Most LinkedIn sequences are written as a drip campaign: connect, then fire a message the moment the request is accepted, then two more on a timer regardless of what the person did. That is not a sequence. That is a countdown to being archived.

A sequence that earns replies treats every step as part of one conversation with one person. The connect note, the wait, the first message, the single follow-up: each one assumes the last one was read by a human who has not yet decided you are worth answering. Here is the shape we use.

Step 1: the connect note does the heavy lifting

The connection request is not the throat-clearing before the real pitch. It is the most important message in the sequence, because it is the only one a stranger reads before deciding whether you exist. Three sentences, one specific reason, no calendar link. We wrote the whole thing up separately in how to write a connection request that does not sound like spam. If the connect note is generic, no follow-up rescues it.

Step 2: wait, and do not send the “thanks for connecting” message

The fastest delete on LinkedIn is the automated message that lands two seconds after someone accepts your request. It tells the recipient one thing: a script is running. Everything you send after that is read through that lens.

Wait at least a day. A real person who wanted to connect with you does not immediately produce a paragraph. The pause is not a delay tactic, it is the difference between a conversation and a funnel.

Step 3: the first message gives before it asks

The first message after an accept has one job: be worth the reply. The mistake is treating the acceptance as permission to pitch. It is not. It is permission to say one useful thing.

Reference why you connected. Reuse the specific reason from your connect note so the thread reads as continuous, not as a fresh cold open.
Give something small. A relevant observation, a link they would actually open, a one-line answer to a problem their role implies. Not a deck.
End with at most one question. Yes/no, or a single open question that is cheap to answer. “Is this still a priority for your team this quarter?” beats “Do you have 15 minutes?”

Step 4: one follow-up, not five

The follow-up that works is a single, short, additive message a few days later. The follow-up that gets you blocked is the fourth “just bumping this to the top of your inbox” with no new information.

A good nudge adds something: a reason the timing changed, a result you just saw, a genuinely shorter version of the ask. If you have nothing to add, do not send it. Silence is a real answer, and respecting it is the thing that lets you message that person again in six months without being remembered as the bot that would not stop.

The rule that makes all of this safe: stop on reply

The single most important behavior in any sequence is that it stops the instant the person replies. Sounds obvious. It is the thing automated sequences get wrong most often, because the timer does not know the conversation moved. Sending scheduled step 3 to someone who already answered step 2 is how you turn an interested reply into a lost deal.

Before any step fires, the sequence has to read the actual conversation and confirm the person has not responded, been marked as a do-not-contact, or already converted. If any of those is true, the step does not send.

Cadence: spread it like a human, not a cron job

Days between steps, not hours. A connect, then a day or two of silence, then the first message, then three to five days before the single follow-up.
Business hours, recipient’s timezone. Messages sent at 3:00 AM read as automated even when the words are perfect.
Vary the openers. Identical first lines at scale trip duplicate-message detection. The body can be templated; the first sentence should not be.

What Marno does for you

Marno runs the sequence as a conversation, not a countdown. Every step does a pre-flight check of the real thread before it sends, so a reply, a do-not-contact mark, or a conversion pauses the sequence automatically. Steps are spaced over days with human-pace jitter and a per-timezone working-hours window, and personalization tokens carry the specific reason forward from the connect note. You write the structure once; it behaves like a person every time.