Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Social Media Marketing › Social Media Management

Setting Up Social Listening

How to set up social listening: what to listen for, building and testing queries, what a tool cannot see, routing what it finds, and a monthly insight note.
Tutorialintermediate12 min readUpdated

Listening, monitoring and reporting

Three activities overlap, and keeping them apart is what stops listening turning into another dashboard nobody reads. Reporting looks at the brand's own accounts: what was posted, who saw it, what they did. The social media reporting guide covers it. Media monitoring watches earned coverage in news, broadcast, trade titles and social, and logs it for the PR team; the media monitoring and coverage reporting guide covers it, and names social listening as one of its sources. Listening is the conversation about the brand and its category beyond the brand's own accounts, read for what it means.

ActivityWhat it looks atThe question it answersOwned by
Social reportingThe brand's own posts and accountsHow did what we published perform against our goal?Social team
Media monitoringEarned coverage across news, broadcast, trade titles and socialWhat did the press and commentators say, and where?PR or communications
Response monitoringMentions and messages that need an answerWho needs a reply or a handover, now?Community manager
Social listeningThe wider conversation about the brand, competitors and categoryWhat are people saying, what is changing, and what should we do about it?Whoever owns insight, often social with research

In practice one set-up feeds all four. The difference lies in what happens to each finding: a complaint is routed to service, a journalist's post to PR, a spike to the crisis owner, and a pattern to the monthly note. Listening also feeds other programmes on the site. The voice of the customer guide treats social posts as one stream of unsolicited feedback; the competitive intelligence guide tracks competitors' presence; the audience research guide uses community conversation to find the questions content should answer. Listening hears only the people who post in the sources it covers, so when the question is how many category buyers think of the brand, the comparison of brand trackers and social listening explains when a survey answers it instead.

Before you start

  • A social reporting routine. Listening adds to the reporting the brand already does; the reporting guide is the prerequisite, because its habit of comparing against your own history applies here too.
  • The people who will act on findings. A named contact in customer service, in PR or communications, in product and the crisis owner, each of whom has agreed to receive what listening routes to them.
  • A list of brand, product and campaign names. Including old names, nicknames and the ways customers actually spell them.
  • A competitor list. The alternatives customers consider, taken from the competitor work in the competitor analysis guide, not only the rivals the team watches.
  • A way to search. A listening tool, or the platforms' own search and alerts for a smaller programme. The steps are the same either way.
  • A home for the notes. A shared sheet for queries, tests, exclusions, tags and the monthly note.

The steps

  1. Write the questions listening should answer

    Start with the questions, not the tool. Listening that begins by collecting every mention produces a volume chart and no decisions. Write no more than a handful of questions the business actually wants answered. What do people say about us and our products, and how is that changing? What do they say about our competitors, and where do they prefer them? What problems and wishes come up in our category that nobody serves? What is rising that we should know about early? Who outside our audience talks about us? Attach each question to the person who will use the answer, such as the product lead for wishes and complaints, or the content lead for questions people ask. A question with no owner is a candidate for deletion.

  2. List the terms for the brand, products, people and competitors

    For each question, list the words people actually use. The brand name and its common misspellings, abbreviations and nicknames; product names, old and new; campaign names and tags; the names of spokespeople and founders who speak publicly for the brand; competitor brand and product names; and the category phrases customers use, which are often not the phrases the marketing team uses. Read real posts and reviews to find them, as the audience research guide advises for content: the customers' own words beat the brand's vocabulary. Group the terms by question so that each later query has a clear purpose. Leave out the names of private individuals; the brand needs to know what is said, not to follow particular people.

  3. Build queries with exclusions

    Combine the terms into queries, one per question or theme. Most queries need three parts: the terms that must appear, terms that make a mention relevant when the brand name is a common word, and exclusions that remove unrelated meanings. A brand called after an everyday word, a place or a person will catch thousands of posts that have nothing to do with it unless the query requires a product or category term alongside it or excludes the other meanings. The PR monitoring guide gives the same advice for its queries: names and misspellings, product, spokesperson and campaign names, with exclusions for unrelated uses. Write each query in the shared sheet with its purpose, its owner and the date, so that changes can be tracked.

  4. Test them on a sample and tune

    Before trusting any query, read what it returns. Take a sample of results for each query and mark each one as relevant or not. Where many results are irrelevant, add exclusions or require a second term; where obvious mentions are missing, which you can check by searching for posts you know exist, widen the terms. Repeat until a sample reads as mostly relevant. Record the test, the sample, the changes made and the date, because a query that worked last quarter can drift when a word acquires a new meaning or a competitor launches something with a similar name. Volume counts from untested queries are noise, however good the chart looks.

  5. Note what the tool cannot see

    Every listening method has blind spots, and the honest set-up writes them down. Private messages, closed groups and communities, and any source the tool does not cover are outside its reach, and coverage differs between tools and platforms. Ask the vendor which sources it includes and how; if you are searching by hand, list the sources you check. Write the gaps next to the queries in the sheet and repeat them in the monthly note, so that a quiet month is read as quiet in the sources covered, not as silence. Where an important conversation happens in a space the method cannot see, such as a closed group for your category, decide whether it matters enough to take part openly under that community's rules, as the guide to sharing content in communities describes, rather than trying to watch it.

  6. Set alerts and route what they find

    Decide which findings need someone now and send them there. A customer with a service problem goes to customer service, through the same handover the community management guide sets out. A post by a journalist or a commentator goes to PR. A sudden rise in mentions about one subject, or any mention of safety, injury or personal data, goes to the crisis owner named in the social media crisis plan. A competitor launch or a product wish goes into the monthly note rather than an alert. Set the alert levels from your own normal range rather than any published figure, and adjust them after the first month: alerts that fire constantly are soon ignored, and alerts that never fire are not doing anything.

    FindingWhere it goesWhen
    Service problem from a customerCustomer service, through the community managerWithin the brand's own response standard
    Journalist, commentator or well-known accountPR or communicationsPromptly, under the PR owner's own rules
    Spike on one subject, or safety, injury or personal dataCrisis owner, under the crisis planAt once
    Product wish or recurring complaintProduct owner, via the monthly noteMonthly
    Competitor moveWhoever owns competitive intelligenceMonthly, or at once if material
    Content idea or customer questionContent leadMonthly

    Alerts that need an answer land with whoever is on the inbox, and the social media inbox triage SOP checks them against the crisis plan's triggers in the same shift routine that sorts comments and messages.

  7. Tag a sample by hand

    Automatic sentiment scores are a rough guide at best, and they misread sarcasm, slang, mixed posts and mentions of competitors. Each month, take a sample of relevant mentions and tag them by hand: the subject (product, service, price, delivery, a competitor, the category), the type (question, complaint, praise, wish, comparison) and the tone as a person reads it. The community management guide recommends the same: sentiment read by a person on a sample rather than only by a tool. Keep the tags short and stable so that months can be compared. Hand tagging is also where the most useful findings show up, because a person notices the question nobody is answering or the competitor that keeps being recommended.

  8. Write the monthly insight note

    The note is the output that matters. Write it in the what, so what, now what form: what the listening showed this month, against last month and the months before; so what it means for the questions and their owners; now what each owner will do. Read counts with the care the guide to reading small numbers advises: a jump from a handful of mentions to a few more is not a trend, and a single post that travelled can dominate a month. Where the note reports earned Share of Voice, define it as the share of voice guide does, against the category, and remember that volume is not sentiment. Break the conversation down by the Segmentation that matters to the owners, such as customers against prospects, or one product line against another. Keep it to a page, and list the gaps in coverage at the end. The note feeds the monthly social report rather than replacing it.

  9. Review the queries and the questions

    Every quarter, look at the whole set-up. Retest the queries on a fresh sample. Add new product and campaign names, remove retired ones and check the competitor list against what customers actually mention. Ask each question owner whether the answers changed anything, and drop questions that have stopped being useful. Check the data you keep: posts, names and handles collected from social profiles are personal data, and the competitive intelligence guide notes that collecting them brings data protection duties. Under the GDPR, cited in this page's sources, personal data is processed under a lawful basis and with principles such as data minimisation, which the guide to keeping personal data out of analytics summarises. Keep only what the questions need, aggregate where you can, and ask whoever handles data protection in your organisation what applies to you; nothing here is legal advice.

A coffee roaster's first quarter of listening (illustrative)

Illustrative example — not a real result

A coffee roaster's first quarter of listening

Brewvane Coffee Roasters, an invented roaster that sells beans online and to 40 cafés, sets up listening for its brand, its two best-selling blends and three competitors, with four questions: what customers say about the blends, what they say about delivery, what they say about the competitors, and what they ask for that nobody sells.

Customers often shorten the name to "Vane", so the first brand query includes it, and the first test returns 500 posts in a week, of which 31 are about the roaster: the rest use "vane" in its everyday senses. Requiring a coffee term alongside the short name and excluding weather and sailing terms brings the next sample to 46 relevant posts out of 52. The team writes down what it cannot see: the closed home brewing groups where many of its customers talk, and private messages.

Alerts route delivery complaints, 14 in the first month, to the customer service lead, and two posts by a food writer to the owner who handles press. The hand-tagged sample of 120 mentions shows praise for one blend and recurring complaints about a grinder setting printed on the bag. The first monthly note's most useful line comes from 9 separate posts asking whether the roaster sells a decaf version of its best-selling blend, which it does not. The so what: a product question with a clear audience. The now what: the product lead tests a small decaf batch and the content lead writes a post about the grinder setting. By the end of the quarter the delivery complaints have fallen to 5 a month after a courier change, and the decaf batch has a waiting list.

Verify it worked

  • Read a fresh sample from each query and mark each result relevant or not. Most should be relevant, and posts you know exist should appear.
  • Open the shared sheet. Every question has an owner. Every query has a purpose, an owner, a date and a test record, with the coverage gaps beside it.
  • Trace a real service complaint, a journalist's post and a safety mention through the alerts. Each reaches its named person.
  • Check the hand-tagged sample. Each mention has a subject, a type and a tone read by a person.
  • Read the monthly note. It follows what, so what, now what, gives each finding an owner and an action, and lists the gaps.

If it goes wrong

  • Results fill with unrelated posts. The brand name is an everyday word. Require a product or category term, add exclusions and retest the sample.
  • Obvious mentions are missing. The terms are too narrow. Add misspellings, nicknames and old product names, then search for posts you know exist.
  • Alerts fire constantly, or never. The levels were not set from your own normal range. Adjust them after the first month.
  • A quiet month reads as silence. The blind spots were not written down. List the sources covered and the gaps in the note.
  • Findings reach the note but nothing changes. They have no owner. Add a now what with a named owner, or drop the finding.

Next steps

Common mistakes

  • Starting with the tool. Collecting every mention before deciding what the questions are produces charts and no decisions.
  • Trusting untested queries. A brand name that is also an everyday word fills the results with posts about something else.
  • Reading a quiet month as silence. It is quiet only in the sources the method covers; write the gaps down.
  • Taking automatic sentiment at face value. Read a sample by hand and tag it.
  • Alerts with nowhere to go. A finding that is not routed to someone who acts on it is wasted, and alerts that fire constantly get ignored.
  • Treating small counts as trends. A handful of mentions more or less is not a change in the conversation.
  • Keeping more personal data than the questions need. Names and handles are personal data; keep only what the work needs.
  • A note that never says now what. Every finding in the monthly note ends with an owner and an action, or it is dropped.

Frequently asked questions

What is the difference between social listening and social monitoring?

The words are used loosely, but a useful split is this. Monitoring watches for mentions that need a response, such as a complaint that tags the brand or a question about an order, and routes them to the person who answers. Listening reads the wider conversation about the brand, its competitors and its category over time and asks what it means: what people want, what is rising, what they say about the alternatives. The same set-up usually does both; the difference is what happens to what it finds.

Can a listening tool see everything people post about a brand?

No. Private messages, closed groups and any source the tool does not cover are out of its reach, and coverage differs between tools and between platforms. Ask the vendor which sources it includes and how, and write the gaps down next to the queries, so that nobody reads a quiet month as proof that nobody is talking.

Do we need a paid tool to start social listening?

Not to start. The questions, the terms, the exclusions, the routing and the monthly note are the same whether the searching is done by a tool or by a person using the platforms' own search and alerts on a few sources. Starting by hand shows which questions matter and which terms work, which is the best brief for choosing a tool later. This page does not name or compare tools.

Terms used on this page

Frameworks

Where this page is used

Also related

Sources

  1. Regulation (EU) 2016/679 (GDPR) — EUR-Lex

Related