A LinkedIn buying signal is a public action that narrows who is worth contacting. They are not equal: an explicit request for a vendor and a thumbs up on a post differ by an order of magnitude in what they tell you. Rank them by how much of the person's own language they carry.
Disclosure: Signl is ours, and the thresholds quoted below are its real defaults, checked against our code on 24 October 2026. No reply rates or conversion figures appear anywhere in this article. We have not measured them in a way we could reproduce, and this topic is full of numbers nobody can reproduce.
Key facts
- The test for signal strength is how much of the person's own language the signal carries.
- A comment carries words. A reaction carries none. Everything else sits between them.
- Every signal has something it cannot predict, and the useful question is what that is.
- Signals from several people at one company are an account signal, not several leads.
- A signal tells you someone has a problem. It never tells you they have budget or authority.
What makes something a signal
Most lists of LinkedIn signals are lists of actions the platform exposes. That is the wrong organising idea, because it treats "followed your company" and "wrote three sentences about their broken process" as members of the same category.
A better test: how much of the person's own language does the action carry? Language is what lets you tell a buyer from a bystander, judge their stage, and write an opening line that does not read as a template. An action with no language attached forces you to guess from a job title, which is how outbound became what it is.
That single test produces the ranking below, and it also explains the thresholds. In Signl, a reaction has to clear a default confidence of 60 before an email is bought, against 40 for a comment, because the reaction gives the classifier less to work with and should therefore need more certainty elsewhere.
The ladder
Six tiers, strongest first. The third column is the one we think most content on this query omits, and it is the column that keeps you honest.
| Tier | Signal | What it tells you | What it does not predict |
|---|---|---|---|
| 1 | A request for a recommendation, with a constraint | They are shortlisting now, and the constraint tells you the fit | Whether you are already too late. Threads fill in hours |
| 2 | Named dissatisfaction with an incumbent | A live problem, an objection you can use, and a budget that already exists | That they will switch. Most complaints end in staying |
| 3 | A stated problem with a number or a date | A real, sized pain, in their words | That solving it is anyone's priority this quarter |
| 4 | A substantive question or disagreement | Engagement with the subject, and their vocabulary | Whether the problem is theirs or their client's |
| 5 | A short on-topic comment or a repost with words added | Topic interest, and a name to research | Anything about their situation |
| 6 | A reaction, a follow, a profile view | Attention, at most | Almost everything. Treat as a tiebreak, not a trigger |
Two practical rules fall out of the table.
Work tier 1 and 2 by hand. There are few of them, they decay in hours, and a generic message wastes the best signal available in the channel. How to find buyers asking for recommendations covers the search patterns that surface tier 1.
Automate tiers 3 to 5 and ignore tier 6 from strangers. Tier 6 on your own posts is a different matter: someone reacting to your post already knows who you are, so the relationship carries the weight the signal does not.
Signals that look strong and are not
Four traps, in rough order of how much time they waste.
Congratulations under a funding or promotion post. High volume, zero information. The person congratulated someone. Signl's free rules set aside generic praise of six words or fewer before any model runs, which catches most of this at no cost.
Engagement pod comments. Where pods operate, the same twenty names appear under every post by one author, with comments that read as on-topic and mean nothing. LinkedIn's Professional Community Policies tell members not to "agree with others ahead of time to like or re-share each other's content" (read 24 October 2026). No tool detects this reliably, including ours. The manual test is whether one author's comment section is the same faces every week. If so, drop the source.
Comments from people whose problem is not theirs. "My clients struggle with this" is an agency, a consultant or a competitor. They may be worth a partner conversation and they are not a prospect, and the pronoun is the fastest way to tell.
Job seekers in a hot thread. The more popular the post, the higher the share. Headlines saying open to work or actively looking are the cheap filter; a classification step catches the rest. In Signl, anyone judged a job seeker, a vendor selling their own service, or noise is never enriched, however confident the model was.
Stacking, and why it changes who you write to
One person engaging once is a lead. Three people from the same company engaging with the same topic inside a month is something else: an internal conversation you can see the edges of.
Handled as three leads, you send three messages into one account and look automated. Handled as an account signal, you ask a better question, which is who at that company owns this decision and has said nothing in public. That person is often the right recipient, and the three commenters are the evidence that gets you in.
Signl groups the inbox by company on request, showing each company's engaged people, their roles, and the roles your ICP sells to that nobody there holds yet. The scoring principles behind combining signals are in how to score a buying signal.
Freshness
Engagement signals decay faster than company signals, which is the main structural difference between this and the funding or hiring signals in our signals cluster.
We do not publish a decay curve. What we can show is where we put our own money: Signl checks a post hourly during its first two days and looks for new posts every three hours on hourly sources, then falls back to a daily full check. That cadence is a bet that most of the value in a thread arrives in the first 48 hours, which matches what we see in our own queues but is not a measurement we would ask you to rely on.
Collected engagement is deleted 90 days after first collection, so there is also a hard horizon by design.
As a working rule: tier 1 and 2 signals are worth acting on the same day, tiers 3 to 5 within the week, and anything older is context for a conversation you were having anyway.
Turning a signal into a message
The ladder exists to decide two things: whether to contact someone, and what you may say.
What you may say is governed by where the signal happened, not by how strong it was. On your own post you can say they commented. On a creator's post you can mention the topic. On a competitor's post you say neither, because naming the source tells someone they were watched; the full treatment of that is in finding leads in competitors' comments.
And the quote has to be real. Signl requires the fragment an opener answers to appear verbatim in the comment, and substitutes a plain line when it does not, because an invented quote is worse than no quote. Eighteen worked rewrites are in how to message someone who commented on a post.
What none of these signals can tell you
The limits, stated plainly, because a signal oversold is how this channel gets a bad name.
- Budget and authority. A perfect comment from someone with neither is still a perfect comment.
- Timing. "We still do this manually" has been true at some companies for nine years.
- Whether anyone else is already there. Public signals are public, and your competitors can read them.
- Anything private. Followers, connections, post viewers and private messages are not available, from us or from anyone selling them legitimately.
The automated collection of public engagement is also contractually contested: LinkedIn's User Agreement section 8.2 prohibits scraping and bot access (read 24 October 2026). That is a separate question from data protection law, which governs whether you may contact the person at all; our position on lawful basis is in GDPR and B2B prospecting. This is not legal advice.
Next, if you are setting this up: LinkedIn social listening for the full picture, or LinkedIn comment tracking for the operational side of keeping signals over time. If you are deciding whether to run this way at all, intent-based outbound vs cold outreach weighs the model, and first-party vs third-party intent data places public engagement among the other kinds of signal you can buy.
