Skip to content

Choose whether the unit is a post, thread or claim

Separate the object you count from the evidence you read, so one long discussion cannot silently become many independent observations.

Guides

What should count as one observation in a discussion study?

A thread can contain a product question, several replies and multiple claims. Counting each sentence as another respondent inflates apparent agreement. Before reviewing public discussions about repairable headphones, choose a primary unit that matches the decision: a conversation for support design, or a documented claim for a claim audit.

Example research prompt
Review the supplied public headphone threads as conversations. Keep replies nested, identify repair questions, and flag missing parents without estimating unique people.
01 / Guides

Map the nested objects

Draw a simple hierarchy from account to thread, reply and claim. Record which identifiers are actually available. Include replies attached to a visible parent when context can be inspected; exclude detached quotations whose speaker or target cannot be established. An absent parent is a context limitation, not another independent topic.

02 / Guides

Keep one counting rule

For a support study, give every eligible thread one row and attach reply evidence beneath it. A thread mentioning both battery life and replacement pads may receive two topic codes but remains one thread. Preserve this distinction in table headings so category totals are not mistaken for a respondent count.

03 / Guides

Test a difficult example

Review a thread containing repeated replies from its author, quoted advice and an unrelated joke. Ask two reviewers to apply the chosen unit independently. If they count different objects, revise the definition before reviewing the rest. Do not resolve disagreement merely by choosing whichever count makes the pattern stronger.

04 / Guides

Deliver linked views

Hand over a thread table and a separate claim index connected by identifiers. State the number of eligible units actually reviewed and which units lack usable context. If the decision requires people rather than conversations, stop and redesign the study; thread-level evidence cannot establish how many distinct people hold a view.

Keep the boundaries clear.

  • One conversation may support several topics without becoming several independent observations.
  • Platform badges indicate research context, not guaranteed source access. Unavailable evidence remains unavailable.

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