Experimentation Throughput Strategy

Use this skill to increase the rate of product experimentation without turning the experiment pipeline into an unreliable traffic jam. It focuses on testing capacity, isolated versus overlapping strategies, interaction effects, and the process/tooling needed to coordinate experiments at scale.

Source Traceability

Primary source: Next-Level A/B Testing by Leemay Nassery. Guidance is transformed and paraphrased from Chapter 1 on rate, quality, and cost; Chapter 2 on testing availability, isolated and overlapping strategies, and interaction effects; and Chapter 9 on balancing rate with quality, cost, and usability.

Related skills:

• ab-testing-platform-strategy for platform architecture and ownership.
• experiment-verification-monitoring for monitoring conflicts and active-test

health.

• experiment-sensitivity-optimization for reducing traffic needs by improving

metric sensitivity.

Reference Routing

Workflow

1. Map the current experiment pipeline and where tests wait.
2. Identify whether the bottleneck is traffic, coordination, tooling, review,

QA, analysis, or decision latency.

1. Decide whether isolated testing, overlapping testing, or a hybrid model fits

the product surface and metric precision needs.

1. Define interference and interaction-effect safeguards.
2. Add visibility: current tests, upcoming tests, capacity, ownership, and

conflict flags.

1. Write rollout rules so teams know when to schedule, overlap, defer, or split

experiments.

Output Format

# Experimentation Throughput Plan

## Bottleneck
[What is limiting experiment rate and what evidence shows it.]

## Recommended Strategy
[Isolated | Overlapping | Hybrid | Keep current strategy] because [reason].

## Capacity View
| Surface or Audience | Current Tests | Upcoming Tests | Constraint | Owner |
|---------------------|---------------|----------------|------------|-------|

## Interference Safeguards
- Conflict dimensions:
- Monitoring:
- Escalation:

## Process And Tooling Changes
1. [Change]
2. [Change]
3. [Change]

## Decision Rules
- Overlap when:
- Isolate when:
- Defer when:
- Revisit when:

Quality Bar

• Do not increase experiment count by ignoring validity risks.
• Do not default to isolated testing when capacity is the primary constraint and

experiments can run independently.

• Do not default to overlapping testing when experiments change the same user

journey, metric, or surface in ways that can interact.

• Make the coordination mechanism explicit; "teams will communicate" is not a

scalable throughput strategy.