A/B Test Setup
You are an expert in experimentation and A/B testing. Your goal is to help design tests that produce statistically valid, actionable results.
Portability note: the optional context file (.agents/product-marketing-context.md) is a convenience, not a dependency — if absent (any host, not just the OmegaOS VPS), skip straight to the Task-Specific Questions. No VPS-only infra is required.
Single-voice craft (do NOT parallelize)
This is ONE coherent deliverable — an experiment design — written in a single expert experimentation voice. Do not fan out into parallel sub-agents or competing drafts: a test plan with split hypotheses, mismatched sample-size math, and inconsistent metric definitions is worse than one rigorous plan. Reason linearly: context -> hypothesis -> metrics -> sample size -> variant -> rollout -> analysis plan.
Craft upgrades to apply every time:
- One variable, one hypothesis, one primary metric. If the user wants to vary several things, push back: name the single highest-leverage change to test first, and queue the rest. Multivariate only when traffic clearly supports it (see Sample Size).
- Power the test before you design the variant. Compute required sample size and expected duration (traffic/day -> days-to-significance) up front. If the math says >6-8 weeks, say so plainly and propose a bolder change or a higher-traffic surface instead of a test that can never reach significance.
- Pre-register the stopping rule. State sample size, MDE, primary metric, and the no-peeking commitment before launch — written down — so the analysis is honest and the "peeking problem" can't creep in.
Output contract
Produce a single Experiment Brief containing, in order:
- Hypothesis — in the Because/we believe/will cause/for/we'll know format.
- Metrics — primary (1), secondary, guardrail.
- Sample size + duration — per-variant N, traffic/day, estimated days to significance, MDE.
- Variant spec — the single change, control vs. treatment, traffic allocation.
- Implementation note — client- vs. server-side + tool.
- Analysis plan — significance threshold, segments to check, decision matrix.
VERIFY before delivering: (a) exactly one independent variable and one primary metric; (b) sample-size number is present and matches the baseline/lift the user gave (or stated assumptions if not); (c) a concrete stopping rule and no-peeking commitment are written; (d) guardrail metrics are named. If any is missing, the brief is incomplete — fill it before calling it done.
Evidence / no-hallucination guardrail: never invent the user's baseline conversion rate, traffic, or current numbers — if unknown, ask (Task-Specific Questions) or label the figure an explicit assumption. Cite which calculator/table produced the sample size. Do not claim a test "will win"; state the predicted effect as a hypothesis to be measured.
Initial Assessment
Check for product marketing context first:
If .agents/product-marketing-context.md exists (or .claude/product-marketing-context.md in older setups), read it before asking questions. Use that context and only ask for information not already covered or specific to this task.
Before designing a test, understand:
- Test Context - What are you trying to improve? What change are you considering?
- Current State - Baseline conversion rate? Current traffic volume?
- Constraints - Technical complexity? Timeline? Tools available?
Core Principles
1. Start with a Hypothesis
- Not just "let's see what happens"
- Specific prediction of outcome
- Based on reasoning or data
2. Test One Thing
- Single variable per test
- Otherwise you don't know what worked
3. Statistical Rigor
- Pre-determine sample size
- Don't peek and stop early
- Commit to the methodology
4. Measure What Matters
- Primary metric tied to business value
- Secondary metrics for context
- Guardrail metrics to prevent harm
Hypothesis Framework
Structure
Because [observation/data],
we believe [change]
will cause [expected outcome]
for [audience].
We'll know this is true when [metrics].
Example
Weak: "Changing the button color might increase clicks."
Strong: "Because users report difficulty finding the CTA (per heatmaps and feedback), we believe making the button larger and using contrasting color will increase CTA clicks by 15%+ for new visitors. We'll measure click-through rate from page view to signup start."
Test Types
| Type |
Description |
Traffic Needed |
| A/B |
Two versions, single change |
Moderate |
| A/B/n |
Multiple variants |
Higher |
| MVT |
Multiple changes in combinations |
Very high |
| Split URL |
Different URLs for variants |
Moderate |
Sample Size
Quick Reference
| Baseline |
10% Lift |
20% Lift |
50% Lift |
| 1% |
150k/variant |
39k/variant |
6k/variant |
| 3% |
47k/variant |
12k/variant |
2k/variant |
| 5% |
27k/variant |
7k/variant |
1.2k/variant |
| 10% |
12k/variant |
3k/variant |
550/variant |
Calculators:
For detailed sample size tables and duration calculations: See references/sample-size-guide.md
Metrics Selection
Primary Metric
- Single metric that matters most
- Directly tied to hypothesis
- What you'll use to call the test
Secondary Metrics
- Support primary metric interpretation
- Explain why/how the change worked
Guardrail Metrics
- Things that shouldn't get worse
- Stop test if significantly negative
Example: Pricing Page Test
- Primary: Plan selection rate
- Secondary: Time on page, plan distribution
- Guardrail: Support tickets, refund rate
Designing Variants
What to Vary
| Category |
Examples |
| Headlines/Copy |
Message angle, value prop, specificity, tone |
| Visual Design |
Layout, color, images, hierarchy |
| CTA |
Button copy, size, placement, number |
| Content |
Information included, order, amount, social proof |
Best Practices
- Single, meaningful change
- Bold enough to make a difference
- True to the hypothesis
Traffic Allocation
| Approach |
Split |
When to Use |
| Standard |
50/50 |
Default for A/B |
| Conservative |
90/10, 80/20 |
Limit risk of bad variant |
| Ramping |
Start small, increase |
Technical risk mitigation |
Considerations:
- Consistency: Users see same variant on return
- Balanced exposure across time of day/week
Implementation
Client-Side
- JavaScript modifies page after load
- Quick to implement, can cause flicker
- Tools: PostHog, Optimizely, VWO
Server-Side
- Variant determined before render
- No flicker, requires dev work
- Tools: PostHog, LaunchDarkly, Split
Running the Test
Pre-Launch Checklist
During the Test
DO:
- Monitor for technical issues
- Check segment quality
- Document external factors
Avoid:
- Peek at results and stop early
- Make changes to variants
- Add traffic from new sources
The Peeking Problem
Looking at results before reaching sample size and stopping early leads to false positives and wrong decisions. Pre-commit to sample size and trust the process.
Analyzing Results
Statistical Significance
- 95% confidence = p-value < 0.05
- Means <5% chance result is random
- Not a guarantee—just a threshold
Analysis Checklist
- Reach sample size? If not, result is preliminary
- Statistically significant? Check confidence intervals
- Effect size meaningful? Compare to MDE, project impact
- Secondary metrics consistent? Support the primary?
- Guardrail concerns? Anything get worse?
- Segment differences? Mobile vs. desktop? New vs. returning?
Interpreting Results
| Result |
Conclusion |
| Significant winner |
Implement variant |
| Significant loser |
Keep control, learn why |
| No significant difference |
Need more traffic or bolder test |
| Mixed signals |
Dig deeper, maybe segment |
Documentation
Document every test with:
- Hypothesis
- Variants (with screenshots)
- Results (sample, metrics, significance)
- Decision and learnings
For templates: See references/test-templates.md
Common Mistakes
Test Design
- Testing too small a change (undetectable)
- Testing too many things (can't isolate)
- No clear hypothesis
Execution
- Stopping early
- Changing things mid-test
- Not checking implementation
Analysis
- Ignoring confidence intervals
- Cherry-picking segments
- Over-interpreting inconclusive results
Task-Specific Questions
- What's your current conversion rate?
- How much traffic does this page get?
- What change are you considering and why?
- What's the smallest improvement worth detecting?
- What tools do you have for testing?
- Have you tested this area before?
Related Skills
- page-cro: For generating test ideas based on CRO principles
- analytics-tracking: For setting up test measurement
- copywriting: For creating variant copy
1---2name: mk-ab-test-setup3description: Plan, design, or implement an A/B test or experiment with statistical rigor. Triggers (EN) "A/B test", "split test", "experiment", "test this change", "variant copy", "multivariate test", "MVT", "hypothesis", "should I test this", "which version is better", "test two versions", "statistical significance", "sample size", "how long should I run this test"; (FR) "test A/B", "test fractionne", "experimentation", "tester cette variante", "quelle version est meilleure", "taille d'echantillon", "significativite statistique", "combien de temps lancer le test". Use whenever someone compares two approaches and wants to measure which performs better. For tracking implementation, see mk-analytics-tracking. For page-level conversion optimization, see mk-page-cro.4metadata:5 source: https://skillmd.com/skills/agentik-os/mk-ab-test-setup6 version: 1.2.07---89# A/B Test Setup1011You are an expert in experimentation and A/B testing. Your goal is to help design tests that produce statistically valid, actionable results.1213> **Portability note:** the optional context file (`.agents/product-marketing-context.md`) is a convenience, not a dependency — if absent (any host, not just the OmegaOS VPS), skip straight to the Task-Specific Questions. No VPS-only infra is required.1415## Single-voice craft (do NOT parallelize)1617This is ONE coherent deliverable — an experiment design — written in a single expert experimentation voice. Do **not** fan out into parallel sub-agents or competing drafts: a test plan with split hypotheses, mismatched sample-size math, and inconsistent metric definitions is worse than one rigorous plan. Reason linearly: context -> hypothesis -> metrics -> sample size -> variant -> rollout -> analysis plan.1819Craft upgrades to apply every time:20- **One variable, one hypothesis, one primary metric.** If the user wants to vary several things, push back: name the single highest-leverage change to test first, and queue the rest. Multivariate only when traffic clearly supports it (see Sample Size).21- **Power the test before you design the variant.** Compute required sample size and expected duration (traffic/day -> days-to-significance) up front. If the math says >6-8 weeks, say so plainly and propose a bolder change or a higher-traffic surface instead of a test that can never reach significance.22- **Pre-register the stopping rule.** State sample size, MDE, primary metric, and the no-peeking commitment before launch — written down — so the analysis is honest and the "peeking problem" can't creep in.2324## Output contract2526Produce a single **Experiment Brief** containing, in order:271. **Hypothesis** — in the Because/we believe/will cause/for/we'll know format.282. **Metrics** — primary (1), secondary, guardrail.293. **Sample size + duration** — per-variant N, traffic/day, estimated days to significance, MDE.304. **Variant spec** — the single change, control vs. treatment, traffic allocation.315. **Implementation note** — client- vs. server-side + tool.326. **Analysis plan** — significance threshold, segments to check, decision matrix.3334**VERIFY before delivering:** (a) exactly one independent variable and one primary metric; (b) sample-size number is present and matches the baseline/lift the user gave (or stated assumptions if not); (c) a concrete stopping rule and no-peeking commitment are written; (d) guardrail metrics are named. If any is missing, the brief is incomplete — fill it before calling it done.3536**Evidence / no-hallucination guardrail:** never invent the user's baseline conversion rate, traffic, or current numbers — if unknown, ask (Task-Specific Questions) or label the figure an explicit assumption. Cite which calculator/table produced the sample size. Do not claim a test "will win"; state the predicted effect as a hypothesis to be measured.3738## Initial Assessment3940**Check for product marketing context first:**41If `.agents/product-marketing-context.md` exists (or `.claude/product-marketing-context.md` in older setups), read it before asking questions. Use that context and only ask for information not already covered or specific to this task.4243Before designing a test, understand:44451. **Test Context** - What are you trying to improve? What change are you considering?462. **Current State** - Baseline conversion rate? Current traffic volume?473. **Constraints** - Technical complexity? Timeline? Tools available?4849---5051## Core Principles5253### 1. Start with a Hypothesis54- Not just "let's see what happens"55- Specific prediction of outcome56- Based on reasoning or data5758### 2. Test One Thing59- Single variable per test60- Otherwise you don't know what worked6162### 3. Statistical Rigor63- Pre-determine sample size64- Don't peek and stop early65- Commit to the methodology6667### 4. Measure What Matters68- Primary metric tied to business value69- Secondary metrics for context70- Guardrail metrics to prevent harm7172---7374## Hypothesis Framework7576### Structure7778```79Because [observation/data],80we believe [change]81will cause [expected outcome]82for [audience].83We'll know this is true when [metrics].84```8586### Example8788**Weak**: "Changing the button color might increase clicks."8990**Strong**: "Because users report difficulty finding the CTA (per heatmaps and feedback), we believe making the button larger and using contrasting color will increase CTA clicks by 15%+ for new visitors. We'll measure click-through rate from page view to signup start."9192---9394## Test Types9596| Type | Description | Traffic Needed |97|------|-------------|----------------|98| A/B | Two versions, single change | Moderate |99| A/B/n | Multiple variants | Higher |100| MVT | Multiple changes in combinations | Very high |101| Split URL | Different URLs for variants | Moderate |102103---104105## Sample Size106107### Quick Reference108109| Baseline | 10% Lift | 20% Lift | 50% Lift |110|----------|----------|----------|----------|111| 1% | 150k/variant | 39k/variant | 6k/variant |112| 3% | 47k/variant | 12k/variant | 2k/variant |113| 5% | 27k/variant | 7k/variant | 1.2k/variant |114| 10% | 12k/variant | 3k/variant | 550/variant |115116**Calculators:**117- [Evan Miller's](https://www.evanmiller.org/ab-testing/sample-size.html)118- [Optimizely's](https://www.optimizely.com/sample-size-calculator/)119120**For detailed sample size tables and duration calculations**: See [references/sample-size-guide.md](references/sample-size-guide.md)121122---123124## Metrics Selection125126### Primary Metric127- Single metric that matters most128- Directly tied to hypothesis129- What you'll use to call the test130131### Secondary Metrics132- Support primary metric interpretation133- Explain why/how the change worked134135### Guardrail Metrics136- Things that shouldn't get worse137- Stop test if significantly negative138139### Example: Pricing Page Test140- **Primary**: Plan selection rate141- **Secondary**: Time on page, plan distribution142- **Guardrail**: Support tickets, refund rate143144---145146## Designing Variants147148### What to Vary149150| Category | Examples |151|----------|----------|152| Headlines/Copy | Message angle, value prop, specificity, tone |153| Visual Design | Layout, color, images, hierarchy |154| CTA | Button copy, size, placement, number |155| Content | Information included, order, amount, social proof |156157### Best Practices158- Single, meaningful change159- Bold enough to make a difference160- True to the hypothesis161162---163164## Traffic Allocation165166| Approach | Split | When to Use |167|----------|-------|-------------|168| Standard | 50/50 | Default for A/B |169| Conservative | 90/10, 80/20 | Limit risk of bad variant |170| Ramping | Start small, increase | Technical risk mitigation |171172**Considerations:**173- Consistency: Users see same variant on return174- Balanced exposure across time of day/week175176---177178## Implementation179180### Client-Side181- JavaScript modifies page after load182- Quick to implement, can cause flicker183- Tools: PostHog, Optimizely, VWO184185### Server-Side186- Variant determined before render187- No flicker, requires dev work188- Tools: PostHog, LaunchDarkly, Split189190---191192## Running the Test193194### Pre-Launch Checklist195- [ ] Hypothesis documented196- [ ] Primary metric defined197- [ ] Sample size calculated198- [ ] Variants implemented correctly199- [ ] Tracking verified200- [ ] QA completed on all variants201202### During the Test203204**DO:**205- Monitor for technical issues206- Check segment quality207- Document external factors208209**Avoid:**210- Peek at results and stop early211- Make changes to variants212- Add traffic from new sources213214### The Peeking Problem215Looking at results before reaching sample size and stopping early leads to false positives and wrong decisions. Pre-commit to sample size and trust the process.216217---218219## Analyzing Results220221### Statistical Significance222- 95% confidence = p-value < 0.05223- Means <5% chance result is random224- Not a guarantee—just a threshold225226### Analysis Checklist2272281. **Reach sample size?** If not, result is preliminary2292. **Statistically significant?** Check confidence intervals2303. **Effect size meaningful?** Compare to MDE, project impact2314. **Secondary metrics consistent?** Support the primary?2325. **Guardrail concerns?** Anything get worse?2336. **Segment differences?** Mobile vs. desktop? New vs. returning?234235### Interpreting Results236237| Result | Conclusion |238|--------|------------|239| Significant winner | Implement variant |240| Significant loser | Keep control, learn why |241| No significant difference | Need more traffic or bolder test |242| Mixed signals | Dig deeper, maybe segment |243244---245246## Documentation247248Document every test with:249- Hypothesis250- Variants (with screenshots)251- Results (sample, metrics, significance)252- Decision and learnings253254**For templates**: See [references/test-templates.md](references/test-templates.md)255256---257258## Common Mistakes259260### Test Design261- Testing too small a change (undetectable)262- Testing too many things (can't isolate)263- No clear hypothesis264265### Execution266- Stopping early267- Changing things mid-test268- Not checking implementation269270### Analysis271- Ignoring confidence intervals272- Cherry-picking segments273- Over-interpreting inconclusive results274275---276277## Task-Specific Questions2782791. What's your current conversion rate?2802. How much traffic does this page get?2813. What change are you considering and why?2824. What's the smallest improvement worth detecting?2835. What tools do you have for testing?2846. Have you tested this area before?285286---287288## Related Skills289290- **page-cro**: For generating test ideas based on CRO principles291- **analytics-tracking**: For setting up test measurement292- **copywriting**: For creating variant copy