Most social listening for sales is set up to fail on day one, because it monitors the wrong vocabulary.
A watch on your category name (“sales engagement platform”, “cold email tool”) returns vendors, consultants, job posts and people writing commentary about the market. Almost none of them are buyers. Buyers do not use your category name, because they are describing a problem rather than shopping for a solution.
The fix is to watch the language people actually use when they have the problem.
The four kinds of mention, ranked
Not all mentions are equal, and the gap between the best and worst is large enough that treating them the same wastes the whole exercise.
1. Recommendation requests, the strongest signal in B2B social
“Anyone know a good tool for X?” “Looking for recommendations on Y, what are people using?”
This is someone publicly asking their network to help them buy. There is no stronger declared intent available anywhere, including paid intent data, because the person has self-identified, stated the requirement and invited responses.
The window is 24 to 72 hours. After that they have shortlisted, usually from the first few replies, and a late answer arrives after the decision.
This is the one to build the watch for. Everything else on this list is secondary.
2. Problem statements
“Our deliverability has fallen off a cliff and I have no idea why.” “Spending four hours a week stitching reports together.”
Someone describing the symptom, not looking for a vendor. Weaker intent, far larger volume, and a much longer window, weeks rather than days, because the problem persists.
The right response here is rarely a pitch. It is being useful, publicly, and letting that produce the conversation.
3. Competitor mentions
Someone naming a competitor, with a sentiment attached. Complaints, price objections, migration questions, “has anyone moved off X”.
Valuable and requiring care. A complaint about a competitor is not a complaint about the category, and swooping into a frustration thread with your product reads badly to everyone watching, including the people who were not complaining.
4. Category discussion
The vocabulary most people monitor and the weakest band. Industry commentary, vendors marketing, analysts analysing.
Useful for understanding the conversation. Not a prospecting list.
Writing the watches
The practical part. Keyword and topic watches are only as good as the phrasing you give them.
Write problem language, not product language.
Instead of monitoring your category, monitor what the problem sounds like:
| Don’t watch | Watch |
|---|---|
| “cold email platform” | “emails going to spam”, “deliverability dropped”, “domain blacklisted” |
| “sales engagement” | “reps spending hours on”, “no visibility into pipeline” |
| “data provider” | “bounce rate”, “bad data”, “contacts out of date” |
Build a dedicated recommendation-request watch. This is worth constructing carefully, since it is the highest-value band. The phrasings that recur:
- “anyone know a good”
- “looking for recommendations”
- “what are people using for”
- “alternatives to”
- “has anyone tried”
- “any suggestions for”
Combine each with your domain’s nouns. The combination is what makes it precise, “looking for recommendations” alone returns restaurant advice.
Watch competitor names with sentiment words. Competitor plus “alternative”, “switching”, “frustrated”, “cancelled”, “migrating”.
Then prune for a fortnight. Every watch returns noise initially. Spend ten minutes a day for two weeks removing phrasings that return vendors, recruiters and commentary. A watch nobody pruned becomes a feed nobody reads, which is how most social listening quietly dies.
How to set up keyword and topic watches covers the mechanics in more depth.
How to respond
The part that decides whether this works, and the part most teams get wrong.
For recommendation requests: reply publicly, briefly, once
Answer in the thread. Be short. Be specific about fit, including where you are not a fit.
What works: naming what you do in one line, saying which situation it suits, and offering to answer questions. What does not: a paragraph of positioning, a link with no context, or claiming to be the answer to a question you only partly fit.
Declare that it is your product. Every time. The audience will find out in one click and an undisclosed recommendation of your own tool is the fastest way to lose a thread, and the people reading are often more valuable than the person who asked.
Then stop. One reply. Arguing with other vendors in the thread is visible to everyone and makes all of you look worse.
For problem statements: be useful first
Answer the question without mentioning your product. If someone describes a deliverability symptom, tell them what it probably is and how to check.
This is slower and it is the highest-return activity in this entire category, for a reason that is easy to miss: the person with the problem is not the only reader. A genuinely useful public answer is seen by everyone following the thread, and some of them have the same problem and have not posted about it.
Then follow up privately, days later, connecting the dots. The sequence (public help, private follow-up) converts far better than either alone.
For competitor mentions: never pile on
The rule is simple. Do not criticise the competitor. Address the problem.
“That’s usually caused by X, here’s how to check” is appropriate whether or not they switch. “We don’t have that problem” is not, and it makes the reader defensive about a choice they already made.
Someone actively asking for migration advice is different, and even then the useful answer is the honest one: what transfers, what does not, and what the real cost of moving is. No platform in this category can import your sequences from another, so anyone promising a painless migration is misleading them.
What this does not replace
Three honest limits, because social listening is oversold.
Volume is low. Even in an active market, genuine recommendation requests in your niche might be a handful a week. This is a high-quality, low-volume channel and it cannot be your pipeline.
It skews to the publicly active. The people posting are a particular slice of your market, usually smaller companies, usually in tech-adjacent sectors. If you sell to manufacturing operations directors, your buyers are not posting.
It is not a replacement for buying signals tied to accounts. A recommendation request from a company outside your ICP is still outside your ICP. Score these leads against your criteria like any other, rather than treating the signal as sufficient.
Where it genuinely earns its place is as the highest-intent band in a stacked system, one signal among several, weighted heavily when it fires. A company already showing hiring activity in your domain, whose head of department just asked publicly for recommendations, is a different proposition from either alone.
Timing, in one table
| Mention type | Window | Why |
|---|---|---|
| Recommendation request | 24–72 hours | They shortlist from the first replies |
| Problem statement | 2–4 weeks | The problem persists |
| Competitor complaint | 1–2 weeks | Frustration fades or escalates |
| Category discussion | None | Not a timing signal |
The 24–72 hour window on the top band is why this needs to be a monitored feed rather than something checked weekly. A recommendation request found on Friday for a Tuesday post is a missed one.
Choosing keywords that work
This is where the play succeeds or fails, and most attempts fail here.
Too broad: “sales,” “marketing,” “AI.” You will drown in posts from people with nothing in common.
Too narrow: your product category name. Almost nobody writes “I need a cold email platform.” People write about problems, not categories.
About right: the problem in the words your buyers use.
Not “email deliverability software” but “emails going to spam.” Not “lead enrichment tool” but “our data is a mess.” Not “sales engagement platform” but “our reply rates have fallen off a cliff.”
Three types that work consistently:
Complaint phrasing. How someone describes the problem when frustrated. These are the highest-intent posts on the platform and they rarely use vendor vocabulary.
Transition language. “We’re moving off,” “evaluating alternatives to,” “finally replacing.” People announce switching decisions publicly more often than you would expect.
Milestone language. “Just hit 50 people,” “opening our US office,” “first SDR hire.” Growth moments that create the problem you solve.
Competitor names. Blunt but effective. People complaining about, praising or comparing a competitor are in-market by definition.
Setting a relevance floor
Two controls stop this becoming noise.
A reaction floor. Posts below a threshold are frequently talking to nobody. A floor filters out the long tail without losing much.
A daily post cap. Topic watches can match far more than expected on a busy day. A cap keeps a trending phrase from consuming a month’s allowance.
Preview expected volume before saving a watch. If a keyword looks like it will match 400 posts a day, it is too broad, narrow it before committing.
The short version
Four bands, very unequal. Recommendation requests are the strongest declared intent available anywhere in B2B, with a 24–72 hour window. Problem statements are weaker, more numerous and last weeks. Competitor mentions need care. Category discussion is not a prospecting list.
Watch problem language rather than category language, “emails going to spam” rather than “cold email platform”, and build a dedicated watch for recommendation phrasings combined with your domain’s nouns. Then prune for a fortnight or the feed dies.
Reply to recommendation requests publicly, briefly, once, and always disclose that it is your product. Answer problem statements usefully without pitching, then follow up privately days later. Never pile on a competitor complaint.
And keep it in proportion: low volume, skewed to the publicly active, and best used as the heaviest signal in a stack rather than a channel of its own.
Next: setting up keyword watches, or when two weak signals beat one strong one.
Frequently asked questions
What is the strongest buying signal on social media?
A public recommendation request, someone asking their network what tool or vendor to use. The person has self-identified, stated the requirement and invited replies, and the window is roughly 24 to 72 hours before they shortlist from the earliest responses.
How do you set up social listening for sales?
Watch problem language rather than category names, buyers describe symptoms, not product categories. Add a dedicated watch for recommendation phrasings like “anyone know a good” or “alternatives to” combined with your domain’s nouns, then prune noise daily for two weeks.
Should you pitch your product in a social thread?
On a direct recommendation request, yes, briefly, once, and always disclosing that it is your product, since the audience will discover it in one click. On a problem statement, no: answer usefully first and follow up privately days later.
Is social listening enough to build a pipeline?
No. Genuine recommendation requests in a given niche are a handful a week, and the people posting skew toward smaller, tech-adjacent companies. It works as the highest-intent band within a stacked signal system, not as a channel on its own.
How do you find prospects talking about a problem on LinkedIn?
Watch keywords and topics rather than companies, using the words buyers actually use (complaint phrasing, transition language, milestone announcements and competitor names) rather than your product category.
Are post authors or engagers the better signal?
Authors. Writing about a topic indicates active concern; reacting to a post is much weaker. Capture engagers as a secondary source when allowance permits.
How quickly should you act on a topic signal?
Within 48 hours. These decay faster than any other signal type, which is why topic watches check every few hours rather than daily.