Playbook
LinkedIn outreach playbook for founders: work the people who already engage with you
Stop spamming people on LinkedIn.
By Imran Patel, Founder and CEO, Syft. Published .
Most LinkedIn outreach goes to strangers: a list, a sequence, a pitch in the connection note. The people who react to your posts, comment, view your profile or follow you are different. They already know your name.
This playbook is how to turn them into relationships, and eventually into customers. It works by hand, or with the AI instructions below pasted into Claude, ChatGPT or Grok.
Summary
- Reach out only to people who came to you: DMs, connection requests, comments, reactions, profile views and follows.
- Check each one against a one-line ICP. Most won't fit; leave them alone.
- Send connection requests blank, or with one line about what they engaged with. No pitch.
- DM only when their timing turns: a job change, a DM, a return after a quiet stretch, several posts in one week. Three sentences, one follow-up.
- About 10 minutes a day and 20 minutes a week.
- On Imran Patel's account, blank requests were accepted 59% of the time from profile viewers and 32% from post engagers. Belkins' 2025 average is 26%.
Why relationships first
Chet Holmes' buyer's pyramid in The Ultimate Sales Machine (2007) puts about 3% of a market as buying now and 7% as open to it. The other 90% is not looking. It is a rule of thumb from a sales book, not a measurement.
A pitch only works on the 10%, and everyone else is pitching them too. A relationship works on the 90%: you are already in their network when their timing changes.
What counts as spam
- A pitch in a connection note.
- A sequence of DMs to someone who never answered.
- The same message pasted to a list.
Connecting is not spam. A blank request, or one line about something real, is how people meet on LinkedIn. Pitching before there is a relationship is the problem. LinkedIn restricts accounts whose invitations are often ignored, left pending or marked as spam.
Who it is for
Founders, execs and sellers who post on LinkedIn at least once a week. If you don't post yet, start with the people who engage with posts by others in your space (rule 1).
The 10 rules
- 1
Start with the people who came to you
In order of warmth:
- They sent you a DM.
- They sent you a connection request. Check who they are, accept, then DM.
- They commented on your post.
- They reacted to more than one of your posts.
- They viewed your profile.
- They followed you.
- They engage with posts by peers, influencers or competitors in your space. Follow those posts. Reply to the person's comment first, so they see your name, then connect.
Everyone else is a stranger. Leave them out.
- 2
Keep the ones who fit
Write your ICP in one line: the kind of company, its size, and the role that buys. Check every name against it. Most of your audience won't fit. That's normal. Keep posting for them, but don't reach out.
- 3
Connect without a pitch
A connection request is not a pitch. Send it blank, or with one line about what they engaged with. No link, no calendar, no product. Don't invent context when there isn't any; blank is fine.
Send it from the founder's account when you can. If a teammate shares more mutual connections with them, send it from that teammate.
In Belkins' 2025 data (15.1 million outreach touchpoints), requests without a note were accepted 27.6% of the time and requests with a note 25.3%. Requests with a note got more replies: 8.2% against 5.3%. A short, real line helps. A pitch doesn't belong in it.
- 4
Be quick
Engagement goes cold in days. Connect within a day of the comment, reaction or view.
- 5
Nurture, don't sell
Once you're connected, your posts reach them. That does most of the work.
Most of your audience isn't ready to buy. When someone leans in, send one useful thing: a guide, an invite, a post they would want. One send. No sequence.
- 6
Reach out when the timing turns
These are the moments to DM:
- A job change into a role that buys what you sell.
- They DM you.
- They come back after a quiet stretch.
- They engage with several posts in one week.
DM that day. Keep it to three sentences and end with an easy way out. Make “yes” cheap: something they can answer in one line without committing to a call.
- 7
Follow up once, then stop
Most messages get no reply. Silence usually means not now, not no. Follow up once with something relevant, then leave it. Unanswered messages stacked on each other read as automation. If it looks like you're talking to yourself, stop.
If they accepted your request but never replied, mention the connection when you email or call later. You're a familiar name, not a cold one.
- 8
Don't say how you know
“I saw you viewed my profile” reads as surveillance. Talk about the topic or their work, not about what you tracked.
- 9
Keep a record
Write down who you connected with, what they engaged with, what you sent and when. A spreadsheet is enough. If you have a CRM, put the people who fit there too, so your team sees the history before they reach out.
Without a record, someone who has commented on your posts for six months gets a cold email from your company that reads like you never met.
- 10
Pace yourself and prune
LinkedIn doesn't publish its sending limits. It restricts accounts that send many invitations in a short time, or whose invitations are often ignored, left pending or marked as spam. Stay at a pace that looks like a normal person, every day, rather than in bursts. Fewer, better-placed requests beat volume.
LinkedIn caps you at 30,000 connections. Withdraw requests nobody accepted after three weeks; a withdrawn request can't be re-sent for up to three weeks. When you get near the cap, remove connections who don't fit and never engage.
Messages that fit the rules
Connection note after a comment
Thanks for the comment on the pricing post. Good to be connected.
DM after a job change
Congrats on the new role at {company}. If {problem} lands on your plate there, happy to share what we've seen work. No worries if not.
DM after repeat engagement
You've been in the comments on the last few posts, appreciate it. What are you working on at {company} these days?
Value send to someone who leaned in
I wrote up how we handle {topic}. Want me to send it over?
A routine
- Every day, 10 minutes: new comments, reactions, profile views and followers from the last 24 hours. Check fit. Connect with the ones who fit.
- Every week, 20 minutes: job changes, repeat engagers, people back after a quiet stretch. DM the ones whose timing turned. One follow-up for anyone who went quiet. Withdraw requests older than three weeks. Update the record.
- Every month: look at the numbers and rewrite your ICP line if the wrong people keep showing up.
What to measure
- Requests accepted out of requests sent.
- Replies out of DMs sent.
- Conversations and meetings started.
Split each one by where the person came from (comment, reaction, profile view, follower) and by who they are (title, persona, industry). It tells you which posts bring in buyers and which bring in everyone else.
Results from one founder's account
Connection requests sent from Imran Patel's LinkedIn account in 2026, by who received them and what the note said.
17 of 29 (59%)
People who viewed his profile
Note: Blank. Sent Apr 13 to Jul 2, 2026.
265 of 817 (32%)
People who reacted to or commented on his posts
Note: Blank. Sent Apr 23 to Sep 13, 2026.
4 of 48 (8%)
Founders running LinkedIn thought-leadership ads, from a list (no prior contact)
Note: A one-line pitch offering a free list of their ad engagers. Sent Aug 27 to Sep 3, 2026.
How we measured
- One LinkedIn account: Imran Patel, founder and CEO of Syft.
- The requests were sent from his account by Syft's LinkedIn agent, our own product, through a Chrome extension in his browser.
- Accepted means the person accepted the request. Requests still pending count as not accepted (12 profile viewers, 103 post engagers). Requests that failed to send, for example because the profile was not found, are left out.
- Counted from Syft's database on October 2, 2026.
Limits
- One account, and not a controlled test. Each row has a different audience, a different note and a different time, so the gap between rows can't be put down to the note alone. Belkins' larger dataset puts the difference between a note and no note at about 2 points.
- It doesn't compare people who engaged with cold contacts sent the same blank request. We haven't run that test.
- Replies and meetings aren't counted here.
Run it with an AI assistant
Paste the instructions into Claude, ChatGPT or Grok, in a chat that can run code on files, and attach your LinkedIn export. Tested on a real export: it finds DMs waiting on you, inbound requests, people to reconnect with and old requests to withdraw, and drafts messages that follow the rules above. It never sends anything.
- Get your LinkedIn export: LinkedIn > Settings > Data privacy > Get a copy of your data > the larger archive. LinkedIn emails a link, usually within 24 hours.
- Paste everything inside the block and answer its first question.
The export has your DMs, connection requests, posts and comments. It has no reactions, profile views or followers, so paste those names in yourself.
You are my LinkedIn relationship assistant. I sell on LinkedIn and I want to stop sending cold pitches to strangers.
Work only with people who already came to me.
Rules:
1. Before anything else, ask me for my ICP in one sentence: the kind of company, its size, and the role that buys.
Also ask for the company names of my current customers, partners and investors (I can skip this).
The export has no company size or industry, so judge fit from title and company name, and mark size as
unverified.
2. Only consider people who came to me: they messaged me, sent me a connection request, commented on my posts, or
are in the reaction, profile-view or follower names I paste (as "name, what they did, date"). Leave out people
selling to me, my coworkers, the customers, partners and investors I named or my DMs name, and anyone who asked not
to be contacted or whom I asked to stop. If a customer is waiting on my reply, list them separately with no draft. Tasks E and F are the only exceptions.
2a. The files: messages.csv = my DMs. Invitations.csv = requests sent and received (Direction column; no title,
company or status). Connections.csv = current title and company (the header starts at "First Name"; skip the
notes above it). Comments = comments I wrote, not comments on my posts. It escapes quotes with a backslash. Its
links use activity IDs and Shares uses share IDs: to find my comments under my own posts, divide each ID by 2^22
(milliseconds since 1970) and match to the share within 5 seconds. A reply of mine there that starts with someone's
name answers their comment. Shares = my posts. No file lists who reacted to my posts, viewed my profile or
follows me; use only names I paste for those. Read the files with code, not by eye. Match people on profile URL
after decoding %-escapes, not on name.
3. Connection notes: only for people who engaged and are not yet connected to me. Blank, or one sentence about what
they engaged with. Under 200 characters. No link, no calendar, no product, no pitch.
4. DMs: only when timing turned: their last message to me is from the last 90 days and unanswered; they came back
after 6+ months of silence; they commented on 2 or more of my posts in one week; I told you they changed jobs; or task C picked them for a reconnect.
Three sentences at most. No link or calendar in the first DM. End with a question they can answer in one line,
and give them a way out ("or not a priority right now?") unless they asked for the call. If someone
engaged but none of these apply, suggest commenting on their next post instead of a DM.
5. Follow-ups: one, with something relevant. If they still don't reply, stop and mark them stopped in the log.
6. Never mention profile views or anything that sounds like tracking. Talk about the topic or their work.
7. Never send anything. Give me drafts to review.
8. If a question needs data that isn't in the files, say so. Don't guess.
Tasks, in this order:
A. People who fit my ICP and engaged with me in the 30 days before today, warmest first (DM > connection request
with a note > comment > request without a note), with the reason for each.
B. DM threads from the last 90 days where the last message is theirs and needs an answer. Skip thank-yous,
emoji-only replies and people selling to me. Then connection requests I received and haven't accepted, with
their note. The export has no title for these, so list them for me to check rather than judging fit.
C. People I exchanged 10 or more DMs with, on at least 3 different days, with no DM or comment either way in the last
12 months. Draft a reconnect DM only if the last message is theirs and unanswered; otherwise suggest no message.
D. For the top 10 across A to C that meet rule 3, 4 or 5: a connection note or a DM. Connection notes can go to
people with no title in the files; mark them "check fit". If fewer than 10 qualify, say
how many and why. Where a draft needs something only I know, leave a [placeholder].
E. Connection requests I sent more than 3 weeks ago where the person is not in Connections.csv, those with my personal note first, then oldest first. Flag
any I wrote a personal note for, and any who told me in a DM they won't connect.
F. Only if I have more than 25,000 connections: connections with no DM, no comment and no invitation in the files,
whose title does not fit my ICP. Give the count before listing.
Output each list as a table I can paste into a spreadsheet: name, company, title, why, suggested action, draft text.
At most 15 rows per table; give the total and offer the rest as a CSV. Write "not in files" for any blank field.
End with a log table: date, person, what you drafted, status (draft / sent / replied / stopped). I will paste it back
next week with what I actually sent; count follow-ups from it.FAQ
- What acceptance rate should I expect on LinkedIn connection requests?
- Belkins measured an average of 26% across 15.1 million outreach touchpoints in 2025. On Imran Patel's account, blank requests to people who viewed his profile were accepted 59% of the time (17 of 29), and blank requests to people who engaged with his posts 32% (265 of 817).
- Should I add a note to a LinkedIn connection request?
- Only a short, real one, and never a pitch. In Belkins' 2025 data, requests without a note were accepted 27.6% of the time and requests with a note 25.3%, but requests with a note got more replies (8.2% against 5.3%).
- How many LinkedIn connection requests can I send a day?
- LinkedIn doesn't publish a number. It restricts accounts that send many invitations in a short time, or whose invitations are often ignored, left pending or marked as spam. Imran Patel's account sent a median of 10 requests a day on the 89 days it sent any, with a high of 38.
- When should I withdraw a pending connection request?
- After three weeks without an answer. Once you withdraw, LinkedIn won't let you invite the same person again for up to three weeks.
- What is LinkedIn's connection limit?
- 30,000 first-degree connections. When you get near it, remove connections who don't fit your ICP and never engage with you.
- How much time does this playbook take?
- About 10 minutes a day to connect with new engagers who fit, 20 minutes a week for DMs, follow-ups and withdrawals, and a monthly look at the numbers.
- How often do I need to post on LinkedIn for this to work?
- At least once a week. If you don't post yet, start with the people who engage with posts by others in your space.
- Is it against LinkedIn's rules to automate outreach?
- LinkedIn prohibits software and browser extensions that scrape or automate activity on its site, and says accounts that use them risk being restricted or shut down. Every step in this playbook can be done by hand, and the AI instructions on this page only draft messages; you send them.
Sources
- LinkedIn Help, “Types of restrictions for sending invitations”The 30,000-connection cap and what triggers invitation restrictions. No published sending limit.
- LinkedIn Help, “Withdraw an invitation”A withdrawn invitation can't be re-sent to the same person for up to three weeks.
- LinkedIn Help, “Prohibited software and extensions”LinkedIn's ban on software and browser extensions that scrape or automate activity on its site.
- Michael Maximoff, “What are B2B LinkedIn Outreach Benchmarks?”, Belkins, updated June 29, 202615.1 million outreach touchpoints from 2025. 26% average acceptance; 27.6% without a note and 25.3% with one.
- Chet Holmes, The Ultimate Sales Machine (2007)The buyer's pyramid: about 3% of a market buying now, 7% open to it.
- Kellen Casebeer, “LinkedIn Outbound: A Deal Lab Playbook (2026)”Source for rules 1, 3, 6, 7 and 10: inbound requests, blank requests, cheap yeses, one follow-up, pacing.
Further listening and reading
GTM Field Notes, episode 1
Kellen Casebeer (The Deal Lab) on LinkedIn outbound
Syft blog
LinkedIn signals are a commodity. Here are 6 plays that still create pipeline.
Lessons from Kellen Casebeer on the LinkedIn outbound plays that still work when everyone has access to the same signals. By Imran Patel.