Social Listening

What to Actually Listen For

Four distinct things, each serving a different decision:

Brand mentions — direct references to you, including misspellings and unlinked mentions. Drives response and reputation management.

Category conversation — people discussing the problem you solve without naming any brand. The richest source of positioning and content insight.

Competitor mentions — what people praise and complain about elsewhere. Complaints about competitors are a direct map of unmet demand.

Intent signals — people actively asking for recommendations in your category. The closest thing to a lead that listening produces.

Setting Up Queries That Work

Poorly constructed queries are why most listening tools get abandoned. A brand name that is also a common word returns thousands of irrelevant results and buries the real ones.

Build queries with Boolean logic: the brand term plus category qualifiers, minus known false positives. Test and refine over several days before trusting the output.

Include common misspellings, the brand name without spaces, and product names separately from the company name.

Where Listening Tools Fall Short

Coverage is uneven. Most tools cover X, public Instagram and Facebook, Reddit, forums, news and blogs reasonably. They cover private groups, WhatsApp, Discord, Telegram and closed communities barely or not at all — and for many categories that is where the real conversation happens.

Sentiment analysis remains unreliable, particularly with sarcasm, mixed sentiment and non-English content. Use it to prioritise what to read, not as a metric to report.

Turning Listening Into Action

The failure mode is a dashboard reviewed monthly and acted on never.

What makes it useful: routing mentions requiring response directly to the person who responds, a standing item in product or marketing meetings covering what the category is complaining about, and a small number of tracked themes rather than a volume metric.

One genuinely useful output is a recurring list of questions people ask about your category, which feeds content, sales enablement and product documentation simultaneously.

Building Queries That Do Not Drown You

A listening setup fails in one of two directions: it returns everything, so nobody reads it, or it returns nothing, so it seems unnecessary. Both are query problems.

Start from the question, not the brand name. Are people complaining about delivery times? produces a usable query. Monitor our brand produces a firehose.

Handle ambiguity deliberately. A brand name that is also a common word needs context terms and exclusions, and building those is most of the setup work. Test the query and read the first hundred results before trusting it — the noise is obvious on inspection and invisible in a volume chart.

Account for how people actually write. Misspellings, abbreviations, transliteration and local-language variants. In Indian markets in particular, a query in English alone will miss a large share of what is being said.

And separate the streams. Complaints, competitor mentions, feature requests and general chatter want different routing and different response times. One combined feed gets ignored as a single undifferentiated volume.

The Blind Spots Every Tool Shares

Listening tools see public posts on platforms that permit access, which is a smaller slice of the conversation than dashboards imply.

Private and semi-private spaces are invisible — messaging groups, closed communities, direct messages. For many Indian consumer categories this is where most word of mouth actually happens, which means the tool is measuring the visible minority.

Platform access changes without notice, and historical coverage varies by platform, so a trend line can move because access changed rather than because sentiment did.

Automated sentiment is weakest exactly where it matters: sarcasm, code-switching between languages, and category-specific terms where a normally negative word is neutral. Spot-check it against a sample you read yourself before quoting a sentiment percentage to anyone.

Used honestly, listening is an early-warning and qualitative-insight tool — it tells you what is being said and in what terms. Treated as a measurement system, producing sentiment scores and share-of-voice percentages presented to two decimal places, it gives false precision about a biased sample.