Loops: Sales Nurture That Only Sends When There's a Reason To

Every SDR has the same graveyard: the 90% of a finished campaign that never replied. They weren't bad fits — they passed qualification, they got the sequence, they just didn't bite this quarter. Most teams do one of two things with these people: nothing, or a monthly "just checking in" drip that quietly burns the domain and teaches prospects to ignore you.

Today we're launching Loops, MarketBetter's sales nurture agent. A loop takes the people who finished your campaigns without replying and keeps them warm — not on a calendar, but on a condition: each person gets one short email, in your voice, only when there's an update that actually fits them. Most days, most people get nothing. That's the feature.

Why nurture is broken

Classic nurture is a drip: the same newsletter to everyone on the list, every N weeks, forever. It fails for a structural reason — the trigger is time, not relevance. The prospect can tell. Open rates decay, unsubscribes climb, and the mailbox reputation you need for real outbound erodes in the background.

The fix isn't better newsletter copy. It's inverting the trigger: start from something genuinely new and worth saying, then find the handful of people it's relevant to — and say it to only them. That's expensive to do by hand, which is why nobody does it. It's exactly the kind of judgment-per-person work an AI agent can do at scale. Here's how Loops does it.

Who gets in — and why everyone else is out

A loop belongs to a rep. When you turn it on, it sweeps your finished campaigns and your last 90 days of sends, and considers every person who was delivered and settled. Then it applies entry checks, in order:

• Hard stops: replied, bounced, unsubscribed, or left the company — out.
• Sequence state: if any sequence that included them is still running (or an enrolment hasn't sent yet), they wait. Nurture never talks over active outreach.
• Soft holds: an out-of-office on file holds them until the return date. Automated replies get classified, not ignored.
• CRM suppression: opted out in your CRM means opted out here.
• The customer/opportunity/relationship gate: existing customers, open opportunities, and accounts your team is already talking to are kept out. Nurture is for the cold middle, not live deals.
• Address re-validation: every email address is re-verified before the person can enter. Addresses go stale fast; a loop that runs for months can't trust a validation from last quarter.
• Fit check: an AI fit review against your seller criteria, with company context. "Fit unclear" is a visible state, not a silent drop.

Every single person gets a row with a state and a plain-English reason — "Replied," "Open opportunity," "In another campaign right now," "Your team is already talking to them." You can open the loop's audience view and see exactly who the loop is working, who's waiting, and why each excluded person is out. If two reps' loops want the same person, the CRM account owner's loop wins and the other rep can see that too.

The monthly brief: research about you, never about them

Once a month, the nurture agent builds a brief: 5 to 10 dated, sourced updates your non-responders might care about. It pulls from three places:

1. Your own updates — your RSS feed, blog, or news page. Product launches, new content, announcements.
2. Industry news — grounded search across the industries you sell into, with every candidate article opened and read on its own page. A search snippet is a lead, not a source; if the agent can't read the page and confirm the publish date from the page itself, the item doesn't make the brief.
3. Your library — the context files your workspace has already uploaded (case studies, one-pagers, positioning docs).

Just as important is what the brief refuses to include: company news about the prospect, hiring, pricing moves, lawsuits, layoffs, and competitor news are all excluded by rule. Research is about you and your market — never about the prospect. Nobody wants a nurture email that opens with evidence you've been watching them.

You get 24 hours to review each brief before anything is written to anyone. Skip an update and it's gone permanently — a skipped item never comes back, even if next month's research finds the same page again (items are deduplicated by source URL, not by row). Still-valid items carry over so the brief stays full without re-researching the world every month.

The daily decision: pick, write, check — or wait

Once a day, the loop looks at the people whose spacing has passed, in rotation — account priority, weighted by how long it's been since they last heard from you — up to a daily send budget. For each person, three steps, all from the already-reviewed brief:

1. Pick. A model chooses one update that's valid on the send date, hasn't already gone to this person, wasn't skipped by you, and is under the company cap (no more than 3 people at the same company ever get the same update). If nothing clears those bars, the decision is wait — and the code honors it. No filler send, no fallback newsletter.
2. Write. One short email — 50 to 100 words — in the rep's voice, about that one update, written near send time so it's never stale by the time it goes.
3. Check. A second model, deliberately from a different model family, reviews the draft and fails closed. It requires the email to contain a recipient-specific implication — a concrete reason this update matters to this person's role or account — not a generic blast with a name token. The writer gets one rewrite using the checker's notes; if that still doesn't pass, the email is blocked and the person simply waits for the next update.

Every outcome — send, wait, or block — is recorded with its reason, and you can read the reasons on the loop's timeline: "Every current update has already gone to them," "Three people at their company already got this update," "The checker held the email back."

What personalization actually uses

The writer and checker see three things about the recipient: their current role (re-read at decision time, not from a stale import), the account's recent intent signals, and the pages people from that account visited on your site over the last 90 days. URLs are stripped of tracking parameters, your page-exclusion rules apply, and account visits are treated honestly — as account context, never as a claim that this person was on your pricing page. The model is explicitly told the difference.

The safeguards, because nurture runs unattended

A sequence runs for two weeks under your nose. A loop runs for quarters. The safety model is built for that:

• Preview mode first. A new loop runs as a dry run: it makes real decisions and writes real emails you can read, but sends nothing until you flip sends on. Previews don't consume updates or company caps, so turning sends on later doesn't find people locked out by their own previews.
• A last-second gate. Between the decision and the actual send — which can be days apart when business-hours holds push a send to Monday — every email is re-checked: a reply, a pause, an unsubscribe, a skipped update, sends being switched off, or the update expiring all stop it at the door.
• Freshness. An email about news more than two days old at send time doesn't go.
• Mailbox self-protection. If a sending address crosses an 8% bounce rate on recent sends, the loop pauses itself and tells you why, with the math.
• Engagement exit. Two loop emails with no opens or clicks and the person exits. Nurture that nobody reads is just noise with your domain on it.
• Live-deal exits. A reply, a new opportunity, an account heating up, or a teammate starting a conversation pulls the person out automatically — the loop's job is to hand people back to real selling, not to keep them.
• A holdout. A slice of every loop is held out from sends entirely, so you can measure what the loop actually causes instead of taking the dashboard's word for it.

Steer it, or let your agents steer it

Each loop has a focus — steering you write in plain language ("lean on the manufacturing case study this month," "soft-pedal the pricing update") — and you can scope a focus to tagged segments of the loop. Changed your mind? A focus edit invalidates queued emails written under the old steering; they're re-decided, not sent anyway.

And because loops live behind the same API as the rest of MarketBetter, they're fully reachable through our MCP server: an AI assistant can discover the loop endpoints, inspect a loop, read the brief, and update the steering — with no feature-specific glue. If your team is already running sales workflows through AI assistants, your nurture loop is one more thing they can manage.

What the pace really looks like

A loop with 800 people, 42-day spacing, and a 20-a-day budget sends at most 19 emails a day, and each person hears from you at most once every six weeks — and only when an update fits them. In practice many people wait longer, because most days nothing new is relevant to them. That's the inversion: a sales cadence guarantees touches on a schedule; a loop guarantees relevance and lets the schedule float.

If your current nurture is a sequence tool blasting a static list — or worse, a stalled-deal spreadsheet nobody re-works — this is the difference between activity and signal.

FAQ

Is this a drip campaign? No. A drip sends on a schedule regardless of content. A loop sends only when a specific, dated, sourced update fits a specific person — and waits otherwise. The daily budget is a ceiling, not a quota.

What does the AI know about my prospects? Their role, their account's intent signals, and the pages their account visited on your site. The research itself is only ever about your company and your market — the brief never contains information gathered about the prospect.

Can I see emails before they go? Yes, twice. Every new loop starts in preview mode where you can read every email it would have sent, and every monthly brief has a 24-hour review window where you can skip any update before it's used.

What stops it from spamming an account? Per-person spacing (weeks, not days), a 3-person-per-company cap per update, the two-email no-engagement exit, and a mailbox-health circuit breaker that pauses the loop at an 8% bounce rate.

How do people leave the loop? Automatically: a reply, a new opportunity, an unsubscribe, a job change, an account heating up, or a teammate opening a conversation all remove or pause them — with the reason shown.

---

Loops is rolling out now. If you want your finished campaigns to keep working the 90% who didn't reply — without a newsletter and without the risk — book a demo and we'll turn a loop on with your data.