How to automate outreach follow-ups

A sequence is a named, ordered list of follow-up steps, each with its own delay and its own subject/body content.

• Add steps in order; each one waits a number of days after the previous step (the first step waits that many days after the initial contact email).
• Each step can hold up to five variants, giving you an A/B (or A/B/C/D/E) test on subject and body per follow-up.
• Every variant needs a whole-number weight, and a step's variant weights must add up to exactly 100 before the sequence can be turned on.
• A sequence is off by default; enabling it validates that every step has at least one variant with weights summing to 100.

No lead joins a sequence just because it exists: enrollment is always an explicit, per-lead action.

• Opt a lead into a sequence right when you send its first cold email, via the sequence picker on the send composer.
• Or, for leads you've already contacted (status Sent), select them from Your Leads and use Enroll to add them to a sequence's follow-up track.
• Re-enrolling a lead that's already on a sequence resets its progress back to the first step, so the UI warns you before doing this to more than one lead at once.

A background job checks for leads whose next follow-up is due and sends it, without you needing to trigger anything manually.

• It only ever considers leads still in Sent status with a due next-follow-up date; a reply, conversion, or unsubscribe already takes a lead out of contention automatically.
• A failed send is retried a day later rather than abandoning the lead.
• If a run is interrupted mid-send, its in-progress claim on a lead is treated as abandoned after 10 minutes and picked up again, so a crash can't leave a lead stuck forever.

Each step's variants are sent using weighted-random selection, so your test traffic splits the way you set it up.

• A variant with 60% weight gets roughly 60% of that step's sends, a variant with 10% weight gets roughly 10%, and so on.
• A lead's variant for a step is picked once, on its first attempt, and reused for any retry, never re-rolled mid-way through.
• If a previously-assigned variant is later deleted from the step, the next send for that lead simply picks a fresh one instead of erroring.

An optional, per-step toggle that lets BrandOps nudge weight toward whichever variant is actually performing better, instead of leaving weights fixed forever.

• Off by default. You turn it on per step, and it requires at least two variants on that step to enable.
• A weekly job scores each variant by opens plus double-weighted clicks per send (a click counts for more since click-through requires real intent, while opens can be inflated by automatic image pre-fetching).
• It only adjusts once every variant in the step has at least 30 sends in the trailing 90 days, and only when the leading variant is clearly ahead of the runner-up; otherwise it leaves weights untouched that week.
• Each run shifts at most 5 weight points toward the leader, and no variant is ever pushed below a 10% floor, so a test never fully closes on autopilot.
• This is a bounded, easy-to-audit shift toward the current leader, not a full statistical experimentation engine, and it's driven by opens/clicks rather than by tracking actual replies.