borghei/Claude-Skills / tools/linkedin/linkedin-employee-advocacy
linkedin-employee-advocacy
VerifiedDesigns and runs an employee advocacy programme: who posts what and how often, review rules, and what is never scripted. Use when getting a team posting, writing an advocacy policy, or fixing a programme that has stalled.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 14.9 KB |
| assets/advocacy_charter_template.md | markdown | 5.3 KB |
| assets/roster_template.json | config | 1.3 KB |
| assets/sample_roster.json | config | 2.7 KB |
| references/governance-and-consent.md | markdown | 12.1 KB |
| references/measuring-the-programme.md | markdown | 9.3 KB |
| references/programme-design.md | markdown | 11.8 KB |
| scripts/advocacy_planner.py | script | 13.8 KB |
| scripts/advocacy_rules.py | script | 10.9 KB |
references/measuring-the-programme.md
View on GitHub# Measuring the Programme
What to measure, at what level, and how to judge the result without borrowed
benchmarks. This file contains no industry figures on purpose: published
advocacy statistics are rarely comparable to your team, your audience or your
sector. The reference point is your own trailing history.
## Contents
1. Three questions, three levels
2. What not to measure
3. Building a baseline from your own history
4. Reading a change
5. Outcomes and honest attribution
6. Programme health indicators
7. Review rhythm
8. Decisions the numbers should drive
9. Measurement checklist
## 1. Three questions, three levels
Keep the levels apart. A report that mixes them answers none of the questions.
| Level | Question it answers | Measures | Who sees it |
|-------|---------------------|----------|-------------|
| Programme health | Is this sustainable and voluntary? | Members active, plan versus actual cadence, hours spent, review turnaround, pauses and leavers | Owner and members |
| Audience | Are the right people seeing and responding? | Team-level impressions and engagement from the product's own analytics, and the mix of who engaged | Owner and members |
| Outcome | Did anything the business cares about change? | Conversations started, introductions, applications, meetings where a post was mentioned | Owner and sponsor |
Individual post numbers belong to the individual. Each member can see their
own in the product. The programme collects team totals that members choose to
share, and never requires screenshots of personal analytics.
## 2. What not to measure
| Tempting measure | Why it harms the programme | Use this |
|------------------|----------------------------|----------|
| A leaderboard of members by reach or engagement | Turns a voluntary activity into a ranked one; rewards existing audience size, not contribution | Team totals and a note of thanks for specific posts |
| Posts per person as a target | Counts effort; produces filler | Plan versus actual at team level, as a sustainability check |
| Follower counts | Move slowly, depend on tenure, and say nothing about who the followers are | Audience mix on a sample of posts |
| Likes from colleagues | Measures internal politeness | Engagement from outside the company |
| An invented money value per impression | Produces a large number nobody believes | Counted outcomes with a sentence on how each was attributed |
| Comparison between members | Audiences differ by an order of magnitude for reasons unrelated to the programme | Each member against their own previous period, privately |
## 3. Building a baseline from your own history
Before launch, or as early as possible, record where the team starts.
1. Choose a window. Eight weeks is enough to smooth one unusual post; four is
the minimum.
2. For the members who are willing, collect for that window: number of posts,
and total impressions and engagements as shown by the product.
3. Sum to team level. Record the count of members contributing figures, since
the total means nothing without it.
4. Compute per-post medians, not means. One post that travelled far will
distort a mean for months.
5. Write down what was unusual in the window: a launch, a conference, a
holiday period.
A baseline sheet needs only these columns:
| Window | Members reporting | Posts | Median impressions per post | Median engagements per post | External share of engagement | Notes |
|--------|-------------------|-------|-----------------------------|-----------------------------|------------------------------|-------|
| Eight weeks before launch | | | | | | |
| Weeks 1 to 8 | | | | | | |
| Weeks 9 to 16 | | | | | | |
"External share" is the proportion of engagement that came from people
outside the company. It is the single most useful audience figure, because a
new programme's early engagement is mostly colleagues.
## 4. Reading a change
A difference between two windows is only interesting if it is larger than the
windows normally differ from each other.
- **Compare like with like.** Same members, same window length. If the roster
changed, compare only the members present in both.
- **Look at the spread, not just the middle.** If the median is flat but the
weakest posts improved, the programme helped the people who needed it.
- **Distrust small counts.** Twelve posts in a window is twelve observations.
A move that depends on one of them is not a trend.
- **Wait for two windows.** One better window is an observation. Two in the
same direction is the beginning of evidence.
- **Name the confounders.** A launch, a funding announcement or a single
widely shared post will move every number. Say so in the report instead of
crediting the programme.
A plain way to state a result: "Median engagements per post were 14 before
launch and 19 and 22 in the two windows since, across the same six members.
The second window included the product launch." That sentence is more useful
than a percentage.
## 5. Outcomes and honest attribution
The outcome level is where programmes are defended, and where the temptation
to overclaim is strongest.
Count things that happened and can be pointed to:
- an inbound message that mentions a post;
- a prospect or candidate who says in a call that they had been reading a
team member's posts;
- an application whose source field or cover note names a person;
- an introduction made in a comment thread;
- an invitation to speak or write.
Record each in a simple log with the date, what happened, which member, and
how you know. "How you know" is the column that keeps the log honest.
What cannot be claimed: that a deal closed because of a post, or that a given
number of impressions is worth a given amount. A post is one touch among many.
Report it as an assist, in words.
Ask the question at the source. Adding "what prompted you to get in touch?"
to the first sales call, the application form and the interview is the
cheapest attribution system there is, and the only one that captures what
happened outside tracked links.
## 6. Programme health indicators
These tell you whether the programme will exist next quarter.
| Indicator | Healthy sign | Warning sign | First response |
|-----------|--------------|--------------|----------------|
| Plan versus actual, team level | Within reach of plan most weeks | Persistently well under | Lower the plan; ask what is hard: time, subjects or permission |
| Members active in the window | Steady, with ordinary gaps for leave | Shrinking each window | Talk to people who paused; do not replace them with conscripts |
| Hours actually spent | Near what was agreed | Well over | Recalibrate the time model; improve the weekly note |
| Review turnaround | Within a working day | Drifting past it | Add checker time or narrow the triggers |
| Share of posts reviewed | Stable and a minority | Rising toward all posts | Triggers are being applied as "review everything"; retrain checkers |
| Similarity between members' posts | Distinct voices and subjects | Convergence on the same announcements | Stop sending copy; split crowded topics |
| Pauses and leavers | A few, for ordinary reasons | A cluster after a specific event | Find out what happened and fix that |
## 7. Review rhythm
| When | Who | Looks at | Produces |
|------|-----|----------|----------|
| Weekly, ten minutes | Owner | Review queue, anything stuck, what goes in the next note | The weekly note |
| Every four weeks | Owner with members | Plan versus actual, hours, topics, what was hard | Adjusted roster and re-run plan |
| Every eight weeks | Owner | The baseline sheet, external share, outcome log | A one-page readout for the sponsor |
| Quarterly | Owner and sponsor | Health indicators, the governance audit, whether the goal is still right | Continue, change or stop |
The readout for the sponsor fits on one page: what the team published, who
responded (in segments, not names), what happened as a result (from the log),
what was hard, and what changes next.
## 8. Decisions the numbers should drive
A measure that changes no decision can be dropped.
| Observation | Decision |
|-------------|----------|
| Engagement is mostly internal after two windows | Review topics and audiences; the posts may be written for colleagues |
| One member accounts for most of the team total | Protect their time; do not present the programme as broader than it is |
| Plan versus actual is low for one role family | Lower that family's starting cadence in the rules file |
| Outcome log is empty after a quarter | Check whether the question is being asked at the source before concluding nothing happened |
| Health indicators fine, outcomes absent, goal unchanged for two quarters | Reconsider the goal or the roster, or stop |
| Members ask for more help with drafting | Add a conversation, an interview or an editor; still no supplied copy |
Stopping is a legitimate outcome. A programme that people keep alive out of
obligation costs more than it returns.
## 9. Measurement checklist
- [ ] Baseline recorded before or at launch, with the number of members reporting
- [ ] Medians used for per-post figures
- [ ] Like-for-like comparison across windows
- [ ] External share of engagement tracked
- [ ] No individual leaderboard anywhere, including informal ones
- [ ] Outcome log has a "how we know" column
- [ ] The attribution question is asked at first contact
- [ ] Confounders named in every readout
- [ ] Each measure maps to a decision
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| brainstormingobra/superpowersYou MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation). | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| brand-guidelinesanthropics/skillsApplies Anthropic's official brand colors and typography to any sort of artifact that may benefit from having Anthropic's look-and-feel. Use it when brand colors or style guidelines, visual formatting, or company design standards apply. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| frontend-designanthropics/skillsGuidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Helps with aesthetic direction, typography, and making choices that don't read as templated defaults. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| mcp-builderanthropics/skillsGuide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK). | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |