Wedges written from the detector, not a merge field.

What is a Nerxus One sequence?

A sequence is an ordered set of steps a contact is enrolled in, advanced by a periodic sequence_tick job that asks ComplianceGate before anything leaves. The contact moves queued, to active on a step, through waiting or sending or sent, and then exits with a reason.

The buyer is the marketer. You are dripping your own book, on your domain, about a finding you can fulfill. The SMB's customers never sit in this enrollment table. If you sell the shop a win-back sequence for their no-shows, that build lives in Forge. This page is your hunt.

A step has an order, a channel, a template, a delay after the previous landing, a send window in the recipient's timezone, and a branch. The default branch on reply is exit. On click of a DEMO link the branch can escalate to a human. On no engagement it walks to the next step. A/B arms can split subject or body and promote a winner after a threshold you set.

The engine is channel-agnostic on purpose. It asks the channel registry to validate and send. Adding a channel later does not rewrite the tick. That is also why a new channel cannot invent a send path around the gate. The step still has to come back with a GatedMessage.

Enrollment is a row, not a vibe. You can read next_run_at, the current step, and the last verdict. If the queue is down, an accepted enroll can sit. The UI is supposed to say queued, not sent. When you come back, you are looking at what the tenant actually advanced.

Campaign members come from Pipeline or from a signal filter over saved businesses. Relay does not import a rented CSV and call it a wedge. If you already paid for a list, Scout can score it first. A sequence that starts from an unscored file will still pass the gate, and it will still sound like a file.

How do the No-Website, No-SSL, Permit-Lead, Chair-Hour Loss, and Service-Gap wedges differ?

Each library wedge is keyed to a detector so the first email can name that detector's evidence instead of a generic intro. No-Website attaches a live pre-built site when Forge has one. No-SSL names the trust gap and can attach a fixed sample page. Permit-Lead uses a recent permit hook for home services. Chair-Hour Loss is the med and dental no-recall or no-show finding with a modeled loss calculator. Service-Gap names the high-margin service they forgot to list and can attach a built service page.

Relay's suggest_sequence picks from that library by vertical and signal when you launch from a Radar hook. You still review. The library is a starting template plugin. Tenants clone and edit. Spam-lint and the required physical address and unsubscribe still run on the clone. A cute rename does not retire the gate.

Other seeded shapes exist in the spec, including a speed-to-lead proof and an AI-invisible restaurant letter. This page locks the five named above because they are the ones the marketing site will keep repeating. If your hunt is hiring-for-a-receptionist, you can still build that cadence. Do not expect the homepage to pretend it is one of the five.

Audience can be a signal filter over saved leads: all no_ssl rows in this list, or five-star shops with no booking widget. That filter is the bridge from the detector catalog to enrollment. A sequence aimed at No-Website that then mails every saved pin is a campaign you misnamed.

The signature play across the wedges is the same motion with different nouns: the finding, the evidence, the DEMO or calculator, the book link that is yours. If you strip those out to raise open rate, you have a sequencer again.

No-Website without a Forge preview is a weaker letter, and it is still an honest letter if you say you noticed the missing site and you can build one. Do not attach a generic template URL and call it their DEMO. Claimable liveUrl is tracked. A stock theme on a random host is a different product's screenshot.

Chair-Hour Loss dollars stay modeled. The calculator can show a band with assumptions. The sequence should not promote the midpoint as a bookkeeping fact. If two screens ever print different chair-hour numbers, that is the same defect Pipeline already had on DEV and FEE: two pickers, one truth required.

Permit-Lead ages. A permit from yesterday is a hook. A permit from four months ago is a cold card Pulse will already have painted stale. Enrolling a decayed permit into a 'hot project' letter is how you sound like you scraped a public record and stopped reading the date.

What does stop on reply actually do?

Stop on reply exits the enrollment the moment a reply is ingested on any active step, so the contact stops receiving the rest of the drip. The thread opens in Inbox. Pulse can start the SLA clock on the deal. You talk to a person, or you do not mail them again from this sequence.

Exit reasons are stored: replied, goal_met, unsubscribed, bounced, suppressed, manual_stop, completed, or compliance_block. Replied is the happy stop. Unsubscribed writes suppression and is a hard block on later steps. Bounced pulls the address. Completed means the steps ran out without a human answer, which is a finished drip, not a win.

A click on the DEMO is not automatically a reply. The branch can notify you or move to a warmer follow-up. Do not treat a click as permission to text. Stop-on-reply is about inbound language, not about a pixel.

OOO and auto-responders are classified as such so a vacation note does not empty the sequence and does not look like Interested. The tick can hold and reschedule. If you have ever kept getting a six-step cadence after you answered a vendor, you already know why this branch is the default.

Manual stop is for you. Pause and stop tools exist on the campaign and on the enrollment. Deliverability degradation on an identity can pause sending on that mailbox without pretending the audience unsubscribed.

How do quiet hours work on a sequence?

Quiet hours are a hold on the step until the recipient's local window is open. The gate can attach holdUntil. send() then returns held and never touches Resend. The tick tries again when the window opens.

Default quiet bounds, where a jurisdiction does not say otherwise, run from 21:00 to 08:00 local. Email still respects a business-hours send window you set on the step. Voice and SMS, on the rare consented path, are stricter, which is one more reason those steps stay unreachable for cold contacts.

Timezone comes from the contact when you have it, and is unknown when you do not. Unknown is conservative. The product would rather hold than guess a coast. You can set the window on the step; you cannot set a window that overrides a jurisdiction rule the gate already applied.

Hold is visible. Collapsing hold into sent would make a midnight delivery look compliant in the audit. The Compliance Center exists to show held, blocked, and sent as different rows.

Quiet hours are not a courtesy footer in the copy. They are a state the sequencer understands. If you need mail to land at 2 a.m. because a guru said that is when owners read, this product will still hold.

What are warmup caps and why does the sequencer honor them?

Warmup caps are the daily send limit on a new or recovering identity, raised on a schedule as deliveries, opens, replies, and low complaints accrue. The sequencer never exceeds the current cap on that mailbox, even if the campaign still has two hundred enrolled contacts.

Caps stack. Plan entitlement, warmed cap, per-identity hourly throttle, and the send window all apply. The lowest limit wins. Authoring shows projected daily volume against that floor so you can see an undeliverable design before you launch.

Rotation across identities on one verified domain spreads volume. Each identity carries its own warmup and reputation. A fresh mailbox on a warm domain is still a fresh mailbox. Do not paste a six-month-old cap onto a record you created yesterday.

The warmup_ramp job is what moves the number. You do not type a heroic daily count into a settings field and call it warmed. If the identity is paused for complaints, the sequence waits. That wait is the product working.

A sequence that 'needs' more volume than the domain can stand is a sequence you should split or delay. Instantly-class tools will sell you more mailboxes. This page will tell you the cap is the feature.

Step funnel metrics exist so you can see where people leave: delivered, open, click, reply, goal. They are counts, not ROI. This site will not turn a 12 percent reply rate into a revenue promise. If an arm wins on reply and loses on goal-met, promote with your eyes open. A booker beats a curious opener.

Detach and re-enroll were once refused because the database could not store them. When the migration closed that gap, the stale refusal had to die in the same breath. A guard that outlives its gap is the same defect as a tooltip that outlives its fix. If you can detach now, the UI should let you, and the gate should still run on the next send.

How does a sequence pick a channel without skipping the gate?

The step names a channel. At runtime the tick asks the channel's validate, which calls ComplianceGate, then send() only if a GatedMessage exists. A voice or SMS step on a contact with no consent record is skipped and logged. The engine does not rewrite the step into email in silence, and it does not force the forbidden send.

Email is the workhorse and the only cold-ok default. LinkedIn remains human-paced assist. RVM stays off until you acknowledge the risk and have an opted-in relationship. Warm voice needs disclosure copy and a consent row. SMS needs the Forms-born consent and a registered number.

Because the locks live on the channel definition, a future plugin cannot become autonomous-safe by omitting itself from a switch. autonomousSafeChannels() is derived from the registry. If the gate never agreed, Autonomous cannot grow a member.

Idempotency is per enrollment and step. Retries after a worker blip do not double-send a Day 0 letter. That key is formatted in one place so two jobs cannot invent two spellings of the same send.

If you want a three-channel cadence for a consented client reminder, you can design it. If you want that cadence for a cold plumber you found on Tuesday, the gate will empty the non-email steps and leave you with email, which is the product telling the truth.

LinkedIn steps, when you connect an account, stay human-paced. The sequencer can queue a view or a connect draft. It will not fire a hundred DMs overnight. Account restriction is a ToS risk, not a TCPA caption, and it is still a reason the default is assist.

Delay is wait after the previous step lands, not wait after you clicked launch. Day 0 means the first GatedMessage that actually sent. A held Day 0 because of quiet hours does not start the Day 2 clock. The tick measures from landing, so a coast-to-coast audience does not compress into one Tuesday night.

How does Radar's finding become the first step of a sequence?

A finished diagnostic emits a hook Relay can consume. Launch outreach from this diagnostic, or the hook_ready event, builds a draft campaign: the audience is the diagnosed business or list, the template is populated from top findings and the DEMO url, and the suggested sequence is the library row that matches vertical plus signal.

No copy is written from a blank page in that path, and no finding is re-derived. Same findings, no rework. That is the two-artifact contract the diagnostic already paid for. You still press review and launch.

Chat can do the same walk. Draft an email campaign to all leads with no SSL is an intent Relay already knows: resolve the audience, enrich missing people if you have that key, verify risky addresses, draft the No-SSL letter, run the gate, show a preview card with count, sample, verdict, and volume versus capacity.

If twelve of the rows are UK sole traders and forty are corporate subscribers, Relay can say it will mail the forty and hold the twelve until you confirm a lawful basis. That sentence is the gate talking, not a tone setting.

Skipping Radar still lets you enroll a list. The first step will then sound like a list. The wedges above assume you did the diagnostic, or at least kept the cheap detector from Scout.

What else exits a sequence besides a reply?

Goal-met exits when the marketer meeting is booked or a claimable DEMO is claimed, depending on how you set the goal. Unsubscribed and complained exit and suppress. Bounced exits the address. Manual stop is you. Completed is the drip finishing cold. compliance_block is the gate saying there is no remaining legal path.

Suppression is shared with Pulse so a suppressed contact reads as suppressed on the board. Re-enrollment of a suppressed handle should fail closed. A marketer who exports the list to 'just send this one' from a personal mailbox has left the product, and the product cannot save that domain.

A lost deal in Pipeline does not by itself exit an enrollment. Stop the sequence, or let suppression and stage automations queue the task. Reactivation later is a new enrollment, email-first, still gated, still honor-unsub.

A/B promotion is not an exit. The loser arm stops getting new enrolls once you promote. People already on the loser finish or exit by the normal reasons.

If you need a sequence that never exits, you do not need this sequencer. You need a newsletter tool, and those have their own consent rules this page will not pretend to replace.

Who should use detector-named sequences, and who should stay in Instantly?

Use these sequences if you sell a named fix to local SMBs and you can attach a DEMO or a calculator in week one. The library exists so you stop rewriting the same No-SSL letter every Tuesday.

Stay in a volume sequencer if your offer is generic, your list is rented titles, and your edge is mailbox count. Instantly will feel faster. This product will feel like it is arguing with you. The argument is the gate, the cap, and the finding.

Agencies that run many verticals in one tenant still get one isolation boundary. A sequence you built for Austin plumbers does not leak into another tenant. v1 does not sell white-label sub-accounts of the sequencer.

If you only need an autoresponder after a form submit, Forms can enroll on submit, and that enroll is Supervised because it is outreach. Do not hide a cold sequence behind a thank-you page and call it transactional.

Common sequence mistakes that burn a domain

The first mistake is enrolling the whole saved pile into No-Website because the campaign was handy. The wedge is the filter. A mixed pile is how a dentist gets a permit letter.

The second mistake is turning off stop-on-reply so 'the cadence can finish.' The cadence finished when they spoke. Everything after that is how you earn a spam complaint and a paused identity.

The third mistake is raising the daily cap in your head because warmup 'looks conservative.' The cap is conservative on purpose. A new domain that sends like a warmed one is a domain Resend or the mailbox providers will notice.

The fourth mistake is adding an SMS step to a cold enrollment and assuming the skip means it will send later. The skip means never, until a Forms consent row exists. Later is a different product motion.

The fifth mistake is launching from chat without opening the preview card. The card is where audience count, verdict, and volume versus capacity sit. Approve is the send. Skimming the card is how a blocked segment still surprises you on Friday.

Get started if you want the next batch. Bring a vertical, a metro, and one detector you can fulfill. The library already has a name for that letter.

A sixth mistake is treating hold as sent because the calendar said the step 'ran.' Hold is visible. Collapsing hold into sent would make a midnight delivery look compliant in the audit. The Compliance Center exists to show held, blocked, and sent as different rows.

How does sequence_tick decide the next send time?

sequence_tick decides the next send time by reading next_run_at on the enrollment, then asking whether the recipient window is open, the identity still has warmup room, and ComplianceGate will brand a GatedMessage. If any of those fail, the tick holds or skips and writes why. It does not invent a send to keep the campaign looking busy.

Delay is wait after the previous step lands, not wait after you clicked launch. Day 0 means the first GatedMessage that actually sent. A held Day 0 because of quiet hours does not start the Day 2 clock. The tick measures from landing, so a coast-to-coast audience does not compress into one Tuesday night.

Timezone comes from the contact when you have it, and is unknown when you do not. Unknown is conservative. The product would rather hold than guess a coast. You can set the window on the step. You cannot set a window that overrides a jurisdiction rule the gate already applied.

If the queue is down, an accepted enroll can sit. The UI is supposed to say queued, not sent. When you come back, you are looking at what the tenant actually advanced. A worker blip plus a retry is idempotent on enrollment and step order. A repeat whose prior outcome was sent or held returns that prior result.

Caps stack on that same decision. Plan entitlement, warmed cap, per-identity hourly throttle, and the send window all apply. The lowest limit wins. Extra enrollments wait. That wait is the product working, not a broken launch.

What does a compliance_block exit look like on an enrollment?

A compliance_block exit looks like an enrollment that stopped because no remaining channel has a compliant path, with the block stored and the reason readable. It is not a silent skip that looks like success. Fix the reason, then replay. Blocked and failed can be re-evaluated because the fix is a change you make.

Exit reasons are stored: replied, goal_met, unsubscribed, bounced, suppressed, manual_stop, completed, or compliance_block. Completed means the steps ran out without a human answer, which is a finished drip, not a win. compliance_block is the gate saying there is no remaining legal path.

A voice or SMS step on a contact with no consent record is skipped and logged. The engine does not rewrite the step into email in silence, and it does not force the forbidden send. If every remaining step is a locked channel, the enrollment should exit compliance_block rather than sit forever looking active.

Relay explains a block in chat with the rule and the remediation. The Compliance page keeps the four hard locks visible. There is no hidden toggle labeled trust this list. Autonomous cannot widen the path that just blocked.

If you add an SMS step to a cold enrollment and assume the skip means it will send later, you misread the log. The skip means never, until a Forms consent row exists. Later is a different product motion, and it still needs a registered A2P number.

How do A/B arms promote a winner without rewriting the gate?

A/B arms split subject or body and promote a winner after a threshold you set, and promotion does not rewrite ComplianceGate. The loser arm stops getting new enrolls once you promote. People already on the loser finish or exit by the normal reasons. Every arm still needs a GatedMessage.

Step funnel metrics exist so you can see where people leave: delivered, open, click, reply, goal. They are counts, not ROI. This site will not turn a reply rate into a revenue promise. If an arm wins on reply and loses on goal-met, promote with your eyes open. A booker beats a curious opener.

Spam-lint and required-element checks still apply to both arms. A cute subject that trips all caps or a missing unsubscribe does not get a hall pass because it is the B variant. Clone the library wedge, then edit. A rename does not retire the gate.

Promotion is not an exit. Replied, unsubscribed, and goal_met still beat a winner arm. Stop-on-reply remains the default branch on both. Turning it off so the cadence can finish is how you earn a spam complaint and a paused identity.

Detach and re-enroll were once refused because the database could not store them. When the migration closed that gap, the stale refusal had to die in the same breath. If you can detach now, the UI should let you, and the gate should still run on the next send.

Can a form submit enroll someone in a sequence?

A form submit can enroll someone in a sequence if you set that auto-action, and that enroll is outreach, so it is Supervised and still gated. A contact form is not a hidden cold drip. Do not hide a cold sequence behind a thank-you page and call it transactional.

Forms-born SMS consent is a different ledger line from an email enroll. An email they typed on a quote request is not SMS permission unless they also ticked SMS. The consent field cannot be pre-checked. A bundled marketing opt-in is a publish block.

Booking a marketer meeting is a goal-met exit. The follow-up sequence is supposed to die when the calendar event is created. Their patient book on a Forge site is a different calendar and does not exit your drip, and it should not have enrolled them in No-SSL either.

Client onboarding questionnaires attach to Steward after Close. Those are existing relationships. Enrolling them in a promotional sequence is still outreach and still Supervised. Transactional is a narrower word than marketers like.

If you only need an autoresponder after a form submit, keep the enroll small, email-first, and review the preview card. The card is where audience count, verdict, and volume versus capacity sit. Approve is the send.

What happens when warmup capacity is lower than the audience?

When warmup capacity is lower than the audience, the sequencer drips at the cap instead of honoring a design that cannot be delivered. Authoring shows projected daily volume against available capacity before you launch. Extra enrollments wait. Hold is not sent.

A campaign of two hundred contacts on a cap of thirty is a week, not an afternoon. Hourly throttle sits under the daily cap so a morning blast cannot empty the day in ten minutes. Combined with quiet hours and recipient business hours, the drip is what launch means.

Rotation across identities on one verified domain spreads volume. Each identity carries its own warmup and reputation. A fresh mailbox on a warm domain is still a fresh mailbox. Do not paste a six-month-old cap onto a record you created yesterday.

The warmup_ramp job is what moves the number. You do not type a heroic daily count into a settings field and call it warmed. If the identity is paused for complaints, the sequence waits. That wait is the product working.

A sequence that needs more volume than the domain can stand is a sequence you should split or delay. Instantly-class tools will sell you more mailboxes. This page will tell you the cap is the feature. The public site will not invent a send volume to make the form look finished.