Personalization from LinkedIn

LinkedIn is the richest source of personalization data for B2B cold email. Every prospect's profile is a research document: their title, tenure, career path, posts, comments, skills, and connections. The goal is to extract one specific, relevant data point from LinkedIn and use it as the personalized element in your cold email opener. Not a summary of their profile. Not a compliment about their headline. One data point that connects their situation to a problem you solve.

The principle: LinkedIn research should take 2-3 minutes per prospect for Tier 2 personalization. If it takes longer, you're over-researching. If it takes less, you're under-personalizing. Extract one token, write one line, move on.

What to Extract from LinkedIn

The 5 LinkedIn data layers

What to extract per layer

Layer 1: Profile data (minimum viable personalization)

Layer 2: Recent posts (highest value)

Layer 3: Career history (medium value)

Layer 4: Connections (for warm intro context)

The 2-3 Minute LinkedIn Research Process

For Tier 2 personalization (the default for most outbound), spend 2-3 minutes per prospect.

The checklist (in order of priority)

1. Check Activity → Posts (30 seconds)
   → Did they post in the last 30 days?
   → YES: Read the most recent relevant post. Extract the topic and their take
   → NO: Move to step 2

2. Check profile header (15 seconds)
   → Custom headline? Note it
   → Tenure in current role? Note if < 1 year
   → Company size? Note if relevant to your angle

3. Check About section (15 seconds)
   → Read the first 2 lines only
   → Any stated priorities or goals? Note them

4. Check Experience (30 seconds)
   → Previous company: did they use tools in your category?
   → Recent role change? (< 6 months = new-broom signal)
   → Promoted or lateral move?

5. Check for mutual connections (15 seconds)
   → Any warm intro path?

6. Write the personalization line (60 seconds)
   → One sentence, 15-25 words
   → Connect the strongest data point to the email's angle

Total: 2-3 minutes per prospect. For a 200-person list, that's ~7-10 hours of research. For a 50-person ABM list, that's 2-3 hours.

When to go deeper (Tier 1, 5-10 min per prospect)

For Tier 1 ABM accounts (top 10-20 prospects), extend the research:

Turning LinkedIn Data into Email Openers

The bridge formula

{LinkedIn data point} + {implication for their situation} = opener

Data point alone is an observation. "I saw you posted about attribution." So what?

Data point + implication is personalization. "Your post about attribution being a data problem matches what we see at Series B teams. The gap usually shows up when..."

Examples by data type

Bridge rules

• The implication must connect to a problem you can help with. A data point about their vacation photos doesn't bridge to your product. A data point about their SDR hiring spree bridges to sequencing infrastructure
• One sentence max for the bridge. The opener is 15-25 words total. Don't write a paragraph analyzing their LinkedIn activity
• If you can't find a bridge, use a category observation instead. "Most Series B teams we talk to..." is weaker than a LinkedIn-sourced opener but stronger than a forced, weak personalization

What NOT to Reference from LinkedIn

The line between thoughtful and creepy

Rule of thumb: Reference what they've published (posts, articles, talks). Don't reference what they've consumed (likes, comments on others' posts, profile views).

LinkedIn Research for AI Personalization

When using AI for bulk personalization (per bulk-personalization-with-ai skill), LinkedIn data must be structured for the AI.

What to feed the AI per prospect

{
  "name": "Jane Kim",
  "title": "VP Sales",
  "company": "Acme Corp",
  "tenure_months": 4,
  "headline": "Building the outbound engine at Acme",
  "recent_posts": [
    "Attribution is a data problem, not a tool problem. Here's why...",
    "We just hit 10 SDRs. The processes that worked at 3 are breaking"
  ],
  "previous_company": "Ramp",
  "previous_tools": ["Outreach", "Salesforce"],
  "about_section_first_line": "Focused on scaling pipeline for Series B growth"
}

AI input rules

• Structured data, not raw profile text. Don't dump the entire LinkedIn profile into the prompt. Extract the specific fields listed above
• Include 1-2 recent posts (titles or first sentences). This is the highest-value personalization data. The AI can reference their take
• Include previous company and tools. The AI can generate openers like "At Ramp you were on Outreach. Curious if..."
• Mark "no posts" explicitly. If the prospect hasn't posted, tell the AI: "recent_posts: none." The AI should fall back to profile data, not hallucinate a post

Prospects with No LinkedIn Activity

Not everyone posts on LinkedIn. For prospects with empty activity feeds:

Fallback data sources (in order)

Fallback rules

• If they have no posts, no custom headline, and no about section: Use company-level data (stage, size, hiring) or a category observation. Don't force LinkedIn-based personalization when there's nothing to personalize from
• A clean Tier 1 segment observation beats a forced Tier 2 LinkedIn reference. "Most Series B SaaS teams we talk to..." is better than "I see you've been at [Company] for 2 years" (which adds nothing)
• Mark inactive profiles in your data. When building lists, flag prospects with no recent LinkedIn activity. Route them to category-observation templates instead of LinkedIn-personalized templates

Scaling LinkedIn Research

Time per prospect by personalization tier

Scaling options

Scaling rules

• Manual research is the ceiling on quality. A human reading the prospect's LinkedIn profile and writing a custom line produces the best personalization. But it doesn't scale past 100 prospects per week per SDR
• AI with structured LinkedIn data is the best balance. Extract profile data programmatically (Evaboot, Phantombuster). Feed structured data to AI. AI generates the opener. Human spot-checks 10-20%. This scales to 200+ prospects per batch at near-manual quality
• Don't skip LinkedIn research entirely for prospects above $30K ACV. The 2-3 minutes per prospect pays back on any deal worth > $30K. The only reason to skip is volume pressure on low-ACV campaigns

Pre-Send Checklist

• [ ] LinkedIn data point is from the last 30 days (posts) or current (profile data)
• [ ] Data point connects to a problem, not just an observation
• [ ] Nothing references what they liked, viewed, or consumed (only what they published)
• [ ] Opener is 15-25 words with the LinkedIn data point + implication
• [ ] If no LinkedIn data was found, fallback to company-level or segment-level personalization (not a forced LinkedIn reference)
• [ ] For AI-generated personalization: cross-checked that the referenced post or data point actually exists in the input data (no hallucination)

Anti-Pattern Check

• "Loved your LinkedIn post!" with no reference to what the post said. Generic. Every SDR tool generates this. Reference the specific topic and their specific take, or don't mention the post at all
• Referencing a post from 8 months ago. Feels like deep-stalking. LinkedIn research should use posts from the last 30 days. Older than 30 days = stale
• "I noticed you liked [Person]'s post about [Topic]." Referencing their likes (not their own posts) crosses from thoughtful to surveillance. Only reference content they created
• Spending 15 minutes researching a $10K ACV prospect. The ROI doesn't justify it. 2-3 minutes for Tier 2. 5-10 minutes for Tier 1 ABM only. Cap it
• Forcing LinkedIn personalization when there's nothing to personalize. A prospect with no posts, no custom headline, and a minimal about section. Forced personalization ("I see you've been at [Company] for 3 years") adds nothing. Use a segment observation instead
• Copying the entire about section into the email. "You mentioned in your LinkedIn bio that you're passionate about..." is a paragraph of their own words quoted back to them. Extract one insight, write one sentence, move on
• Using LinkedIn data that doesn't connect to the email's purpose. "I see you're a marathon runner" in a cold email about sequencing tools. Professional data only. The data point must bridge to a problem
• AI-generated personalization referencing a post that doesn't exist. The AI hallucinated "Your post about pipeline velocity" when the prospect never posted about that. Cross-check every AI-generated LinkedIn reference against the actual input data