People asking for a recommendation are the highest intent you can find in public, and they use a small, repeatable set of phrasings. Search those phrases rather than your product category: the person who writes "can anyone recommend" has a budget, a deadline and no vendor yet.
Disclosure: Signl is ours. It runs keyword searches as standing sources, which is the automated version of what this article describes. Everything below works manually too, and the parts that do not work well are marked.
Key facts
- Buyers describe problems, not categories. A search for your category name mostly returns vendors talking to each other.
- The strongest requests carry a constraint: a stack, a budget, a headcount, a deadline. The constraint is the proof the search is real.
- Recommendation requests are rare. Expect a handful a month in most categories, not a feed.
- The authors of these posts are leads, and so are the people answering them in the comments. Most setups only collect one of the two.
- A public reply usually beats a direct message on this signal, because the person asked the room.
Why phrasing beats category
Search your category name and you get your competitors' marketing, your competitors' employees, and the consultants who write about the category. The buyer is in there, drowned.
The reason is simple: buyers do not know your category name until late. They know their problem, and the first thing they do about it is ask people they trust. That request uses the words of the problem, plus one of about twenty stock openers that have nothing to do with your industry.
Those stock openers are the search surface. They are also stable, because "can anyone recommend" is how English speakers have asked for referrals for decades.
The pattern library
Twenty-two patterns, in four groups by what they tell you. Combine each with two or three words from your own problem space, for example "close" and "reconciliation" for accounting software, or "deliverability" and "warmup" for email tooling.
Group 1: direct vendor requests, the strongest signal
| Pattern | What it returns | The false positive to expect |
|---|---|---|
| can anyone recommend | Open requests, often with constraints | Recruiters asking for candidate referrals |
| looking for recommendations | Same, slightly more considered | People asking for books, podcasts and courses |
| any recommendations for | Short, high volume, noisiest of the three | Restaurant and travel posts, if your words are generic |
| who do you use for | Asks about incumbents, so it reveals the field | Agencies asking peers which subcontractor to use |
| what are people using for | Category-level, earlier stage | Vendors fishing for market research |
| hiring or buying | Explicit build-versus-buy moment | Genuine job posts |
Group 2: switching and dissatisfaction, the most actionable
| Pattern | What it returns | The false positive to expect |
|---|---|---|
| alternatives to | People already shortlisting | Blog promotion by competitors |
| moving away from | Decided to leave, not decided where to | Posts about leaving a job |
| migrating off | Technical, usually with a deadline | Infrastructure posts with no purchase attached |
| cancelled our | Fresh dissatisfaction, strong emotion | Event and subscription cancellations |
| frustrated with | Pain without a purchase decision yet | Venting with no commercial intent at all |
| after [competitor] announced | Reaction to a competitor's price rise, shutdown or acquisition | Their own staff and investors |
The last row is worth setting up permanently against your two closest competitors. A price change or a shutdown produces more switching intent in a week than a quarter of outbound, and we built an entire article around one of these events in Trigify alternative.
Group 3: stated problems, earlier but higher volume
| Pattern | What it returns | The false positive to expect |
|---|---|---|
| still doing [task] manually | Clear, quantifiable pain | Consultants describing client work |
| spending too much on | Budget pressure, named line item | General economy commentary |
| takes us [N] days to | A measurable problem with a number attached | Humblebrag posts about speed |
| does anyone else find | Seeking validation, pre-purchase | Pure community chat |
| the problem with [category] | Opinionated, often from buyers who know the space | Thought-leadership posts by vendors |
| we tried [approach] and | A failed attempt, which is a qualified need | Retrospectives with nothing left to solve |
Group 4: build and hire signals, slower but larger
| Pattern | What it returns | The false positive to expect |
|---|---|---|
| we're building our own | Build-versus-buy in progress | Side projects |
| first [role] hire | A function being started from scratch | Congratulation posts about someone else |
| taking [function] in house | Agency churn, which may be your opening | Posts about hiring generally |
| starting from scratch on | Greenfield, no incumbent to displace | Career-change posts |
Reading a hit in fifteen seconds
Three questions, in this order. If the first fails, stop.
- Is the problem theirs? "We" and "our" mean buyer. "My clients" means agency. "I used to" means neither.
- Is there a constraint? A stack, a number, a date, a team size. No constraint usually means idle curiosity.
- Can you serve them? Country, size, industry. A perfect request from a company you cannot invoice is not a lead.
Signl applies the same three as an automated check against your ICP, and a known country outside your list caps the match whatever the post says. The ladder behind question two is in LinkedIn buying signals.
Turning a phrase into a standing source
A search you run by hand on a Tuesday finds the requests posted on a Tuesday. Requests posted on Wednesday are gone by the time you look, because the thread fills up and the asker has already picked someone.
So the useful version of this is continuous. In Signl a topic search is a source: it looks for new posts matching your words, collects the authors of those posts and the people commenting on them, qualifies each one, and finds a verified work email for the ones who fit. A search counts as 2 sources against the plan's 25, because a daily search plus the comments on everything it finds costs about twice what watching one profile does.
Two configuration notes from running these.
Keep the problem words specific and the pattern words generic. "can anyone recommend" plus "deliverability" works. "can anyone recommend" plus "software" returns the internet.
Expect the first week to be mostly wrong, and read the rejections. Every person Signl sets aside keeps a reason in words in the Filtered out tab. If people you would sell to are sitting in there, the fix is your ICP, not a lower bar.
What to say to someone who asked for a recommendation
Answer the question. In public, under their post, like a person who knows the answer.
This is the one engagement type where naming your product is fine, because they asked. Keep it to one sentence, put the honest caveat in, and do not pitch the thing they did not ask about. If three other vendors have already piled in, saying something genuinely useful about the choice beats being the fourth link.
Direct messages work on this signal too, but they are strictly worse: slower, invisible to the other forty people reading the thread, and they lose the context that made the request answerable. Save the message for the stated-problem patterns in group 3, where there is no question to answer. Rewrites for those are in how to message someone who commented on a post.
Honest limits
- Volume is low. This is a precision instrument. Run it alongside competitor and creator sources, which is the point of the wider setup in LinkedIn social listening.
- Search behaviour changes. Quoting and operator support in LinkedIn's own search has differed between tabs and over time. Test every phrase by hand where you are using it, and keep the ones that behave.
- The best requests happen in private. Slack groups, WhatsApp, and DMs between peers carry most referral traffic, and none of it is visible or should be.
- You are late by design. A post asking for recommendations is already a shortlist being formed. Being useful in the thread is most of what you can control.
- Automated collection is contractually contested. LinkedIn's User Agreement section 8.2 prohibits scraping and bot access (read 24 October 2026), which is a separate question from whether data protection law lets you contact someone. Our lawful-basis position is in GDPR and B2B prospecting. This is not legal advice.
