Sequences - multi-step LinkedIn follow-ups that auto-stop on reply

A sequence is the ordered series of DMs SocialScalr sends after a connection request is accepted. Up to 4 follow-up steps. Per-step delays, merge fields, spintax variants, A/B testing, automatic stop on reply or disconnect. The sequence is the part of outbound where most of the reply rate is won or lost.

Why 4 steps and not more

LinkedIn DM decay curves are brutal. From our customer data across 12,400 sent follow-ups:

• First follow-up: ~18% reply rate at top-decile templates.
• Second follow-up: ~9% reply rate.
• Third follow-up: ~4% reply rate.
• Fourth follow-up: ~2% reply rate.
• Fifth+ follow-up: under 1%, and recipient sentiment turns negative.

We capped at 4 deliberately. Longer sequences buy almost no extra pipeline and start damaging your reputation.

The canonical 4-step structure

1. Day 0: Connection request. Specific signal + your relevance, no CTA. ~270 chars.
2. Day 3 (after accept): First DM. Curious question, soft opt-out. ~75 words.
3. Day 9: Second DM. Shorter, gentle ask. "Worth a 15-min look?" ~35 words.
4. Day 14: Third DM (break-up). "If now isn't the time, no problem." ~30 words.
5. Day 21: Final touch. Short value-add (a relevant article, frame, or a specific question). ~40 words.

See our deep-dive on the structural skeleton for what goes in each message.

Per-step features

Each step in a sequence supports:

• Delay in days (1-30) before firing. Sequence starts the countdown from the previous step's send (or from the connection accept for step 1).
• Time-of-day window. Send only during your configured working hours (default 9am-6pm local). Step waits for the next valid window if delay completes overnight.
• Day-of-week mask. Skip weekends (default) or specific days.
• Merge fields: {{first_name}}, {{company}}, {{title}}, {{custom_field_X}}. Verified against the lead record before send - merge-field failures are surfaced before, not after.
• Spintax: {Hi|Hey|Hello} {{first_name}} - random pick per send.
• Up to 4 message variants per step for A/B testing.

Automatic stop conditions

The sequence pauses immediately on:

• Inbound reply detected from the lead.
• Lead disconnects (rare but happens).
• Lead stage manually moved to Won or Lost.
• Manual pause on the lead or campaign.
• Campaign hits its end date.
• Lead account becomes inactive on LinkedIn (account deleted, suspended, or set to private).

The lead's activity timeline records exactly which condition stopped the sequence so you can audit.

A/B testing

Every step can run up to 4 message variants. SocialScalr round-robins them across leads in the campaign. After enough volume (typically 200+ sends per variant), the Insights tab surfaces which variant has the higher reply rate with confidence intervals.

Best practices:

• Test one element at a time. Don't change opener + CTA + length in one variant - you won't know what moved the needle.
• Hold a champion. Always have one variant be your current best-known template. New variants are challengers.
• Promote when significant. When a challenger beats champion by 3+ percentage points across 400 sends, promote it to new champion.
• Use the A/B significance calculator if you want to verify statistical confidence yourself.

Pacing and throttle

Sequence steps respect the daily caps you've configured. If your daily DM cap is 50 and the sequence wants to fire 75 follow-ups today, the extra 25 queue for tomorrow. The sequence never blows past your safety limits.

Within the day, sends are paced with randomised gaps (5-15 minutes by default) so the activity pattern looks like a human checking LinkedIn periodically, not a bot bursting.

Sequence cloning

Save sequence templates and re-use them across campaigns. Common library entries customers build:

• "B2B SaaS founder outbound" - 4-step, curious-question opener, value-led second touch
• "Recruiter sourcing" - 4-step, role-specific opener, opportunity-led second touch
• "Investor outreach" - 2-step manual (auto-send disabled), tracking-only
• "Event follow-up" - 2-step, conference-context opener, faster cadence

What sequences don't do (intentionally)

• No conditional branching. If a lead engages with a post, do X; otherwise do Y. We considered it; rejected because it adds complexity without measurable ROI in our data.
• No email channel inside the sequence. SocialScalr is LinkedIn-only. Multichannel works via outbound webhook to your email tool.
• No InMail. Sequences send to 1st-degree connections only. InMail is a paid LinkedIn feature outside our scope.
• No more than 4 steps. See the decay curve above.

Plan availability

• Free: 1 sequence, 2 steps max.
• Starter ($29/mo): Unlimited sequences, 4 steps, basic A/B (2 variants per step).
• Pro ($49/mo): Full 4 variants per step, AI personaliser per step, smart send-time analyser.
• Agency ($149/mo for 10 seats): Shared sequence library across team, per-seat owned campaigns.

Related

• How to write a follow-up that gets a reply - the message structure.
• Anatomy of a 40% acceptance note - the connection request that feeds the sequence.
• Campaign setup help - how to wire a sequence into a campaign.
• Note critique tool - score your step 0 (the connection note).