Employee advocacy on LinkedIn usually dies by week six. Here is the version that survives

Take a 140-person company that sells procurement software. Its LinkedIn company page has 6,200 followers and the marketing team posts on it three times a week. Every month someone exports the page analytics into a slide: impressions up 8%, 41 new followers, engagement rate a little under 2%. The slide gets about ninety seconds in the marketing review and nobody in sales has ever asked to see it.

In the same month, one of the account executives wrote a post about a deal she lost because the buyer's finance team was never in the room. It got 40 reactions. Eleven of them came from people at companies on the sales team's target list, two of them finance directors. One of the solutions engineers posted a screenshot of an ugly approval chain he had untangled for a customer, and a procurement consultant with a large following reshared it with a comment of his own. A customer's head of operations mentioned the product by name in a post about her quarter, unprompted.

None of that is on the slide. The slide only knows about the company page, and the company page is the smallest part of what happened.

This article is about running LinkedIn as a revenue channel across everyone who talks about your company, and about why the usual way of organizing that (an employee advocacy program) tends to collapse within a couple of months. Most of what follows can be done with a spreadsheet and some discipline. Near the end we'll say where the spreadsheet stops being enough and what we built at Resonue to take over from it.

Four kinds of accounts talk about your company

Start with an inventory, because most teams have never made one.

The first kind is the company page. You own it, you have its analytics, and it is the only one of the four that LinkedIn hands you numbers for.

The second is your employees. Founders, AEs, product managers, the support lead who writes sharp posts about onboarding. LinkedIn's own advocacy guide claims that on average employees' networks are at least 10x larger than the company's follower base, and that only 3% of employees share content about their company while accounting for a 30% lift in the engagement the company sees. Treat those figures with some care: the guide is several years old and was written to sell an advocacy product LinkedIn has since retired. You don't need LinkedIn's average anyway. Add up the connection counts of your twenty most active colleagues and compare the total to your page followers. For the procurement company above, twenty people averaging 1,800 connections is 36,000, against 6,200 followers. There is overlap between those networks, so the real number is smaller, but it is still a multiple of what the page reaches.

The third kind is outside people who mention you without being asked: customers, consultants, analysts, a former employee who still likes the product, occasionally a critic. You control nothing here. It is often where the most credible posts come from, for the plain reason that the author has nothing to sell.

The fourth, which only some companies have, is contracted content partners: people with an audience in your market whom you pay to post. The wider market calls them influencers. It is a minority practice in B2B and we'll deal with it in one section further down.

(At Resonue we call each of these accounts a Voice, and that is the last you'll hear from us about the product for a while.)

Here is the inventory to build. One row per account, and it takes an afternoon:

To find the externals, search LinkedIn posts for your company name and your product name, filter to the past month, and do it again next month. It is tedious and you will miss some.

The posts_mentioning_us_90d column tends to surprise people. At most companies, five or six employees produce nearly all the posts that mention the company, and marketing had never listed them. Those people are your program. Whatever you launch should be built around them and not around the 134 colleagues who have never posted and have no wish to start.

The page report tells you about the page

There's nothing wrong with company page analytics. They accurately describe one account. The trouble starts when that report stands in for "how LinkedIn is going", because then most of what LinkedIn is doing for you, or failing to do, sits outside the measurement.

There is also a structural reason the page matters less than its report suggests. In the 2024 Edelman and LinkedIn study of 3,484 business executives, 73% of decision-makers said an organization's thought leadership is a more trustworthy basis for judging its capabilities than its marketing materials and product sheets. A company page post reads as marketing material however good it is, because a logo posted it. The AE's post about the lost deal reads as a person who knows something. Same company, same subject, a different amount of belief on the other end.

The same study has a number that should bother anyone who runs this channel: only 29% of the companies producing thought leadership said they can link sales leads back to specific pieces, 42% still judge it by looking for traffic bumps, and 19% have no measurement process at all. So plenty of teams are in the position of the procurement company: they can describe the page in detail and have nothing to say about the rest.

One concession before going further. The page still earns its place. It is where a prospect lands when they check whether you are a real company, it is the account that can run ads, and it is the natural home for hiring posts and announcements nobody wants on their personal profile. Keep posting there. The argument is only that it shouldn't be the whole report.

Why advocacy programs die around week six

"Six weeks" is our shorthand for a pattern, not a measured statistic. Nobody has surveyed this, as far as we can find, so take it as a rough shape. The shape is consistent enough to be worth walking through.

Week one: the CMO announces the program at all-hands. There is a Slack channel, maybe a tool that queues up suggested posts. Thirty people join.

Weeks two and three: marketing drops a message in the channel whenever the company page publishes. "New post is live, please like and share!" A dozen people repost it. Their networks now see the same graphic twelve times with no commentary, or with the suggested caption pasted in unchanged, which is worse. An identical repost gives a reader no reason to stop: they already scrolled past the original. The people doing the reposting can see this in their own notifications, where each repost collects two likes, both from colleagues.

Week four: someone builds a leaderboard to revive things. The same three people win it. Everyone else learns they are in the bottom twenty-seven.

Week five: the head of sales asks what the program has produced. Marketing answers with total impressions. The head of sales nods politely and stops asking.

Week six: the Slack channel has one message in it, from marketing, with a single emoji reaction from an intern.

Look at what failed and none of it was laziness. The ask was wrong (amplify our post, when it should have been say something of your own). Nobody who participated got anything back: no sign that their effort reached a person who mattered, no sense of what worked for them. And the one executive whose support would have protected the program was given a vanity number. Each of those has a fix, and the fixes are the rest of this article.

Start smaller than feels respectable

A program that survives usually starts with five to eight people, all volunteers, drawn from the ones who already showed up in your inventory. Include at least one AE, because sales leadership will believe a colleague's results before it believes marketing's, and at least one person who builds or supports the product, because those posts contain details marketing can't write.

You cannot force employees to post. Their profiles belong to them, their networks were built before they met you and will leave with them, and a mandated post reads like one. Some companies put LinkedIn activity in AE performance plans. We'd advise against it: you get the minimum number of posts at the minimum level of effort, and the good posters start to resent a thing they used to do willingly. What you can do is make posting easier and more rewarding for the people inclined to do it, which comes down to two things. Tell them what is worth talking about this month, and show them what their posts did.

Give it a named owner with about two hours a week. Here is what that person's Tuesday looks like: twenty minutes checking which posts from last week touched the current brief, thirty minutes going through who reacted to the three or four most relevant posts, ten minutes sending two individual notes ("your post on approval chains pulled in four people from target accounts, two of them at Halden, I've told their AE"), and the rest on next month's brief. If nobody has those two hours, don't launch. A program with no owner doesn't even make it to week six.

Brief people like you're running a sprint

The share-this-post message fails partly because it has no shape. It arrives at random, it asks for a mechanical action, and it never ends. A brief is different in three ways: it has a subject, a priority, and dates.

Here is one for the procurement company, tied to a feature launch:

Brief: approval workflows launch. Priority: HIGH. Window: October 6 to October 24. What's happening: multi-step approvals ship on October 13. The problem it addresses is invoices stalling because one approver is on holiday and nobody else has authority. What we'd love covered: that problem, from wherever you sit. An AE has heard buyers complain about it. Support has seen the tickets. Product knows why it was hard to build. What's true and safe to say: it ships October 13, it's included in all plans, it supports up to five approval steps. What to leave out: customer names unless marketing confirms permission, anything about the Q1 roadmap. There is no suggested copy. Write it the way you'd say it to a peer.

A few things about this are deliberate.

The window is under three weeks. Open-ended asks get ignored because there's never a reason to act today. A window that brackets the launch gives people a reason to post this week and gives you a period over which to measure. Events work the same way: for a conference on March 12, open the window three weeks before, since the point is to have conversations warmed up before you arrive at the booth.

The priority exists so that people can ignore most briefs. If you run one HIGH brief a month and a couple of LOW ones ("we're hiring two engineers", "new case study is up"), your volunteers know where to spend their one good post. Marking everything HIGH is the same as having no priorities at all, and your volunteers will work that out within a month.

There is no copy, and that is where most programs lose their nerve. Marketing worries about message consistency, so it supplies a caption, and twelve identical captions later you are back in week three. Give people the facts and the guardrails and leave the writing to them. If someone asks for help getting started, sit with that one person and work from how they already write.

Then close the loop, which is the step nearly everyone skips. When the window ends, go through what your people posted during it and mark each post: did it cover the brief, yes or no. Add a tab to the spreadsheet:

Now you can answer questions you couldn't before. Six of eight volunteers posted about the launch. Those posts drew reactions from 31 people at accounts sales cares about. The March event brief got one post, which says something about the brief or the event. Reading posts and judging "covered or not" by hand is slow and somewhat subjective, and it's still worth an hour a month.

Give each person their own numbers

The leaderboard fails for a reason anyone who has managed salespeople will recognize: ranking motivates the top three and tells everyone else to stop. It's also unfair in a specific way on LinkedIn. A founder with 14,000 followers and a support engineer with 600 connections aren't playing the same game, and raw engagement will always put the founder on top, whatever either of them wrote.

Compare each person to their own past instead. For every volunteer, keep their last 30 days next to the 30 days before: how many posts, total reactions and comments, how many distinct people engaged. Then send it to them privately, with one observation a human wrote. "Your posts with a screenshot of a real workflow get about double your usual comments" is worth more than any rank, because it tells them what to do next and because it is about them.

A comparison to the team average is fine if it stays private and kind. "You're at about 1.3 times the team's average engagement per post" is encouraging. Publishing the same figure in a channel turns it back into a leaderboard.

Be open about what you're tracking. Everything here comes from public posts, but it is still colleagues looking at colleagues' activity, and the volunteers should have opted into that (hence the opted_in column). Don't run numbers on employees who never signed up and then present them with a report. Someone will call it creepy, with some justification, and you'll spend the next month on that conversation when you meant to spend it on the program.

Judge a post by who it reached

Here is the claim we'd defend hardest in this article: a post with 40 reactions from the right accounts beats a post with 900 from strangers, and most teams reward the second one.

The 900-reaction post is usually a personal story, a hiring opinion, a take on remote work. There's nothing wrong with those posts, and they do grow the author's reach over time. But open the list of reactors and sample a hundred names. For a company selling procurement software, you'll typically find recruiters, students, other people's salespeople, and a handful of relatives. Perhaps three people out of the hundred could ever buy.

Now the 40-reaction post about the lost deal. Go down the list, all 40, and put each person in one bucket:

Thirteen people you sell to, plus four customers worth keeping warm, from one post. The 900-reaction post, extrapolating from its sample, reached perhaps 27 such people, at a rate of 3%. The small post ran at 13 of 31 non-colleagues, 42%. They are closer in absolute terms than the like counts suggest, and the smaller post came with context: those 13 reacted to a post about finance being absent from procurement deals, which tells the AE what to open a conversation with.

Track two things per post, then: right_audience_count (people in the second and third buckets) and the rate (that count divided by non-colleague reactions). The count tells you what the post was worth. The rate tells you whether the topic attracts buyers or passers-by.

Always strip out colleagues. At small companies, internal reactions can be a quarter of the total, and they flatter every number you look at.

Doing this by hand costs about fifteen minutes per post for 40 reactors, and it isn't practical for 900, which is why you sample. Do it only for posts that covered a brief and for anything that performed unusually. What to do with the 13 names (which to contact, how warm each one is, what to write to them) is a subject of its own, covered in our article on turning post engagement into sales conversations. Whether several of those people work at the same account, and what that signals, is covered in the one on buying committees.

Let your own history pick your topics

Every quarter a new list circulates: best times to post, ideal length, carousels are back, carousels are finished. These lists are averages across millions of accounts that mostly don't sell what you sell to the people you sell to. Your own posting history is a smaller dataset and a much more relevant one.

Tag each post from the last six months, across all your tracked voices, with one or two topics. Keep the vocabulary short, fifteen topics at most: "approval workflows", "finance and procurement alignment", "hiring", "company news", "customer stories". For each post, compute a simple engagement figure (reactions plus comments, divided by the author's follower count, so the founder doesn't dominate). Average that across every post, and you have your company baseline. Then average it per topic and compare.

You'll hit a trap almost at once. "Customer stories" has two posts and an average three times the baseline, because one of the two went unusually well. "Approval workflows" has fourteen posts and sits at 1.2 times baseline. A naive ranking crowns customer stories on the strength of a single good day.

The fix is to pull small samples toward the baseline:

adjusted = (n × topic_average + 3 × baseline) / (n + 3)

where n is the number of posts on the topic. With two posts at 3.0 times baseline, the adjusted figure is (2 × 3.0 + 3 × 1.0) / 5 = 1.8. With fourteen posts at 1.2, it's (14 × 1.2 + 3) / 17 = 1.16. Customer stories still ranks first, and it should, but now by a margin that reflects how little you know. The constant 3 is a judgment call: it means a topic needs a handful of posts before its own record outweighs the company average. Use 5 if your volume is high.

Then tier them. We'd use these cutoffs: at or above 1.5 times baseline with at least two posts is an S, at or above baseline is an A, down to 0.6 is a B, below that is a C. S and A topics go into next month's briefs. B topics need a better angle. C topics are where you stop asking volunteers to spend their posts. "Company news" is often one of them, and that's useful to know: the page can carry it and the volunteers can be spared.

Two cautions. Engagement rate is a proxy, and the better measure is the right-audience rate from the previous section, if you have enough classified posts to compute it per topic. Most teams won't for months. And a topic can score badly because the posts about it were dull. Reread the posts before you retire the topic.

Competitors: look monthly, then close the tab

It is worth knowing how much of the conversation in your category belongs to you. It isn't worth checking every morning.

Once a month, pick three or four competitors. For each, count posts in the last 30 days that mention them: their page, their visible employees, and outsiders. Sum the engagement on those posts. Your share of voice is your engagement divided by the total across you and them. Note the topics each competitor's posts cluster around, since a competitor owning "AI in procurement" while you own "approval workflows" is a positioning fact you should know about. And glance at tone: are outside mentions of them mostly warm, neutral, or complaints? A rough positive, neutral, negative tally across twenty posts is enough.

What to do with it: if a competitor's share jumped, find out whether it was one post, one person, or a sustained push. One viral post is noise. Six employees who started posting weekly is a program, and you'll be seeing more of it.

And what to refuse: copying their topics because those topics work for them. Their baseline isn't yours. The monthly look is there so you don't get surprised, and it shouldn't be where your editorial calendar comes from.

If you pay content partners

Some B2B companies pay people with an audience in their market to post about them on a recurring basis. Few talk about it publicly. If you don't do it, skip this section, and don't take its existence as advice to start.

If you do, the arrangement usually specifies a cadence (say, two posts a month mentioning the product) for a monthly fee. The most common failure is mundane: the posts quietly stop happening. The partner gets busy, month two has one post, month three has none, and nobody notices until renewal, because nobody's job includes counting.

So count. A tab per partner: agreed cadence, posts delivered this period, shortfall, days since the last post. Three states are enough. Delivered the agreed number: fine. Under target but posted recently: slipping, send a friendly note now. Silent for two full cycles: the arrangement is effectively off and needs a conversation about whether it continues.

Then look at what each partner delivers beyond the count, with the same lens you use for employees: engagement relative to your own average, and above all who reacted. A partner with a big audience of the wrong people is an expensive way to collect likes. You know what you pay each one. Put the delivered side next to that figure and make the call yourself. No formula makes it for you, because the value of a partner who reaches the right 50 people depends on your deal size.

The parts that stay hard

Legal and comms will have opinions, and you should ask for them before launch. If they find out in week four, they'll have more of them and less patience. What usually satisfies them is a one-page guide: no customer names without written permission, no financial figures that aren't already public, no roadmap dates, no commentary on a competitor's outage or layoffs, and a named person to ask when unsure. In regulated industries (financial services especially) some posts may need pre-approval, and that will slow everything down. Accept it and keep the volunteer group smaller. If your company is public, quiet periods apply to employees' posts too.

People leave, and their audience goes with them. The AE with the best posts in your program is building an asset she owns. That's the deal, and it's a fair one. It is also the reason a program with eight voices is sturdier than one built on a single star.

Attribution is the hardest part, and the place where overclaiming does the most damage. If someone at a target account reacted to three of your people's posts in September and their company entered the pipeline in November, the posts influenced that deal in some way you can't quantify. They didn't source it in any sense a CFO would accept. The buyer also saw an ad, got a referral, and met your AE at an event. LinkedIn's own B2B Institute argues that 95% of potential buyers aren't in the market on any given day, which means most of what a post does is make a future purchase slightly more likely, invisibly, months ahead.

So report it as what it is. "Fourteen accounts now in pipeline had people engaging with our posts before the opportunity opened" is a true sentence, and sales leadership can weigh it. "LinkedIn generated $400k" is a sentence that will get you taken apart in the first meeting where someone asks how you know. The only defensible rule we know of is about timing: count an account as influenced only if the engagement came before the stage change. Engagement after a deal opens is nice and shouldn't be counted, since people look up vendors they are already talking to.

Where the spreadsheet gives out

Everything above works by hand for about eight voices and one brief a month. Then the arithmetic catches up with you. Eight people posting weekly is around 35 posts a month. Reading each one for brief coverage, opening each reactor list, classifying perhaps 1,200 people, recomputing topic averages, redoing the competitor count: that is well past the two hours a week the owner has. What gets dropped first is the classification of who reacted, the part that was producing the evidence sales cared about.

This is the work Resonue does, and here is specifically what it covers.

It keeps the inventory as a roster of Voices in the four types described at the start: company page, employees, external voices and content partners, tracked from their public profiles. Each voice gets a verdict line with its posting cadence, its engagement compared to your workspace average, how many engagers its posts captured, how often its posts mention your brand, and the pipeline its engagers' accounts touch.

Briefs live on a calendar with a start date, an end date, and a HIGH, MEDIUM or LOW priority. When posts come in, AI matching marks which ones covered which brief, and each brief shows how many engagers its posts drew, so the "covered or not" tab and the per-brief count fill themselves in. Current briefs also show up when a person opens Compose to write.

In Compose, a draft gets a score from 0 to 100 before anyone posts it, broken into topic fit, hook strength, format fit and tone match. Topic fit is judged against your workspace's own pooled post history, all voices together. It isn't a generic playbook. For the person staring at a blank box, Rosa (the assistant inside the product) can generate a starting draft for that one voice, built from that person's own top posts and the topics they keep returning to, anchored to a brief if you choose one. It happens on demand, one voice at a time, and nothing is ever published on anyone's behalf: the person edits it, or throws it away, and posts from their own account.

Each employee with a seat has a personal view showing their own last 30 days against the 30 before (posts, engagement, people reached) along with the current briefs. There is no ranking on it.

Topics and hashtags are tiered S, A, B and C against your own baseline, with the small-sample correction described earlier built in. Share of voice, share of topic and share of sentiment are computed against the competitors you choose to track. For content partners, you enter the cadence and dates of the arrangement, and each partner shows as Honored, Slipping or Off contract, with the shortfall and days silent. It shows what each partner delivered, and the judgment about whether that was worth the fee stays with you.

The pipeline view applies the timing rule from the previous section: an account counts as influenced only when its first tracked engagement predates its stage change. Stages are moved by hand on a board.

If you'd like to see this on your own company and not a sample workspace, talk to sales. The demo is prepared beforehand on your company's recent posts, so the first thing on screen is your own voices and who has been reacting to them.