Why this is in the vault
The founder provided Justin Welsh's LinkedIn OS course PDF (274 pages, dated to roughly mid-2021) to pressure-test against the live buyer-credibility plan (project_credibility_for_phdata_sales). Most of its tactics are now dead, penalized, or built for a different goal (solo-creator product sales, not phData buyer pipeline) — this note exists to separate the ~30% that still works (profile funnel, writing order, outbound rhythm) from the ~70% that's dated or counter-productive, so the founder doesn't import 2021 LinkedIn tactics into a 2026 plan by default.
Mapping against Ray Data Co
Concrete connections to the live plan: (1) the 5-hr/week "Apply it" schedule directly operationalizes project_credibility_for_phdata_sales — build/write/post/comment/conversation blocks sized against the founder's actual weekly budget, not Welsh's daily-posting cadence; (2) the outbound rhythm (10-15 named prospects, patient commenting, never pitch, reach out after 2-4 months) is adopted wholesale for the HCLS data-leader prospect list, since "the ask is a conversation" matches the plan's own framing; (3) the profile-rewrite proposals (headline, banner, About, Featured) give a concrete draft to react to once the founder unlocks his actual LinkedIn profile for editing; (4) the dead/risky-tactic table (hashtags removed Feb 2024, engagement-partner commenting now named as a policy violation) is a compliance guardrail — it heads off the founder accidentally reintroducing a now-penalized tactic via an old course. Net effect: validates the plan's existing shape (phData pipeline over audience growth) rather than changing it, and hands off four open questions (brand-policy constraint on naming phData/using its logo, posting cadence, the 15-prospect list membership, and whether commenting is a smaller first step than posting) that only the founder can resolve.
Verdict
Skim it and use about 30%. These parts still work: the profile funnel, his writing order (the point first, then the opening lines, then the summary; the published post opens with those lines, p.100), the topic-by-style idea matrix, and the "never pitch; comment for 2-4 months, then reach out" outbound rhythm. The rest is built for a solo creator selling a $150 course: daily posting, engagement groups, polarizing hooks and a tripwire ladder of products. The course dates to about mid-2021 (inferred): it cites "Top LinkedIn Hashtags of 2021" (p.214) and shows a Stripe chart covering Jan 2020 to Jun 2021 (p.10). Several of its tactics are now dead or penalized.
Bias note: his numbers (51M views and the revenue screenshots on pp.7-11) come from one survivor. He was an ex-startup executive who started posting in 2019-21, and the course is the product his system sells. Treat his results as a ceiling, not an expected outcome. It is my inference, not sourced, that organic reach was easier to get then.
Pages covered: the full text layer for all 274 pages (about 120 pages have more than a title), plus rendered images of pp.7, 10, 11, 111, 178, 184, 211, 220, 233, 245, 251 and 256. I skipped repeated funnel-build slides (pp.177-190), example banners and headshots (pp.194-201) and swipe-file screenshots (pp.265-273).
What it teaches (paraphrased)
Foundation (pp.16-53). Narrow from topic to niche to sub-niche. Build a persona from three parts: a relatable 7-beat backstory, 3-5 strong opinions ("be polarizing or be unknown") and 3-5 short stories that each carry a lesson.
Content (pp.54-130). Choose a voice path:
- Leading: you did it yourself.
- Discovering: you studied it so the reader does not have to.
- Reporting: you interviewed people who did it.
Then cross themes with styles (observation, contrarian, X vs Y, listicle and others) to generate about 8 ideas at once. Write in this order: the point first, then a 3-line opener that earns the "see more" click, then a summary that invites comments. Post daily at a fixed time, with a partner who reacts and comments early. Get on calls with the people who DM you and ask what they tried, why it failed and what it is costing them. After 60-90 posts, repurpose the winners.
Tribe (pp.131-172). Spend 45 minutes replying on your own posts and 15 minutes commenting on top creators' posts. Ask creators when they post, bookmark their activity pages and try to comment first. Grow those contacts into relationships. Never connect and then pitch.
Lead capture (pp.173-228). The profile is a funnel: a banner someone can read in about 2 seconds, a tagline in the form "I help {who} {do what} {how}", a free download in Featured that collects email addresses, and an About section that adds proof.
Business (pp.229-263).
- Inbound: posts 5-7 times a week, sending readers to long-form pieces on his site, then a newsletter, then cheap products, then coaching.
- Outbound: a list of 10-15 prospects, Google Alerts and patient commenting.
- Selling: either the tripwire ladder or a mix of 80% content and 20% expiring offers.
Still relevant in 2026?
| Welsh tactic | 2026 status | Evidence |
|---|---|---|
| Profile hashtags (pp.203-216) | Dead. Profile hashtags were removed in Feb 2024 and the Creator Mode toggle in Mar 2024 | Contentin glossary; Lindsey Gamble |
| Accountability partner who comments early (pp.104-111) | Risky. LinkedIn's 2026 post names engagement pods and automated comments as violations. Whether a single manual partner counts as a pod is my inference | LinkedIn News, 2026; Social Media Today |
| Racing to comment first on creators' posts (pp.136-140) | Not named in LinkedIn's post. Low value for reaching buyers (my inference) | Same LinkedIn post |
| Opening lines and holding attention | Still central. The feed now uses large language model (LLM) retrieval and a sequential ranking model. Search Engine Land says it favors "topical relevance and engagement patterns." My inference is that a consistent topic now matters more than before; the paper still uses engagement counts | Search Engine Land; arXiv 2510.14223, Oct 2025 |
| Moving readers off-platform | Right goal, higher cost. SocialPilot reports about 60% less reach for posts with external links, and says the link-in-first-comment workaround is penalized too. Unverified vendor claim | SocialPilot, dated Sep 22 |
| Formats | Carousels reach about 2.5-4x as many people as text posts, trend "stable / slightly up." Vendor data | Carouselli 2025 |
| Organic reach | Views down 50% year over year to Feb 2025 (van der Blom, via Contentin) | Contentin, Dec 2025 |
| Buyer value of thought leadership | Supports our goal. Buyers say good thought leadership raises their support for a vendor. It also reaches "hidden" buyers (finance, legal, procurement) that traditional sales and marketing overlook. The research is LinkedIn-sponsored | DemandGen Report on Edelman-LinkedIn 2025 |
Where it differs from our plan
| Topic | Welsh | Our plan | Better for buyer credibility that turns into phData pipeline |
|---|---|---|---|
| Goal | Audience growth, then an email list, then products | phData pipeline: reps pulling him into deals, buyer conversations | Ours. |
| Monetization | Ladder from a $5-25 product up to $1,250 coaching | No info-product ladder. Side bet: a Snowflake Marketplace dataset listing. A paid newsletter is his call | Ours. Drop pp.248-261. |
| Cadence | 5-7 posts a week | About 5 hrs/week; 6 build-in-public pieces by Nov 30 | Ours, plus about 2 LinkedIn posts a week cut from each piece. I infer that daily posting at this budget would mean filler, which weakens topical consistency. |
| Persona | Polarize; rags-to-riches backstory | Practitioner building in public on Snowflake + Claude, on his own site under his own name, with phData named as his employer | Ours. VP-level buyers want judgment. Keep strong opinions, framed as "here's what broke." |
| Voice path | Lead / discover / report | Build in public | Both. Building in public is Welsh's Leading path. Reporting fits later conversations with buyers. The current conversations list (program veterans, Snowflake reps, his manager) is for calibration, not content. |
| Engagement | Pods; race to comment first on creators | Community replies tied to the build; networking | Ours. Comment on buyers' posts, by hand. |
| Outbound | 10-15 prospects, Google Alerts, never pitch, reach out after 2-4 months | Not specified | Adopt Welsh's, with Healthcare / Life Sciences (HCLS) data leaders as the prospects. The ask is a conversation. |
| Funnel | A free download in Featured that collects email addresses, then a newsletter | EmDash site as the hub, LinkedIn as the main channel, Resend list (2026-09-28 chat decision) | Same shape. The call to action is "read the build notes / talk to me." Keep links out of post bodies. |
| Headline | "I help {who} {do what} {how}" | Not yet decided | Welsh, adapted. The "how" is phData's practice, not his own offer. |
Apply it: a 5-hr/week plan
| Block | Time per week | What it covers |
|---|---|---|
| Build | 1.5 hrs | Work on the Snowflake + Claude side bet that the pieces describe |
| Write | 1 hr | One site piece (the hub), aiming for 6 pieces by Nov 30 |
| LinkedIn posts | 40 min | Two posts cut from the piece, each opening with lines that earn "see more" and ending with a question for buyers. No link in the body; the link lives in Featured. One document post a month (diagram plus a "what broke" list) |
| Comments | 50 min (10 min/weekday) | Hand-written comments on posts from about 15 named HCLS data VPs and Directors, plus phData and Snowflake reps |
| Community reply | 20 min | One reply a week where the build gives a real answer (target 8 by Nov 30) |
| Conversation | 30 min | One networking conversation a week (target 8). Use his questions: what have you tried, why didn't it work. Log patterns to the HCLS problem log |
| Tally | 10 min | Friday count of DMs and comments from buying titles, not followers |
Profile rewrite proposals
I could not see the profile: LinkedIn showed a sign-in wall and I did not log in. These drafts come from the plan only.
- Headline: "Running AI agents in production on Snowflake + Claude | Healthcare & Life Sciences data | phData"
- Banner: One line ("AI agents in production for data teams") plus the site URL. Leave the phData logo off unless the brand policy allows it.
- About: Three short paragraphs: what he builds; proof (Claude Certified Architect – Foundations (CCA-F) 850, Snowflake GenAI 854, builds he can name); then an invitation to read the build notes and compare notes.
- Featured: His best build write-up, the Resend signup and, later, a talk.
Open questions for the founder
- Does phData's social media or brand policy require a disclaimer, or limit naming clients or using the logo?
- Two posts a week, or post only in weeks a site piece goes up?
- Who goes on the 15-prospect list, and should reps be on it?
- Would commenting daily on buyers' posts be a smaller first step than posting, given that posting still feels like a leap?
Related
- [[2026-09-29-nov30-credibility-targets]]
- [[2026-09-29-conversations-to-have]]
- [[2026-04-06-content-os-welsh]]
- [[2026-09-29-agents-need-your-definitions]]
- [[project_credibility_for_phdata_sales]]