How to spot a buying committee forming on LinkedIn, weeks before it reaches your CRM

Take a 70-person company that sells scheduling software to freight forwarders. Between the first week of June and the middle of July, four people at the same mid-size forwarder reacted to the company's LinkedIn posts. An operations manager liked the CEO's post about detention fees. Two weeks later a dispatcher supervisor commented on a product marketer's post about driver no-shows. A few days after that, the head of IT liked the company page's integration announcement, and in July a regional director liked the CEO again.

Nobody at the software company noticed. Each of those reactions landed in a different person's notifications, on a different day, next to forty other names. The CEO saw "an ops manager somewhere liked my post." The product marketer saw one comment among nine. The company page admin doesn't read reaction lists at all.

In late August the forwarder filled in the demo form. The AE opened the CRM, found an empty account record, and started discovery from zero with a single contact, six weeks after four people across three departments had shown their hand.

This article is about not being that AE. The method is simple to describe: stop reading LinkedIn engagement person by person and start reading it company by company. It is tedious to run by hand, and we'll be plain about where it breaks, but you can start on Monday with a spreadsheet and nothing else.

The person is the wrong unit

Most advice about LinkedIn and sales is built around individuals. Someone liked your post, so send them a message. Someone viewed your profile, so connect. It treats each reaction as a small lead.

We think that's the wrong unit of analysis for anyone selling to companies, and the reason is arithmetic. A like costs the person who gives it about one second and no commitment. People like posts because the author is a former colleague, because they're job hunting, because they work for a competitor and are keeping tabs, because the post showed up while they were waiting for a coffee. Any one reaction has a dozen plausible explanations, and "this person's company is about to buy software like ours" is rarely the most likely one.

Meanwhile, the purchase you care about isn't made by a person. Forrester's 2024 survey of business buyers put the average at 13 people involved in a buying decision, with 89% of purchases involving two or more departments. 6sense's 2025 buyer research, run on more than 4,000 buyers, found that buying groups average ten or more members on deals around $250,000. Your deal size may be a tenth of that and your groups smaller, but the shape holds for almost any B2B sale above a credit card swipe: several people, more than one department.

The same 6sense research has the number that should bother a sales leader most. Buyers made first contact with a seller about 61% of the way through their journey, and 94% of buying groups had ranked their preferred vendors before that first contact, buying from the early favorite 77% of the time. By the time the form is filled in, the group has existed for a while, it has talked, and it has opinions about you.

So there's a stretch of weeks or months in which a group is forming and your CRM knows nothing. Some of that group's activity is invisible to you and always will be: internal Slack threads, analyst calls, a peer's recommendation over dinner. A slice of it is public, though. If your company posts on LinkedIn, some of those people will react to what you publish, under their real names, with their employer in their headline. The information is sitting in your reaction lists. It's just filed by post and by person, when the question you need answered is filed by company.

Why one person is weak evidence and four people on three posts is strong

Here's the reasoning, because the threshold matters less than understanding why any threshold works.

One person from an account reacting once tells you almost nothing, for the reasons above. One person reacting five times tells you something about that person. It's worth a look (scoring individuals is a separate subject and we cover it in another article), but it still says little about the company. Enthusiasts exist. Some people like everything.

Two people from the same company is where it gets interesting and also where people fool themselves. LinkedIn shows your activity to your connections, and colleagues are connected to each other. When the ops manager likes a post, the dispatcher supervisor may see it in her feed precisely because he liked it. Two reactions from one company on one post can be a single event with an echo. They aren't independent observations, and treating them as two data points overstates what you know.

That's why a good rule counts distinct posts as well as distinct people. Four people on one post might be a cascade through one office's feeds. Four people spread over three posts, published on different days by different authors, is much harder to explain away. Each post is a separate occasion on which someone at that company saw your content and chose to react. Coincidence gets expensive as an explanation. A shared interest inside that company gets cheap.

The rule we use, and the one built into Resonue, is four or more people from the same company, across three or more posts, within 90 days. The honest status of those numbers: they're a judgment call. Nobody ran a study showing that four is the inflection point. The reasoning for each goes like this.

Four people, because two can be a manager and a direct report, and three can still be one tight team. At four you usually start to see a second function or a second level of seniority, and cross-functional is what a buying group looks like.

Three posts, because it kills the single-cascade problem and the single-viral-post problem in one move.

Ninety days, because a quarter is roughly the rhythm at which companies plan and budget, and because longer windows fill up with people who have since changed jobs. If your sales cycle runs nine months, a 120-day window is defensible. Much shorter than 60 days and you'll miss slow-forming groups, and slow is probably the normal case: the 6sense report cited above puts the average buying cycle at about ten months.

Now the concession. The threshold should really scale with company size, and a flat rule doesn't. Four people out of a 60-person company is 7% of the staff paying attention to you, which is remarkable. Four people out of a 40,000-person bank may be four strangers in four countries who will never be in the same meeting. For large accounts, tighten the rule by hand: require that the people share a business unit or a country, or raise the bar to six. For very small companies, two founders engaging repeatedly may be the entire committee, and waiting for four means waiting forever.

Pick a number, write it down, and keep it fixed for a couple of quarters so you can learn whether it was right. A threshold you keep adjusting after the fact teaches you nothing.

Rolling engagement up by company, by hand

You need one spreadsheet with two tabs. Resist adding a third.

Tab one: the touch log

One row per person per post. This is the raw material and it's deliberately dumb.

Tab two: the account roll-up

A pivot on company_key, filtered to employer_confidence = high, with these columns:

The weekly ritual

Pick a fixed slot. Friday afternoon works because the week's posts have finished collecting reactions. Whoever runs it opens each post published by the company page and by the three to six employees whose posts reliably draw an audience, clicks into the reaction list, and logs every name that isn't a colleague, a relative or an obvious bot. Comments get logged from the comment thread. For a team publishing eight to ten posts a week at 30 to 60 reactions each, the first session takes about two hours. Once you know the regulars by sight, it drops to around 45 minutes.

Two practical warnings. First, a post keeps collecting reactions for a week or more, so revisit last week's posts before starting on this week's. Second, log customers too, and tag them as customers in notes or in a separate column if you like. A customer account where four new people start engaging is an expansion signal, and a churned account reappearing is worth a call on its own.

People whose employer is ambiguous

This is where the hours go, so set rules in advance and stop deliberating case by case.

Use the current primary position, not the headline. Headlines say things like "Helping logistics teams scale" and tell you nothing. If the person lists two current roles, take the one that looks like a full-time job and ignore the advisory board seat.

Freelancers, fractional executives and consultants get employer_confidence = low and stay out of the roll-up. One exception deserves a note: a consultant or agency person who engages repeatedly may be running a vendor evaluation on behalf of a client. You can't tell which client from a like. Log them and let the AE decide whether to ask.

"Stealth startup", "Self-employed", students and open-to-work profiles with no current position get none. Don't guess.

Subsidiaries and groups are the hard case. Roll up to the entity that would sign your contract. If you sell to individual country operations, DHL Freight Sweden and DHL Supply Chain UK are two accounts. If you sell group-wide, they're one. Decide per account type, write the decision at the top of the sheet, and apply it the same way every week.

Job changers need a judgment. If someone engaged in May at one company and moved to another in August, their May touches belong to the old employer. Leave history alone, and log new touches under the new company. (A champion who moves to a new company is a lead in its own right, but that's an individual-level play.)

Normalize names by website domain where you can. "Kuehne+Nagel", "Kuehne + Nagel" and "Kuehne and Nagel (AG & Co.) KG" all become one company_key. A lookup tab is tempting here. A find-and-replace on Fridays is enough.

Reading who is in the group, and who isn't

A flagged account tells you that several people are paying attention. It doesn't tell you what kind of attention. The next step is to look at the titles and sort them into the roles a purchase needs, which is a human job and takes about two minutes per account.

For most B2B software deals, the roles are roughly these: the people who would use the product day to day, the manager who owns the problem and would champion a fix, the person who holds the budget, the technical evaluator (IT, security, data, whoever has to connect and approve it), and the late-stage gatekeepers in finance, legal and procurement. Write in roles_seen which of those you can see, using titles only. Don't infer anything else.

Then read the gaps, because the gaps are where the instructions for the AE are.

One absence is not a signal. Finance, legal and procurement almost never react to vendor posts, and their silence on LinkedIn means nothing. The 6sense and Forrester numbers above describe groups of ten to thirteen. If you can see four of them, you're seeing the four who happen to be active on LinkedIn, which skews toward commercial and operational roles and away from the people who will scrutinize the contract. So when the AE builds an account plan from a flagged group, the plan should already include a line for whoever signs off on security and whoever negotiates terms, even though neither will ever show up in the log.

Multi-threading without being weird about it

Knowing that four people at an account have engaged creates a temptation to contact all four on Tuesday morning. Don't. They sit near each other, or at least share a Slack channel, and "did you also get a message from that scheduling vendor?" is a conversation that ends with all four ignoring you.

The second temptation is to show your work. Compare these two openers.

The bad one: "Hi Marta, I noticed you and three of your colleagues at Nordfracht have been engaging with our content recently, so I wanted to reach out and see if scheduling is a priority for the team."

The better one: "Hi Marta, you commented on Jonas's post about driver no-shows, the bit about carriers confirming and then not turning up. We hear the same from most mid-size forwarders. Are you handling that with phone calls today, or is there tooling in place?"

The first message tells her she's being watched and tells her you've profiled her colleagues. The second refers to something she did in public, on purpose, and asks a question she can answer from her own desk. The account-level picture is why you're writing to her. It doesn't need to appear in the message. (How to write that first message well is the subject of another article in this series. Here we care about the order and spacing of the threads.)

A sequence that works without setting off alarms:

1. Start with one person. Pick the one whose role is closest to the problem and whose engagement is most recent, and prefer a commenter over a liker, because a comment is an open door you can reply to in public first.
2. Wait for a response, or a week, before opening a second thread.
3. Open the second thread from a different sender matched to the recipient. The AE talks to the manager. Your head of product or CTO talks to their IT lead. Your CEO talks to their director, briefly, peer to peer.
4. In the first real conversation, ask the question that turns your private map into shared knowledge: "Who else on your side cares about this?" Once Marta says "you should talk to Henrik in IT", you can contact Henrik openly, with her name, and the fact that Henrik liked your integration post three weeks ago stays what it should be: background that told you the question was worth asking.

If the first person doesn't reply, that isn't a verdict on the account. Go to the second person with a different angle suited to their role, not a copy of the first message.

What cold to warm looks like over 30 days

An account warming up has a recognizable pattern in the roll-up tab, and it helps to know it because several things imitate it.

Here's the freight forwarder again, week by week, as the spreadsheet would have shown it. Week one: one person, one touch. Week two: nothing. Week three: a second person, on a different post by a different author. Week four: the first person again, plus a third person. In the 30-day comparison, touches_prev_30d was 0 and touches_30d is 4, people_30d is 3, across three posts. One more person in the following fortnight and the committee flag trips.

Three properties separate that from noise. The number of people is rising along with the number of touches (one fan liking everything inflates touches and leaves the people count flat). The touches are spread across several posts and ideally several authors. And they're spread across weeks, with gaps, which is how busy people behave when something is slowly becoming a priority.

Now the imitations.

One viral post is the big one. Your CEO writes something that travels, it collects 900 reactions where a normal post gets 50, and the following Friday half your roll-up tab is up 300% against the previous 30 days. Almost none of it means anything. The test is quick: exclude that post_id and recompute. An account that is still gaining without the viral post is warming. An account whose entire gain comes from it has noticed you once, which is worth something, but it goes on a watch list and not to an AE. This is also why posts_90d sits in the committee rule. A viral post can deliver four people from one company in an afternoon, and it's still one post.

Hiring posts attract people who want a job with you. Funding announcements attract investors, vendors and recruiters. Event weeks produce a burst from everyone who visited the stand, which is real interest but already known to the field team. Competitors' employees show up steadily, and you should tag their company_key once and filter them out for good. Your own employees' former colleagues react out of friendship, and cluster suspiciously at whichever company your newest hire just left.

None of this needs sophisticated handling. It needs a person who looks at the top movers each week and asks "which posts did this come from?" before anything gets forwarded to sales.

Who does what: marketing keeps the log, sales reads the accounts

This work falls between two teams, and in our view that, more than the tedium, explains why it rarely gets done. Marketing sees the engagement and doesn't own accounts. Sales owns accounts and never opens a reaction list. A workable split follows.

Marketing owns the log and the roll-up. They publish or coordinate most of the posts and they're already in the analytics. They run the Friday session, apply the employer rules, run the viral-post check, and produce a short list for Monday: newly flagged accounts, top movers against the previous 30 days, and any customer or churned account that reappeared.

Sales owns the reading and the move. The AE or SDR who owns each account fills in roles_seen, decides who to approach first, and writes the next_step. Marketing shouldn't guess at committee composition for accounts it doesn't know, and should never relabel a flagged account as an MQL and push it into a sequence. An MQL implies someone raised a hand. Nobody here has. These are accounts worth a thoughtful, human first move, and an automated five-step cadence is the opposite of that.

RevOps, if you have it, owns the definitions: the threshold, the subsidiary rule, what "influenced" means in reporting. One person should own these in writing, because they will be challenged the first time a number shows up in a board deck.

The meeting is twenty minutes on Monday. Marketing brings the list, sales says what they already know ("we're in a late-stage deal with them, that's why"), and each new account leaves with an owner and a next step or an explicit "watch". Feedback has to flow back too. When an AE reports "I called, it's four interns", that's what tells you whether the threshold is set right.

Giving LinkedIn credit without lying to yourself

Eventually someone asks what all this posting is worth, and the roll-up can answer more honestly than most attribution can, provided you respect one rule about time.

An account counts as influenced only if its first logged engagement came before the deal stage moved. That's why first_touch_date is never overwritten. If the forwarder's first touch was June 3 and the opportunity was created August 27, LinkedIn was in the picture twelve weeks before sales was. If an account's first engagement comes two weeks after the AE's first meeting, it doesn't count, however warm the account looks today. That engagement is a consequence of the sales process: people look up the vendor they just met. Counting it credits the channel for your AE's work.

Report two numbers per quarter: influenced pipeline (open opportunity value where engagement preceded the stage change) and influenced revenue (the same for closed-won).

Then be clear about what these numbers are. They establish sequence, not cause. "They engaged first and bought later" is compatible with "the posts helped them choose us" and equally with "they were already going to evaluate us, and people who are evaluating you tend to read your posts". You can't separate those from a spreadsheet. A comparison makes the number more useful without pretending to settle it: of all opportunities created this quarter, what share had prior engagement, and do those deals differ from the rest in win rate or cycle length? If engaged-first deals close no better than the others, say so. If they do close better, you have an association worth investing in, and you should still call it influence, because that's what it is.

There's a further reason to log dates carefully. Research by John Dawes at the Ehrenberg-Bass Institute for the LinkedIn B2B Institute argues that up to 95% of business buyers aren't in the market for a given category at any one time, since companies switch providers of things like software roughly every five years. If that's even approximately right, most accounts engaging with you this quarter won't buy this quarter. Some of them will in eighteen months, and the only way you'll be able to show that the relationship started on LinkedIn is a first-touch date logged when it happened.

What this method can't see

It only sees accounts where somebody has engaged at least once. An account on your target list with zero rows in the log isn't a no. Most people who read posts never react to them, and whole companies have cultures of reading silently, particularly in regulated industries where employees are wary of public activity. Silence tells you nothing either way. Your outbound plan for those accounts shouldn't change because of an empty row.

It only sees what your team publishes. If you post twice a month and get fifteen reactions, there's nothing to roll up, and the right move is to fix the publishing before building the spreadsheet. As a rough floor, you want a few posts a week across several people before account-level patterns become readable.

It can't see intent on channels you don't own: review sites, communities, a competitor's webinar. It complements the intent data you may already buy and doesn't replace it.

It works with small numbers. Four people across three posts is a sensible flag and a poor statistic. Expect false positives, the four-interns kind, and expect accounts that buy without ever tripping the flag. Judge the method by whether the Monday list produces better first conversations than the AEs would have had otherwise.

Where doing it by hand breaks, and what Resonue automates

The spreadsheet works. We'd recommend a quarter of running it by hand to anyone, partly because it forces the definitions to get settled. It breaks in predictable places.

Volume goes first. At ten posts a week the Friday session is manageable. At thirty posts across fifteen employees, it's most of a day, and the person doing it starts skipping the low-reaction posts, which is exactly where the fourth person from a flagged account tends to show up. Timing degrades next, since weekly logging with relative timestamps gives you dates accurate to within a week. Employer resolution is the slowest step and the most error-prone. And the 30-day comparison needs a clean history that a hand-kept log rarely has after the third month.

Resonue is the product we build, and the account side of it is this method running continuously. You track your company page, your employees and any outside voices who post about you, and every person who engages with a tracked post is recorded and rolled up by company. The companies board shows, for each account, interactions and distinct people over the last 7 and 30 days, plus an account heat figure, which is the sum of the heat scores of the individual engagers at that company (heat weights recent and repeated engagement above old, one-off engagement). Buying committee detection applies the rule described above, four or more people across three or more posts within 90 days, and flags the account on the board and the dashboard. Opening an account lists the people behind the number with their headlines, their individual heat, and which of your voices reached them.

The 30-day momentum report does the cold-to-warm comparison: new engagers, new warm accounts, and a movers list of accounts gaining interactions against the previous 30 days. The Monday list arrives by itself too. A weekly brief email goes out on Monday morning to everyone in the workspace, flagging accounts where a committee is forming and customer or churned accounts that re-engaged in the past week.

The pipeline board handles the attribution rule. Every account that engages lands in an inbox, and you move it through Lead, Opportunity, Customer, Churned or Disqualified by hand. When you set a deal value on an opportunity or a customer, the account counts toward influenced pipeline or influenced revenue only if its first tracked engagement predates the stage change. Engagement that arrives after you moved the card earns nothing, by construction.

One habit from the hand method is worth keeping. The three-post condition protects the committee flag, but a movers list after a breakout post still deserves a skeptical look.

If you'd like to see this on your own accounts, talk to sales. We prepare the demo on your company's own recent posts before the call, so the conversation starts from the accounts that have actually been reacting to you, including, quite possibly, a group of four you hadn't noticed.