenue Share
A commission model where affiliates receive a recurring percentage of the net revenue generated by referred users for the lifetime of those users or for a defined period.
CPA (Cost Per Acquisition)
CPA is a commission model where an affiliate earns a fixed payment for each qualifying action, such as a deposit, registration, or purchase, that a referred user completes.