How to run a social listening review

A social listening review should help you identify a discussion worth answering, a recurring product complaint or a question your content has missed. To do that, preserve the posts behind the summary.

Vaaya can collect public social data through its catalog. Begin with a bounded review of a few sources, then decide whether deeper comment collection will answer a useful question.

Set the listening brief

Choose the brand or topic, aliases, language, platforms and date window. Write exclusions for ambiguous names. A common-word brand needs more context than a distinctive product name.

Connect an agent from Vaaya's installation page. Give it a budget and a request like this:

Review public discussion of [brand/topic] on [platforms] between [dates].
Use these aliases: [aliases]. Exclude [unrelated meanings].
Consult Vaaya for current sources and quotes. Budget: [amount].
Keep original post URLs, publication dates, retrieved times and native metrics.
Separate brand-owned posts from independent discussion. Summarize themes with
supporting examples and disagreements. Describe the sample's coverage gaps.
Do not reply, message users or change any accounts.

1. Discover the route for each platform

Vaaya's social-data reference uses vaaya/discover to find TikHub endpoints. Supply a query describing the platform and operation, such as a Reddit discussion search. Inspect the returned endpoint, action, price and required parameters.

For example, the current catalog lists /api/v1/reddit/app/fetch_dynamic_search through tikhub/fetch, with query required. The endpoint is itself a parameter on the Vaaya call. Do not reuse its inputs for another platform; another search may require keyword or search_query.

For bounded bulk collection, the catalog also includes apify/tiktok-posts with keywords and optional maxItems, plus apify/youtube-videos with searchQueries and optional maxItems. Pick the route that matches the source you need. Keep the first batch small enough to inspect.

2. Check relevance before expanding the sample

Read returned titles or post text and verify that they refer to the intended brand or topic. Identify brand-owned content, reshares, paid or disclosed promotional content where that status is available, and independent discussion. Keep unknown sponsorship status unknown.

Check publication dates against your window. Some endpoints may offer date controls; others require filtering returned rows. If the source cannot support the requested window, state that gap instead of relabeling old content as recent.

Deduplicate by platform post ID or canonical URL. Keep linked reshares together when they repeat the same discussion. A repeated post should not create several independent examples of the same complaint.

3. Read comments only where they help

Choose posts relevant to the brief before fetching comment pages. A high view count does not guarantee useful discussion. For a complaint review, comments on a precise product issue may be more informative than a broadly popular joke.

Discover the platform's comment endpoint and carry forward the returned post ID. Preserve its pagination token or cursor when continuing. Set a page limit and stop if pages repeat or the response no longer advances.

Separate post content from reader reactions. A video making a claim and a comment disputing it belong in different evidence rows. Paraphrase ordinary users when a short summary answers the business question; retain the original link for review.

4. Write a digest with evidence attached

Use this illustrative output schema; it is not a measured listening report.

Field	What the digest records
Theme	A specific question, complaint or use case
Supporting posts	Original URLs and short paraphrases
Source context	Platform, date and owned/independent status
Native metrics	Views, likes or comments, each kept separately
Disagreement	Evidence that complicates the theme
Suggested action	A support check, content idea or item to watch

Describe the number of rows actually reviewed when you run the task, along with the queries and platforms. Avoid converting a handful of posts into an estimate of market sentiment. Your sample can surface an issue without measuring how common it is.

Handle missing coverage explicitly

A successful request with no relevant posts is different from an endpoint error. Inspect the response and query before repeating a paid call. Record unavailable platforms and unread comment pages alongside the findings.

Keep the actual charges from Vaaya's call responses in the working file. For a recurring review, preserve the same query definitions and compare new evidence with previous rows. End the digest with the few source-linked items that your team can investigate or act on.

Questions

Does a social listening search measure all conversation about a brand?

No. It returns a sample shaped by the chosen platforms, queries, dates and available endpoints. Report that coverage rather than calling it the whole conversation.

Can I add views, likes and upvotes into one reach number?

Keep the original metrics separate. They measure different interactions and can include repeat exposure or overlapping audiences.

What does an empty search mean?

It means that request returned no matching rows. Check the query, dates, endpoint response and coverage before making a broader claim about discussion volume.