Understanding Distribution Strategies in Scorecard Routers (Manual QA)

When configuring a Scorecard Router (Manual QA) in Oversai, you can choose how interactions are distributed among your reviewers. This is done through distribution strategies.

Distribution strategies determine how interactions are assigned once the router has selected them using sampling.

This article explains the three available strategies and when to use each one.

🧠 What Are Distribution Strategies?

After the router selects interactions based on your sampling configuration, Oversai needs to decide how those interactions are assigned to reviewers.

Distribution strategies define the logic used to distribute those interactions among your reviewers.

Oversai currently supports three strategies:

• Round-Robin (recommended for most QA workflows)

• Random

• Sequential

Each strategy changes how interactions are assigned, but does not affect how sampling works.

Sampling determines how many interactions are selected per agent, while distribution determines who reviews them.

🔁 Round-Robin Distribution (Recommended)

Round-Robin distributes interactions evenly by rotating assignments between reviewers.

The router cycles through reviewers one by one until all interactions are assigned.

This ensures that workload is balanced and reviewers see a mix of interactions.

Example

Reviewers:

• Ana

• Ben

• Chen

Total interactions selected: 12

Assignment order:

1 → Ana
2 → Ben
3 → Chen
4 → Ana
5 → Ben
6 → Chen
7 → Ana
8 → Ben
9 → Chen
…until all interactions are assigned.

When to use Round-Robin

Round-Robin works best when you want:

• Fair workload distribution

• Equal exposure to different agents

• Balanced QA workloads

• Reduced reviewer bias

For most QA teams, Round-Robin is the recommended default strategy.

🎲 Random Distribution

Random distribution shuffles interactions before assigning them to reviewers.

The total number of assignments per reviewer still respects the configured allocation percentages, but which interactions each reviewer receives is randomized.

Example

Total interactions: 12

Reviewers:

• Ana

• Ben

• Chen

Instead of assigning interactions in order, the router randomly shuffles them before assigning.

Example result for Ana:

Interaction #7, #2, #11, #4, #9, #1

Each reviewer still receives their correct number of interactions, but the selection is randomized.

When to use Random

Random distribution is useful for:

• Compliance audits

• Statistical sampling

• Removing any potential selection bias

• Regulatory or audit scenarios

This approach ensures interactions are assigned unpredictably.

📑 Sequential Distribution

Sequential distribution assigns interactions in continuous blocks to each reviewer.

Instead of rotating or randomizing assignments, the router distributes interactions in order.

Example

Total interactions: 12

Reviewers:

• Ana

• Ben

• Chen

Distribution:

Each reviewer receives a continuous block of interactions.

When to use Sequential

Sequential distribution is useful when:

• You want to keep as many interactions from the same agent assigned to the same grader as possible, enabling more consistent evaluations and better context when reviewing performance

• Interactions are already sorted meaningfully (for example, by time)

• Reviewers focus on specific time segments

• Teams want reviewers to analyze similar interactions together

📊 Quick Comparison

📝 In Summary

Distribution strategies determine how interactions are assigned to reviewers after sampling is completed.

Oversai provides three options:

• Round-Robin → Balanced workload distribution

• Random → Unbiased statistical assignment

• Sequential → Ordered block assignment

If you're unsure which strategy to choose, Round-Robin is the best starting point for most QA programs.