← Other blogs

Social media listening for research: a sample, not a census

Public posts are evidence with a known bias. The method is to say which bias, and keep the claim inside it.

· 3 min read

Social media listening for research means treating public posts as evidence you did not commission: real words from real people, gathered where they already talk. The catch sits in the second half. Nobody answered your question, and nobody chose the sample for you.

What you are actually collecting

Public posts, thread text, review pages and video comments. Not answers. A person who posts about a bad checkout did not sit down to tell you about checkout. They wanted to complain, warn a friend, or get help.

That is still useful, as long as the write-up says what it is.

  1. Who would have to post for this claim to hold?
  2. Where do those people already talk in public?
  3. What would I expect to see if I were wrong?

The third question is the one that keeps a study honest. Write the answer down before you look.

Where the sample leans

What you notice What it often means What stays hidden
A cluster of complaints in one week A change reached a lot of people at once The people who left without a word
Long, detailed praise An enthusiast wrote it The ordinary buyer with no strong view
The same question asked again The documentation is thin Whether anyone found the answer elsewhere

None of those rows gives you a percentage. They give you a lead.

A small workflow that holds up

  • Pick six to ten public sources: a subreddit, a review page, a trade blog, a competitor's pricing page, two or three accounts that set the tone.
  • Write the condition you care about as a sentence, not a keyword list.
  • Check weekly, not hourly. Research questions rarely turn on a same-day alert.
  • Log the quote, the link and the date. A claim without a link is an anecdote with better formatting.

AyeWatch does the checking part. It watches public pages, topics and public profiles on X, Reddit, YouTube, LinkedIn, Instagram and TikTok, and sends a summary when something that matters changes. The feed and push come with every plan. Email, Slack, Discord, webhooks and the API need a paid one, from $9 a month in monitoring credits, with a 7 day trial on the web.

It reads public profiles, so you need no account on the platform. It does not export a dataset and it does not produce sentiment scores.

Keep the claim small

Write "players in several threads described the timer as unfair" instead of "players find the timer unfair". The first sentence can be checked. The second one needs a sample size, and a monitor cannot give you one.

If the brief asks for share of voice or sentiment across a whole platform, a listening suite that licenses platform data is the right purchase. A web monitor answers a different question: what did this page, this topic or this account say, and when.

FAQ

Is social media listening for research reliable?

It is reliable as a source of language, complaints and questions. It is not a representative sample, so treat counts with care and never turn a thread into a percentage.

Do I need an account on each platform?

Not with AyeWatch. It reads public profiles on the platforms it covers, and the monitor runs without logging in to them.

How much does a monitor cost?

AyeWatch has a 7 day trial on the web, then paid plans from $9 a month, billed in monitoring credits. Reading the public sources it watches costs nothing.