Skip to content

Choose conversations as clusters before reading replies

Sample discussion containers deliberately and preserve within-thread dependence instead of treating every reply as an independent observation.

Guides

How should sampling account for replies nested in discussions?

A public repair discussion may contain dozens of replies responding to one unusual incident. Reading all of them can provide rich context, but it does not supply dozens of independent incidents. When a study samples threads first and replies second, the analysis should preserve that two-level design from selection through the final handoff.

Example research prompt
Review supplied repair threads as clusters. Keep reply evidence nested and distinguish patterns recurring across threads from repetition within a single incident.
01 / Guides

Define the cluster boundary

Treat one public discussion thread as a cluster with an identifiable parent and attached replies. Include replies tied to the selected parent; exclude unrelated cross-links unless separately selected under the study rule. Record whether the visible conversation is complete or partial so missing portions do not become invisible losses.

02 / Guides

Select clusters before replies

Choose threads using a declared topical or systematic rule rather than their apparent agreement with the expected finding. Then specify whether to inspect all available replies or a bounded subset. Keep thread selection and reply selection as separate decisions, because each creates a different limit on what the findings can describe.

03 / Guides

Inspect within-cluster concentration

Track whether many matching comments come from one incident, one author or one quoted claim. Report concentration at the thread level alongside reply examples. Do not allow a lively conversation to dominate a summary simply because it generated more visible replies than a quiet thread describing a different repair mechanism.

04 / Guides

Present both levels

Deliver a thread-level evidence table and a nested reply view, with exact reviewed counts for each. Explain which findings recur across threads and which exist only within one conversation. If only one cluster was available, call it a case study rather than implying broader agreement from its many internally dependent replies.

Keep the boundaries clear.

  • Reply volume does not establish the number of independent incidents.
  • Platform badges describe research context, not guaranteed access to evidence on that platform.

Start with a read-only check: ask your agent to report actual source coverage before collecting records or exporting an audience.

Primary references

Documentation reviewed 2026-11-13. Product capabilities, client availability and plans can change. Third-party marks identify their products, not an affiliation.

Know what social media shows about you.
Start with a question.

Get started
montage