Concepts
Campaigns
Using campaigns for A/B testing and offer optimization
What are Campaigns?
A campaign is a collection of related offers for A/B testing. When multiple offers share the same campaign ID, Stacked automatically distributes users across them for performance comparison.
How Distribution Works
When a user qualifies for a campaign, Stacked randomly assigns them to one variation:
- Permanent - Users always see the same variation
- Balanced - Equal distribution across all variations
- Consistent - Users won't switch between variations
Important
Users only see one offer from a campaign, even if they qualify for multiple. This ensures clean test results.
What to Test
| Test Type | Variables to Change | Example |
|---|---|---|
| Reward Amount | Coin values, item quantities | 100 coins vs 200 coins vs 150 coins + item |
| Messaging | Titles, descriptions, CTAs | "Conquer 10 Battles!" vs "Weekend Warrior" vs "Limited Time Event" |
| Timing | Expiry duration, availability | 1-hour vs 4-hour vs 24-hour expiry |
| Difficulty | Completion requirements | Spend 100 gold vs 500 gold vs 1000 gold |
| Targeting | Surfacing conditions | Level 10+ vs VIP only vs all users |
Best Practices
- Start small - 2-3 variations for simpler analysis
- Change one variable - Test one element at a time for clear results
- Set success metrics - Define completion rate, revenue, retention goals upfront
- Run long enough - At least one week or 100+ participants per variation
Tip
Only change one element between variations. Testing both reward and difficulty simultastackedusly makes it unclear which drove results.
Key Metrics
| Metric | What to Track | What It Tells You |
|---|---|---|
| Conversion Rate | % of users who complete | Offer appeal and difficulty balance |
| Time to Complete | How quickly users finish | Motivation strength and task difficulty |
| Revenue Impact | Change in spending | Direct and indirect monetization effect |
| Retention Effect | Next-day and 7-day return rates | Long-term engagement impact |
After Testing
- Winner Takes All - Deploy best performer to everyone
- Segmented Approach - Use different variations for different user groups
- Iterate Further - Refine the winning approach with follow-up campaigns
- Combine Elements - Merge successful aspects from multiple variations
Stacked