Thousands of public posts are useless when the inclusion rule changes halfway through the analysis. Define what counts as a use case before collecting the sample.
This market research workflow produces a transparent report and a dataset of source URLs.
SocialListeningAPI searches public posts on Reddit, LinkedIn, X, Hacker News, TikTok, and YouTube for the phrases you choose and returns normalized records with source URLs.
Step 1: Define a concrete use case
Create a SocialListeningAPI account and claim the 100 free credits. Write an inclusion rule that requires a product or category, a task, and an observed or claimed outcome.
Search phrases might include:
using product name for customer researchbuilt workflow with product nameproduct name saved timehow I use category toolmoved from competitor to product name
Exclude generic promotions, copied announcements, and posts without a clear task.
Step 2: Collect the market research sample
for endpoint in \
reddit/search-posts \
linkedin/search-posts \
x/search-posts \
hackernews/search-posts \
tiktok/search-posts \
youtube/search-posts
do
curl --get "https://api.sociallisteningapi.com/api/v1/$endpoint" \
--data-urlencode 'query=using product name for customer research' \
--header 'x-api-key: YOUR_API_KEY'
doneOne phrase across six sources costs 7 credits. The API documentation explains the normalized response. Claude or ChatGPT can collect the same source records through the MCP.
Step 3: Build and deduplicate the dataset
Store source URL, platform, date, query, product, task, outcome wording, and inclusion status in an external database. Remove duplicate URLs and copies of the same post.
SocialListeningAPI returns public search records. Dataset storage and deduplication happen outside the API.
Do not automate Reddit replies from the dataset. A human must read the community rules and approve any response.
Step 4: Classify and check the sample
Use an external LLM to suggest use-case labels. Have a person check a sample from every label, correct weak groups, and record the error rate. A claimed outcome is evidence of what the author said, not independent proof that the outcome happened.
Do not call this comment analysis. The dataset contains returned posts and videos, not their full comment sections.
Step 5: Publish the method and limits
Report the sample size, query list, collection dates, platforms, inclusion rule, deduplication rule, classification method, human checks, and coverage limits. Publish the source URL dataset when you can do so responsibly.
Conclusion
Only call a use case proven when several independent sources describe the same concrete job. Keep rare examples in the dataset without presenting them as a common pattern.
Readers should be able to inspect the method, source list, and limits before trusting the findings.
Continue with Validate a Product Idea With Public Search Data.
