Social
LinkedIn, X, and Instagram in one place: what would be tracked, the cadence Play 3 publishes on, and a worked example of the numbers.
No account is connected yet — the fixture below shows the shape of the dashboard, not Shepherd's reach.Illustrative
LinkedIn
Illustrative14,200
followers
+1160 · 30d
- Posts, 30d
- 22
- Engagement rate
- 5.2%
Top post: 24 hours, on average (fixture)
X
Illustrative6,800
followers
+540 · 30d
- Posts, 30d
- 18
- Engagement rate
- 3.8%
Top post: Underwriting is a data problem (fixture)
Instagram
Illustrative3,400
followers
+260 · 30d
- Posts, 30d
- 15
- Engagement rate
- 4.6%
Top post: Send us the data center (fixture)
These three cards are a fixture, not Shepherd's data — no LinkedIn, X, or Instagram account is connected (see Integrations, category Social). They show the shape of the dashboard once Play 3 is publishing on a real cadence.
What gets tracked, once an account is connected
Planned| Metric | Definition | Source |
|---|---|---|
| Followers, 30-day delta | Growth per platform, read from each platform's own analytics API. | LinkedIn Pages API, X API, Instagram Graph API |
| Posts published vs. planned | What actually went out against the calendar below. | Compared to lib/clips.ts CADENCE |
| Engagement rate | Likes, comments, reposts, saves over impressions, per post and per platform. | Platform analytics API |
| Signal-value clips that convert | Whether a clip marked high signal_value actually drove more profile-to-broker engagement than a low one, testing the hypothesis in lib/clips.ts. | Platform analytics, joined to clip metadata |
| Engagement to intent | A broker who engages with a post should raise their Intent axis automatically. Today that axis is manual (see the engagement-social claim on Plan) until this is wired. | Feeds lib/scoring.ts, not built yet |
| Founder posts to submission spikes | Julian's suggestion from our Round 4 call: overlay Justin's posting dates on submission and media-request volume, and watch for the lag. Directional, not causal, but it prices the founder-content bet with Shepherd's own data. | Post dates from LinkedIn, submissions from the CRM once wired |
Top clips, by signal value
IllustrativePlay 3| Clip | Platform | Length | Signal |
|---|---|---|---|
Two weeks or one day Your broker sends a risk to eight markets. Seven take two weeks. | 75s | high | |
65% fewer claims Tenured Procore users have 65% fewer claims. We price on that. | X | 40s | high |
The MGA question Whose paper? Every broker asked us that for three years. | 90s | medium | |
Underwriting is a data problem We read loss runs with OCR so an underwriter doesn't have to. | X | 45s | low |
Data centers Data centers are the hardest construction class to place right now. | 45s | high |
Publishing cadence, first 30 days
IllustrativeWeek 1
- Mon
- Record (60 min, Justin + Mo)
- Tue
- Edit: 12 to 16 cuts
- Wed
- LinkedIn clip 1
- Thu
- X clip 1
- Fri
- Instagram clip 1
Week 2
- Mon
- LinkedIn clip 2
- Tue
- X clip 2
- Wed
- LinkedIn clip 3
- Thu
- Instagram clip 2
- Fri
- X clip 3
Week 3
- Mon
- LinkedIn clip 4
- Tue
- X clip 4
- Wed
- Instagram clip 3
- Thu
- LinkedIn clip 5
- Fri
- X clip 5
Week 4
- Mon
- LinkedIn clip 6
- Tue
- Instagram clip 4
- Wed
- X clip 6
- Thu
- Review: what held the first minute
- Fri
- Engagement sync into Play 1