Employee Advocacy Program: The Review Queue, the Policy and Week Three
Written by Martin H.
Someone in your company is still waiting for approval on a post about an event that ended three weeks ago, and the next one never got written. How long did your last ten posts wait?
Approval is the one part of an employee advocacy program everybody budgets for once, usually by naming a committee. The part that keeps it alive in month two is a two minute budget nobody writes down, plus a stack of posts that cleared before anyone needed them. This is the employee side; the customer side, where buyers vouch for you, is a different instrument and it lives in customer advocacy. Every program I have watched fail was killed by a delay rather than by a bad idea, and the delay always looked like somebody being careful.
An employee who waits three days for a post about an event that ended on Tuesday learns that sharing your company's content is a way to be late. The ones who learn it fastest are the most willing ones, and they never tell you they stopped. Two things decide whether the program survives: a queue you never staffed and a library you never wrote.
Where an Employee Advocacy Program Actually Stalls
Start with the queue, because it is the visible one. A post is written on Monday about something that will be stale by Thursday. It enters a review path, and the reviewer is a lawyer with eleven other things on her desk, or a marketing manager wearing three jobs, or a committee that meets on Fridays. It clears on Friday. The employee posts it anyway, watches it land flat, and files away a conclusion. The second time, they do not bother. The queue did not block a post. It taught a behavior.
The regulator points the same way. Asked about pre-approving paid posts, the FTC says "you should review the posts for truth-in-advertising compliance, including any disclosure responsibilities," and where regular monitoring is too much, "you should probably switch to pre-approval of posts," because "it's much easier to review posts before they're posted than to search for them afterwards." That advice is written around paid influencer content, but the mechanism transfers. Pre-approval is sanctioned. The open question was never whether to review, only how long it takes and who owns it.
Then the 2026 layer. LinkedIn published a post on 4 June 2026 saying its systems "have been trained to recognize signals of AI slop," defined as content that "may sound polished on the surface but lacks any real unique perspective," and that such content "is less likely to be widely distributed beyond a person's immediate network." Its initial testing identified generic content correctly "94% of the time." A review built to make every post safe pushes everyone toward the safest wording, and the safest wording is now the least distributed. Speed and originality turn out to be one problem.
The second stall is quieter. A fast queue still fails if there is nothing worth posting. DSMN8's Employee Advocacy Benchmarks 2026, a survey of more than 200 programs by a vendor selling into this market, ranked the barriers in managers' own words: uncertainty about what content to share first at 18 percent, then resistance to using social media and lack of interest in company content at 16 percent each. Its reading is the line to keep: "the biggest barriers to participation are uncertainty and lack of clarity, not motivation." Two of the three top barriers are content problems.
So the queue makes sharing feel late, and the library makes it feel pointless. Approval is the visible cause. Supply is the larger one, and a fix that touches only the first looks like progress for about three weeks.
Two pages already concede that approvals matter, so this is not a discovery. Vulse scores every platform partly on "Approvals and control: Can admins review posts before they go out, and can they publish on behalf of a person with that person's consent?" Sprout Social's own example post describes the curation workflow it introduced, naming three criteria: timing, best practices and reporting needs. Neither gives the reviewer an owner, a clock and an escalation rule, and neither connects the review to how far the approved post travels.
The signal that catches this early: median time from submission to publish, per lane, plus the share of submissions approved after the event had passed. A queue whose median exceeds its own content half-life is not a control. It is a publishing ban with a form.
Three Lanes, One Owner and One Clock
Every stalled program I have looked at has one review path used for everything, staffed as if most of it were sensitive. The fix is not a better tool.
Lane | What may go in it | Who owns it | Clock |
|---|---|---|---|
Pre-cleared | Product facts, claims the company already published, hiring posts, event photos with consent on file | The library owner in marketing | No review at the point of use. The page is versioned monthly |
Standard | Commentary on company news, industry takes, a customer story with written permission | One named reviewer in marketing | Four business hours |
Sensitive | Regulated claims, earnings, litigation, personnel matters, a named customer without a signed release, anything unannounced | A named compliance reviewer with comms | Two business days |
Two details decide whether that table survives a calendar. Every lane needs a named deputy, because a clock that stops when one person takes leave is a clock nobody believes. And a refusal has to arrive inside the clock, in writing, with a reason and an offered rewrite. A late, unexplained no loses your most willing people first.
The pre-cleared lane makes the other two affordable, and it is the one nobody wants to build. Its contents are one versioned page of statements your company will stand behind in an employee's own voice: what the product does, what it costs, which logos may be named. That page is slow work done once, in advance, by someone who does not enjoy it. It is also brand voice guidelines written for the one audience that will quote them back at you verbatim.
I have assumed 70 percent of what employees want to share can be pre-cleared. That is a round number I picked, not a measurement, and it is the most important number here.
What a Review Costs, Printed in Full
A queue clears when capacity meets arrivals. It does not clear faster because people try harder, and it does not clear at all if you staff it for a fraction of what arrives.
Input | Value used |
|---|---|
Posts submitted per week to the review path | 40 |
Median content half-life, meaning how long a submitted post is still worth publishing | 3 business days |
Reviewer capacity, standard lane | 12 posts per reviewer per week |
Reviewer fully loaded cost | USD 9,000 per month, USD 108,000 per year |
Working weeks in the year used for the denominator | 46 |
```
Weekly arrivals A = 40
Share needing full review s = 1 - 0.70 = 0.30
Weekly review load L = A x s = 12 posts
Reviewer capacity C = 12 posts per reviewer per week
Load per reviewer L / C = 1.0
```
At a 70 percent pre-cleared share, one reviewer clears the queue exactly. The program is feasible with a single named owner, and the whole design question becomes whether your 70 percent is real. At a zero percent pre-cleared share the same program needs 3.3 reviewers, because 40 divided by 12 is 3.33. That is the arithmetic behind every stalled program I have seen: the queue was designed as if everything were sensitive and then resourced for a fraction of it. The pre-cleared lane never got built, so every submission arrived at the full review.
```
Reviewer cost per review = 108,000 / (12 x 46) = 196 USD per post reviewed
Cost per published post = 196 / approval rate; at 0.9 = 218 USD
```
About USD 200 per post cleared. Defensible for a post that reaches a decision maker, absurd for a post about an office event. Which is the argument for lanes: not that some posts deserve less care, but that they deserve a different price. It is also the answer to the platform question a launch meeting asks first, because a platform does not reduce the cost of a review. It reduces the cost of distributing what has already been approved. The review is people.
The Four Clauses a Policy Has to Settle
Three of the four have a regulator behind them. The fourth does not, and I will say so.
Disclosure. On whether "#employee" is good enough, the FTC's answer is: "Consumers will likely not understand '#employee' to mean that the person is an employee of your company." It recommends "I work for XYZ" or "#XYZ_Employee", and that it would be "much clearer to use the words 'my company's' or 'my employer's' in the body of the message." Put the disclosure in the pre-cleared copy rather than asking employees to invent one, because they will invent the hashtag.
What may never be said. An endorsement has to be honest and not misleading, and an employee cannot make a claim the marketer could not make. Results claims need the generally expected result, and "results not typical" does not cure it. For a regulated employer the prohibited list is not marketing's to set. PostBeyond's write-up names four bodies a program has to satisfy, the Federal Trade Commission, the Financial Industry Regulatory Authority, the US Health and Human Services Office for Civil Rights for HIPAA, and the Securities and Exchange Commission, and notes that under the SEC's social media protocols "even 'liking' a Facebook post can potentially cross regulatory lines". A policy that lists categories rather than sentences is the only version that stays current.
Who owns a mistake. This is the clause most policies get backwards. The FTC states the advertiser "would be responsible for employees who didn't comply with those instructions (or who make unsubstantiated performance claims)." The company owns it, not the employee. Say so in those terms, because an employee who believes a mistake is personal will stop posting anything interesting. If you are a listed issuer, material non-public information belongs in the sensitive lane with a named approver.
Voluntary in writing. No source I could find addresses whether an employer may require participation. So, labelled as my recommendation rather than a rule: state that participation is voluntary, that no performance review or compensation decision depends on it, and that an employee may leave at any time. A required endorsement is a different object from a willing one, and both the disclosure and the distribution problem above assume willingness.
Why Participation Falls After Week Three
No published source I found gives a decay curve or a participation rate by week. Week three is a frame, not a measurement. It is shorthand for the point where the launch stops generating its own energy and the program has to run on its design.
Supply is the unglamorous half. In the same survey of more than 200 programs, managers reported providing an average of four new pieces of content weekly. Four. If you enroll three hundred people and offer four items a week, you are running a newsletter with a share button, and the participation numbers will say so.
Effort is the half I would put on the wall. Every required step between "content offered" and "published" subtracts from the number of people who will ever do it. I use a voluntary attention budget of two minutes per share, a design assumption rather than a statistic, derived from one behavioral fact: in programs that work, 68 percent of advocates share three or more times a week, per DSMN8. Nobody sustains a repeated multi step task on an unrelated errand at that frequency. A program asking five minutes per post is asking for fifteen minutes a week of unpaid marketing work, and it will get them from the ten percent who enjoy it. The fast lane's job is to get a post from offered to published inside two minutes: copy written, link tracked, disclosure pre-worded.
Two incentives survive that, and both are supply fixes wearing an incentive's clothes.
Recognition that is visible internally. A named mention in the channel your employees already read, from a person whose opinion they care about. The honest incentive already exists: 94 percent of advocates in DSMN8's survey say posting on LinkedIn has benefited their careers. Make that visible rather than inventing a substitute.
Content the employee would have shared anyway. Sprout Social's September 2026 research found that "72 percent of engaged users would post if their teams wrote the content for them." Set that against the 18 percent whose top barrier is uncertainty about what to share. The intervention is a library, not a pep talk.
Three incentives fail, each for a specific reason. Leaderboards concentrate attention on the ten percent already posting and hand everyone else a public number that says they are losing. Quotas turn a voluntary endorsement into a required one: the disclosure question changes shape, because the FTC's pre-approval framing assumes willing endorsers, and the post reads like it was written to satisfy a number. Cash per post makes the endorsement paid, which changes what must be disclosed, prices the one thing the program relied on being sincere, and attracts fraud.
The signal across all three: the skew of posts per participant, and whether it is widening.
The Three Numbers, and the One to Refuse
Most programs report a number divided by enrolled headcount. That denominator is why the number improves every time you recruit and never means anything.
```
Participants 300
Active core (ten percent) 30
Posts from the core 216
Posts from everyone else 24
Posts per person outside core 24 / 270 = 0.09 per month
```
The 270 people outside the core are posting less than once every ten months. That is not participation in any sense a program can defend, and it is not a motivation problem a reward fixes. You are running a 30 person program with 270 spectators, and every metric that divides by 300 is lying to you. The concentration is documented on the vendor side: across 700 million LinkedIn impressions tracked by Vulse, ten percent of people generate ninety percent of them. Take that as a shape, not an industry rate.
Participation rate, over people who posted in the period rather than people enrolled. Reported over enrollees, it falls every time you recruit, which makes success look like failure.
Share-through rate: posts published divided by content items offered to a participant in the period, tracked for the launch cohort. A cohort whose share-through falls while the library also shrinks is not a motivation problem.
Attributed pipeline. The honest version is a tracked link, a UTM convention applied consistently, and a holdout comparing a group offered content against one that was not. Without the holdout you have an attribution claim, not a measurement. The method is in incrementality testing.
The number to refuse as the headline is reach. Not for moral reasons, for an arithmetic one: reach is the only number here that a program can move without changing anything a customer does. One popular employee resharing moves it. A better program does not. Put reach in the appendix of your report and these three in the summary, with the same owner and date as the rest of your marketing KPIs.
The Stop Rule, and When the Answer Is No
Most programs are reviewed rather than governed, and a review in a good quarter always concludes that the program is promising. Write the rule first. These thresholds are mine, not benchmarks.
Stop the program, rather than tune it, when any two of the following hold across two consecutive quarters.
- Share-through rate is below 25 percent of content offered, for the launch cohort. The library is not being used, and the problem sits upstream of the employees.
- Median review time exceeds the content half-life in the standard lane. The queue has become the constraint, which is a resourcing decision rather than a marketing one.
- Participation, measured over people who posted, is flat or falling while enrolled headcount rises. The program is adding spectators, not advocates.
- Attributed pipeline is not distinguishable from zero after a tracked link and a UTM convention have run for two quarters. You cannot show the work, so stop paying for it.
Pause instead of stopping when the signal is supply. Low share-through with a healthy review time is a content problem, fixable without ending anything. The rule exists for the state where both the library and the clock have been fixed and the numbers still do not move. Name the person who can apply it before launch, because a rule with no owner becomes a discussion, and a discussion in a good quarter becomes a delay.
When launching would cost more than it returns
```
Value of one published post V = 40 USD (illustrative, discounted to pipeline)
At 40 submissions a month and 90 percent approval:
review cost = 40 x 196 / 0.9 = 8,711 USD per month
value = 36 posts x 40 = 1,440 USD per month
```
At those inputs the review path costs about six times what the posts return. That is not an argument against employee advocacy. It is an argument against reviewing everything. Set the pre-cleared share to 70 percent and review cost falls to roughly USD 2,600 a month, which is a different business with the same people.
So, plainly: a company that cannot name a pre-cleared library, a named reviewer and a person who will post first should not launch a program. It should write the library first, and it will discover in the writing whether it has anything worth advocating.
Three conditions settle the question whatever the arithmetic says. A small team, because below a threshold the review path costs more than the reach it buys. A regulated business, where the program is a compliance workflow with a distribution benefit and belongs in the compliance budget: under Rule 2210 as currently written, retail communications to more than 25 retail investors in any 30 day period require principal pre-approval before use, with the principal's name and approval date recorded, and FINRA Regulatory Notice 26-14 would replace that with risk-based supervision because pre-approval "presents significant challenges for certain categories of retail communications, particularly communications on social media." No change has been adopted. And no posting leader: DSMN8 found senior executives involved in 79.5 percent of the surveyed programs, while the Hinge Research Institute's employee advocacy study found more than 25 percent of firms say no specific department owns employee social. A program without a named owner and a leader who posts first launches into a vacuum.
The version that works is smaller than the version that gets pitched. One reviewer, a page of pre-cleared statements, three lanes, four clauses and a rule written before anyone was invited. That is a document and an afternoon, not a platform and a year. Which is why the question for your next planning meeting is not which employee advocacy program to buy, but whether you will write the library that makes one worth joining.
Appendix: The Four Clauses, Copy-Ready
Bracketed items are the only places that need editing.
1. Disclosure. "[COMPANY] requires every employee who shares content about the company, its products or its customers to disclose the employment relationship in the post itself. Acceptable forms are 'I work for [COMPANY]', '#[COMPANY]_Employee', or the words 'my company' or 'my employer' in the body of the message. A standalone '#employee' tag is not sufficient, because readers do not reliably understand it to mean an employment relationship. Where copy is provided through the program, the disclosure is pre-written and must not be removed."
2. Prohibited claims. "No post may make a claim [COMPANY] could not make in its own marketing, including any claim about results, performance, savings or comparative superiority. Where results are described, the generally expected result must be stated, and 'results not typical' is not an acceptable qualifier. [IF REGULATED] The following categories are prohibited without written approval from [COMPLIANCE ROLE]: any statement about [REGULATED CATEGORY 1], any statement about [REGULATED CATEGORY 2], and any reference to a named client or prospect."
3. Error ownership. "Where an employee shares content in good faith following the pre-cleared library or an approved submission, responsibility for the content rests with [COMPANY] and not with the employee. [COMPANY] will correct, retract or remove the content and will not treat the matter as a performance or disciplinary issue. This clause does not cover content the employee publishes outside the program, or content that departs from approved copy. Escalation goes to [NAMED OWNER] at [CONTACT]."
4. Voluntariness. "Participation in [PROGRAM NAME] is voluntary. No performance rating, promotion decision, compensation decision or informal expectation depends on participation. An employee may withdraw at any time by [WITHDRAWAL MECHANISM], after which access to the pre-cleared library is removed and no further content is offered. Withdrawal carries no consequence and requires no explanation."
Version the four clauses with a date and a named owner at the top. A policy without a version is a rumor.
Frequently Asked Questions
How long should it take to approve an employee's social post?
It depends on the lane, and that is the point of having lanes. A pre-cleared post needs no review at the point of use. A standard post, meaning commentary on company news, an industry take, or a customer story with written permission, should clear in four business hours. A sensitive post, covering regulated claims, earnings, litigation, personnel matters or a named customer without a signed release, gets two business days with a named compliance reviewer. Anything slower than the clock in the sensitive lane means the lane is understaffed, not that the post was complicated.
Does the FTC require pre-approval of employee social posts?
The commission does not state a requirement in the guidance read for this article. It recommends the practice, and its reason is practical: "It's much easier to review posts before they're posted than to search for them afterwards." It also says that if regular monitoring is too much, "you should probably switch to pre-approval of posts."
What is a good participation rate for an employee advocacy program?
I could not find a defensible published benchmark, so I will not give you one. The figure most often quoted in this field travels without a named source or a denominator, and a rate with no denominator cannot be managed. Build your own, and report participation over people who posted rather than people enrolled. If you want one number to watch, use share-through rate for the launch cohort.
Can we require employees to participate?
No source I found addresses whether an employer may require participation, so this is a recommendation rather than a rule. State in the policy that participation is voluntary, that no performance or compensation decision depends on it, and that an employee can leave at any time. A required endorsement is a different object from a willing one.
