Employee-Led Content: The Whole System

I built this system for a company that sells data infrastructure to scientists. Scientists check claims for a living, so marketing bounced off them. The only thing that landed was proof written by the people doing the work. That constraint turned out to be a gift. A system built for the hardest audience ports to every easier one.

Most companies treat content as a brand function: one marketer, one company page, paid reach on top. I believe the company page is the weakest publishing asset a company owns, and the employees are the strongest one. This piece is the whole system for acting on that belief. It runs at three people and at ten thousand. It ships with the gates, the templates, and the numbers that kill it when it does not work.

A person out-reaches the page

Three reasons live in how the platforms work, and one lives in how people work.

Feeds rank people above brands. A person with 3,000 followers usually reaches more of the right readers than a company page with 30,000. That claim is testable inside a month. Compare your page analytics against one employee posting at the same rate, and look at who showed up.

Trust attaches to a person. A reader sees that the writer did the work and decides the company behind them is worth a call. A homepage cannot produce that feeling, because a homepage contains nobody to judge.

One page speaks in one voice, and ten people speak in ten. Your data engineer gets read by data engineers, and your designer gets read by designers. The gain is coverage, and most companies miss it while they chase follower counts.

Writing in public sharpens the thinking. Explaining your work to a stranger forces you to find the holes in it, and that payoff arrives months before any external result does.

Your real audience is smaller than your follower count

Count the people who can approve a purchase of what you sell. For the company I built this for, the honest count was a few thousand worldwide. The working target was about 800 of the right people knowing the team by name. Your number differs, and the exercise is the same. List the roles that sign, estimate the population, and write the count down.

That count changes everything downstream. A post read by 50,000 people outside your market is worth close to nothing. A post read by 300 people at accounts on your sales list is the entire game. The measurement section below is built on this arithmetic, and every vanity metric dies by it.

Hiring is the second engine, and most teams forget to count it

You compete for talent against companies with bigger budgets. You cannot outbid them. You can be more interesting to read, and people choosing where to work read what a team publishes. The building-in-public format exists for exactly this reader.

Count hiring inside the return. A pipeline-only measurement undercounts the program and produces the wrong call on day 90. A strong candidate who arrives pre-sold is worth as much as a warm lead.

The system is five layers, built in order

Teams usually start at layer four. They pick a scheduling tool, post for a month, and it fades. Nothing anchored it at layer one, and nothing at layer five caught what it produced. Build them in order.

Layer one is the company story

Everything downstream inherits from this layer. If it is vague here, every post is vague. Three things get written down and agreed.

Positioning takes one paragraph: what you do, who for, and why it is hard, in plain words with no adjectives. Write the unglamorous version. The company I wrote this for makes messy scientific data usable by machines, and naming that plainly beat every polished alternative.

The audience gets named specifically enough that you can list real companies, then real roles inside them. Write the list.

The pillars come last, three to five of them. A pillar is a theme where you hold evidence other people do not, and that evidence bar is the whole test.

Pillars differ by company, and the shape holds everywhere. A logistics company owns what breaks in cross-border shipping because it eats those failures daily. A dev-tools company owns the migration stories its support queue absorbs. A design agency owns the before-and-after of real engagements. In every case the pillar sits on evidence the company paid for.

Four or five pillars is the right number. Two gets repetitive by month two. Eight means you own none of them.

Layer two is the people

Each voice builds a small strategy inside the company one.

Each voice picks two or three company pillars. Depth on a few beats a touch on all five.

One personal pillar goes on top, and most companies skip the one part that keeps the program alive. The personal pillar is something real that belongs to the person: the move from one field to another, a craft, a sport, first-time management. It is the reason a reader follows the person rather than skimming another corporate feed. It also stops the writing turning into a chore, and chores get dropped around week five.

Every voice names the one or two readers they write for, people they can picture. Everyone in our industry is nobody.

Then build a coverage map. List the personas that matter to your sale, and check that every persona has a voice writing for it. Senior practitioners pair with the buyers who resemble them, engineers with the technical evaluators, founders with the executives. Overlap is fine. Two people writing the same post wastes effort. Two people covering one pillar from different angles is the goal.

Layer three is the company page, and it follows

The company page supports the people. It does not lead. Four jobs, most valuable first.

• Reshare team posts within a day. Free distribution, no extra work.
• Comment on team posts as the company. It adds a second surface to the post and shows the company stands behind the person. The comment must say something, and a brand account dropping two words of praise is worse than staying quiet.
• Say the things only a company can say: funding, launches, releases, partnerships, hiring, events, anything carrying legal weight.
• Be the archive. Someone who finds you through a person will check the page, and the page must look alive.

Customers already named publicly on your own website are safe to mention. Anything not already public is not. The gates below enforce this.

Layer four is the content engine

The blank page kills these programs faster than the calendar does. Someone sits down on Friday, nothing comes to mind, and they skip the week. Then they skip the next one. A shared library of raw material is the fix, and the whole team feeds it.

• Customer and prospect calls, fed by sales: real questions, real objections, the words people use.
• Support and onboarding questions: what confuses smart users.
• Hard edge cases, from the team that solves them: the specific, odd problems you handle.
• Benchmarks and internal experiments, from engineering and research: numbers with the method attached.
• Failed attempts, from everyone. What you tried that did not work is often your best material.
• Engineering decisions: why you chose one approach over another.
• Reading: papers, posts, and books, with your opinion attached.
• Conference notes: what you heard, and what you disagreed with.
• Old internal docs: explanations already written and still internal.

If you explain something twice inside the company, put a note in the library. Two internal explanations means the outside needs it too.

Put the library where the team already works, a wiki page or a chat channel with a tag. Do not stand up a new tool, because new tools go unused.

The process is seven steps, and most posts take between half an hour and an hour of active work.

• Pick an idea, five minutes, from the library or from the week you just had.
• Draft for twenty to thirty minutes. Ugly first draft, no polish yet.
• Sharpen for ten minutes. One claim per post, and cut the adjectives.
• Run the gates. The reviewer clears the check inside 24 hours.
• Add a visual if it helps. Text alone is fine and often better.
• Publish, two minutes, from your own account.
• Reuse later. The program owner turns good posts into blog posts, docs, decks, and talks.

The process depends on two commitments. Review finishes inside 24 hours, because a four-day turnaround stops people writing, and that clock belongs to the reviewer. And the author presses publish, from their own account, every time.

Ten formats work on a skeptical audience, in rough order of power.

• The walkthrough: how you solved something, step by step. Method is what experts trust.
• The teardown: take something public and show what is broken in it. Concrete, and recognizable on sight.
• The cheat sheet: a reference table people save. Saves are a strong signal of value.
• The failure story: you tried it, it broke, here is why. Rare, and it sticks.
• The benchmark: a number with the full method behind it. Experts check the method before the number.
• The translation: your field explained to the adjacent field. Almost nobody does this well.
• The stake in the ground: an opinion you will defend. It attracts the right people and repels the wrong ones.
• Building in public: choices and their trade-offs. Your strongest hiring material.
• The field note: an event or paper recap with a real position in it. Quick to write, and it shows you are current.
• The human one: career, team, learning something hard. It makes you a person rather than a feed.

Six shapes read as noise to a serious audience, so they stay out.

• Press releases dressed as posts.
• Announcements with nothing behind them.
• Generic news commentary with no position of your own.
• Motivational content.
• Posts that end on a question planted to pull replies.
• Reposted statistics with no opinion attached.

Layer five turns attention into pipeline

Attention that goes nowhere is a hobby. This layer turns it into revenue.

Most of your signal arrives soft. A buyer at a target account views a profile, leaves one careful comment, or follows three of your people in the same week. In a considered-purchase market that is interest. A team that counts only form fills declares failure while the program works.

Five rules keep this layer working.

• The first reply comes from the person they engaged with, never from a rep they never met. Handing a warm exchange to a cold sequence burns the trust the post earned.
• Give before asking. The next step is a doc, a dataset, or an answer, never a calendar link.
• Offer a meeting without pushing one. A prospect who is not ready gets usefulness and patience.
• Log it in the CRM. Otherwise the program looks like it produced nothing and gets cut at the next budget review.
• Ask how they heard about you on every inbound form and every first call. Social attribution is broken, so the self-reported source is the most honest number you get.

Three tiers of people run it

Posting is opt-in. Forcing it is how these programs collapse.

• Voices: five to eight people who want it. Two posts a week, a persona to own, the weekly session. About two hours a week.
• Contributors: ten to twenty people. One post a month and one real comment a week. About an hour a month.
• Amplifiers: everyone else. React or comment when you have something to add. Minutes.

Tier one runs on volunteers. Someone pushed into it writes posts nobody wants to read, including them.

A founder running this alone is the smallest honest version, and it caps early. One voice covers one persona, a vacation stops the cadence, and the audience follows the account rather than the company. Five voices survive all three.

Tier three has a bad version, where everyone drops the same two-word praise on every post within ten minutes. Readers spot it at once, and platforms rank low-value comments down, so it costs reach as well as credibility. A comment must add information, disagree with something, or ask a real question. When none of those apply, react and move on.

Ownership splits six ways.

• The executive sponsor protects the time, clears blockers, and posts personally.
• The program owner runs the library, the weekly session, the review clock, the metrics, and onboarding.
• Each voice owns their account, their ideas, their words, and their publish button.
• Rotating reviewers own the accuracy check, inside 24 hours.
• Legal and leadership own only the named escalations in the gates.
• Sales and customer success feed the library and handle warm handoffs.

Where leadership does not post, the program fades within a quarter, because everyone reads it as optional and they are right.

The cost is two hours a week per voice, and hiding it kills the program

Two hours is about 5 percent of a working week. Across six voices that is twelve hours a week the company is not spending on something else. Say it plainly before anyone signs up, and if you cannot protect the time, do not start. A program run at half commitment produces posts nobody reads and teaches the team the whole thing was theater.

The usual failure is a good person who gets too busy for five weeks running. Three things prevent it. Managers agree to the time before their person joins tier one. The time goes in the calendar, same slot every week. And it is never the first thing dropped when a deadline moves.

A weekly rhythm that fits inside those two hours:

• Monday, five minutes, everyone: drop the week of raw material into the library.
• Midweek, an hour, voices: write and post.
• Midweek, twenty minutes, reviewers: clear the review queue.
• Later in the week, an hour, voices: second post, and replies to comments.
• Friday, thirty minutes, owner and voices: what worked, and what runs next week.
• Monthly, forty-five minutes, owner and sponsor: read the metrics and drop what is not working.

Writing the post is half of it. Replying to comments is the other half, and it is the half people skip. The comment threads are where trust gets built and where buying signals show up first. Budget fifteen minutes after posting and reply to anyone who said something real.

Five gates keep it safe without killing it

Five checks stand between a draft and the feed. Most posts trigger one or two, so the gates run as a filter.

Gate one is accuracy. Every post that makes a factual or technical claim gets checked by someone who did not write it. The check hunts wrong facts, findings stated past the data, methods that do not support the conclusion, and citations that say something else. Slower and right beats faster and wrong. One bad claim undoes more than several good posts build.

Gate two is client exposure, and the default answer is no. Four bans: a client name that is not already public, indirect identification, client data or screenshots or results, and anything about an active deal. Naming a client needs their written approval, and verbal approval does not count. When unsure, describe the shape of the problem with no company attached.

Gate three is legal exposure. Unpublished methods, patent-adjacent work, and anything under a contract or an embargo go to a senior lead, who escalates to counsel when unsure. Ask before drafting. The question costs nothing up front and a lawsuit at the end.

Gate four is claim backing. A number needs a source you can show when challenged. Industries with regulators carry an extra rule here: no medical, financial, or legal claims your license does not cover. Validated, proven, best, only, and first are promises, and a promise you cannot defend in a room of experts comes out. Do not compare yourself to named competitors. Compare to the problem instead.

Gate five is voice, and it stays weak on purpose. Tone is the only thing checked, and opinions pass untouched. Over-editing is the most common death for a content program. Marketing smooths every draft into the same flat register, the author stops recognizing their own writing, and they stop writing. Fix errors, flag risks, leave the voice alone. A rough post that sounds like a person beats a polished one that sounds like nobody.

On AI, three rules hold. The idea comes from the person, and AI tightens rather than decides. The person reads and approves every word before it goes out. Nobody posts from anyone else's account. A wrong claim damages the name at the top of the post, a professional reputation someone spent years building. That reputation belongs to the person alone.

A good post passes four checks

• One claim, stated once.
• One specific detail: a real number, name, dataset, or example.
• Useful on its own, to someone who will never buy from you.
• Yours alone. A competitor cannot publish it word for word.

Four skeletons pass all four checks. The work goes in the brackets.

I pulled [N] public [things] to check how consistent they were.
[Field] appeared [N] different ways. The worst of them: [...].
This is why [the easy assumption] is harder than it sounds.
How we handle it: [...]

We spent [time] building [thing]. We threw it away.
The assumption that broke it: [...].
What we do instead now: [...].
The part that generalizes: [...]

If you run [adjacent function] and someone handed you [your domain]
this morning, here is the short version. [Three to five plain points.]
The part that will surprise you: [...]

We tested [A] against [B] on [dataset].
Method: [...] so you can check it.
Result: [...].
Where it does not hold: [...]

All four open with something concrete and explain afterwards. All four include the part where it did not work, which is the part a serious reader looks for first.

Measure who showed up, and let every other number go

Impressions, likes, and follower counts are not success metrics here. Say that out loud at kickoff, so nobody optimizes for them. Chasing reach turns a focused feed generic inside two months.

The one number that tracks pipeline is the ICP engagement rate. Take everyone who engaged with your posts in a month and look up where they work. Count the share at companies on your sales target list, in relevant roles. If the company is off the list, the engagement does not count, however senior the person. That rule exists to stop you moving the goalposts later.

Do it by hand in a spreadsheet. It takes about an hour a month and needs no tooling. Working out who engaged takes longer than counting how many, and it is the only version that predicts anything. Forty engagements with twelve from target accounts beat nine hundred with four.

The readers who matter most never show up in either count. In a considered market the serious buyer reads for weeks and says nothing, then mentions three posts on the first call. Silent readership is invisible to every dashboard, which is one more reason the source question on the first call carries the program.

Five weekly signals say the machine runs.

• Posts published per voice.
• Review turnaround against the 24-hour commitment.
• Comments and follows from target accounts.
• Conversations started in direct messages.
• Items added to the library.

Five monthly signals say it works.

• Inbound meetings where the person names the content as the source.
• Deals where a contact engaged before the first call.
• Job applicants who mention it.
• Speaking invitations.
• Branded search volume.

Set expectations by month, in writing, before day one.

• Month one: almost nothing. Small numbers are the expected result.
• Month two: the right people start following, and comments get better.
• Month three: first real conversations, and the first time a call opens with your posts.
• Months four to six: inbound steadies, candidates mention it, and sales starts forwarding posts.
• Month six and beyond: it compounds. Old posts keep working, and people arrive already trusting you.

Week six is the dangerous point. The effort is real, the results are invisible, and someone proposes dropping the program. The written 90-day commitment is what keeps that conversation short.

Write the kill criteria before you start

A program without kill criteria never dies. It gets quieter until nobody remembers who owned it. Set the bars now and read them on day 90. The four I use:

• Share of engagers from target accounts: at least one in five.
• Posts published per voice during the pilot: at least four.
• Review turnaround breaches over 24 hours: no more than two.
• Conversations traceable to a post: at least one.

Miss any two of those and stop. Miss one and fix that single thing, then run thirty more days. Hit all four and expand.

These bars are judgment, and the one-in-five share was my starting guess for one company in one market. Argue your own numbers before day one, because a bar set on day 90 bends to whatever happened. What does not bend: the bars exist in writing before the first post.

The pilot runs 90 days in four gated phases

Days 1 to 15, foundation. Confirm positioning, the target list, and the pillars. Recruit the voices, volunteers only, with manager sign-off on the two hours. Each voice picks pillars, a personal pillar, and a persona. Build the coverage map. Seed the library with thirty ideas. Name the reviewers and agree the 24-hour turnaround. Record the baseline: follower counts, inbound meetings per month, inbound applicants per month. The phase is done when one test post clears the review path end to end.

Days 16 to 45, pilot. Each voice posts once a week. The weekly session runs without exception. The company page reshares within 24 hours. Add the source question to inbound forms and first calls. The phase is done when every voice published at least four posts and the turnaround held.

Days 46 to 75, scale. Voices move to two posts a week. Contributors come aboard. The company page starts commenting with substance. Routing runs end to end, from signal to the right person replying. Sales gets briefed on warm handoffs. The three best posts become longer assets. The phase is done when routing worked on a real lead.

Days 76 to 90, decide. Read the four numbers against the baseline. Cut the formats and channels that did not work. Write the winning patterns into a starter kit for the next group. Then expand, hold, or stop, on the numbers, in writing, with a name against the decision.

The same system runs at three people and at ten thousand

Six invariants never change with size, and everything else scales.

• The author presses publish.
• Posting is opt-in.
• Review clears in 24 hours.
• The kill criteria exist in writing.
• Every voice keeps a personal pillar.
• The ICP engagement rate is the headline number.
• Three to ten people: the founder is the sponsor, the owner, and the first voice. One or two voices total. Review is one peer reading the draft in chat, the library is a single channel, and the weekly session is ten minutes.
• Eleven to a hundred: a named owner separates from the sponsor. Three to six voices, one rotating reviewer, the system as written but trimmed to fit.
• A hundred to a thousand: the full system as written. Five to eight voices, the coverage map, rotating reviewers, CRM routing.
• Past a thousand: pods of five to eight voices per business unit, each with its own coverage map. One central owner keeps the bars, the gates, and the rules identical across pods. Legal gets a named lane in gate three, because the blast radius grew.

A three-person company skips the coverage map and the tiers, and it skips nothing else. The founder writing one honest teardown a week, gated by one peer, is the same machine at its smallest size.

The risks are known, so plan for them

Every one of these has a counter already built into the system. The list exists so nobody discovers them at week six.

• People get busy and stop. High odds. Manager sign-off before joining, calendar blocks, a full library, and the weekly session.
• A wrong claim goes out. Medium odds. Gate one, checked by a non-author, and a fast public correction when it happens anyway.
• Client information leaks. Low odds. Gate two, default no, written approval only.
• Review becomes the bottleneck. Medium odds. The 24-hour clock belongs to the reviewers, and breaches count against the kill criteria.
• Over-editing kills the voice. High odds. Gate five stays weak on purpose.
• Engagement looks staged. Medium odds. No praise pods, and every comment has to add something.
• One person dominates. Medium odds. The coverage map, and rotation in who gets amplified.
• Reach drifts off-audience. Medium odds. The ICP rate is the only headline number.
• The program gets killed at week six. High odds. Ninety days agreed in writing before the first post.
• A key voice leaves. Medium odds. The system is built to outlast any one person, and the next paragraph prices it.

The honest objection to the whole system is that building people's public profiles helps them leave. Some will. That cost is real, so name it instead of waving it away. Four reasons to run it anyway. The library holds the raw material and the formats, so the system outlasts any one person. Five voices means no single departure ends it. People who leave well stay advocates, and some come back or refer. And the alternative is a team whose expertise nobody outside can see, which is worse for hiring and worse for sales. A company where nobody wants to build a public reputation has a larger problem than turnover.

The rules fit on one page

Do:

• Write about work you did.
• Be specific, with real numbers, datasets, and examples.
• Share what failed as well as what worked.
• Hold an opinion you will defend.
• Reply to every real comment, and credit teammates by name.
• Say you do not know when you do not know, and correct yourself in public when wrong.

Do not:

• Name a client without written approval, or share client data or results.
• Post a number you cannot source.
• Make claims a regulator licenses, in a regulated industry.
• Compare yourself to named competitors, or post about an active deal.
• Use validated, proven, best, or first without evidence.
• Post from someone else's account, comment for the sake of commenting, or end on a question planted to pull replies.

When unsure, ask in the review channel before drafting. Two minutes there saves an hour later.

The templates, ready to copy

A new voice starts with two weeks of setup. Week one: rewrite your headline to say what you do and who you help, because a job title says neither. Rewrite your profile summary in plain first person. Pick two or three company pillars and one personal pillar. Name your reader. Follow thirty people in your space and comment with substance on five posts.

Week two: post something you learned or fixed this week, small is fine, then post your pick from the formats list. Reply to every comment on both. Your first ten posts will not be good, and that does not matter. Getting the reps in early matters more than getting them right.

The pre-publish checklist:

• One claim only.
• At least one specific detail.
• Useful to someone who will never buy.
• A competitor cannot publish it word for word.
• No client named or identifiable without written approval.
• Every number has a source.
• No regulated claims.
• A technical claim got checked by someone who did not write it.
• No hype words.
• No closing question planted to pull replies.
• Sounds like me.

The weekly session runs thirty minutes. Five go to what went out, ten to one post examined closely, ten to next week from the library, five to anything stuck. A meeting that runs long is a meeting people start skipping.

IDEA:   [one line]
SOURCE: [call / bug / paper / benchmark / question]
PILLAR: [which one]
FOR:    [which reader]
DETAIL: [the specific thing: number, name, dataset, quote]
GATES:  [which checks this is likely to trigger]

Objections, answered

I am an engineer, not a marketer. Good. The system wants an engineer explaining something to another engineer. A post that reads like marketing failed.

I do not have time. Two hours a week. Managers agree to it before anyone joins tier one, so the time exists before the writing starts.

I have nothing to say. You have more than you think, and it feels obvious to you because you know it. Anything you explained to a colleague this month works. The library exists for the days this feeling shows up.

I am worried about being wrong in public. Gate one exists for this. When something slips through anyway, correct it in public. Serious readers respect a visible correction, and it usually builds more credibility than the error cost.

This sounds mandatory. No. The writing tiers are opt-in. Everyone else engages when they have something to add, and stays silent otherwise.

Competitors will copy us. They will copy the format. The evidence, the data, and the people do not copy. A post a competitor can reproduce word for word was too generic to publish.

What is in this for me. A reputation that is yours and travels with you. A network in your field. Better invitations. And clearer thinking about your own work, the part people underestimate until a month in.

Start here

Name the owner and the sponsor. Recruit the volunteers. Set your own four bars and put them in writing. Seed the library with thirty ideas. Date the kickoff, and hold the 90 days.

You are closer than it feels. Most of what a good post needs sits in your call notes, support tickets, and chat threads today. The work is mostly moving that material into the open.

If your market breaks the mechanism, measure it and tell me. Run one person against your company page for a month and count who showed up on each. That number decides better than either of us.