Closed-lost reactivation.

The play

Closed-lost is the only list in your CRM where every record already knows your product, already took a call, and already told you why they said no. It is also the list nobody touches, because "checking in" with someone who said no is the lowest-status email in sales and reps know it.

The reason they said no is the whole play. "Budget" is answered by funding or expansion. "Not the right time" is answered by a new ops leader or new locations. "Went with a competitor" is answered by a leadership change or a contract coming up. "Our champion left" is answered by whoever replaced them. "No decision" is answered by any of those. A change that doesn't answer the loss reason is trivia, and you don't send trivia.

So the system is a weekly loop: pull the deals in the window, research each account for what changed since the close date, grade the change against the loss reason, and write the opener from the pair. Most weeks it sends a handful of emails. They are the best handful you will send.

Who it's for

• You have at least 50 deals marked closed-lost in the last 18 months. Below that, do it by hand on a Friday.
• Your closed-lost deals have a loss reason, or you're willing to spend a day backfilling one from the notes. Without it the play has nothing to grade against.
• Your sales cycle is 30 days or longer, so a lost deal was a real conversation, not a bounced demo request.
• The reps who lost the deals are still around, or the emails come from a founder or marketing leader. "We talked in June" needs a we.

The build

1. Fix the loss reasons first. Five values: budget, timing, competitor, champion left, no decision. Backfill from deal notes where the field is empty. This is the unglamorous day, and the play is worth exactly as much as this field.
2. Pull the window. Closed-lost, close date between 6 and 18 months ago. Under 6 months it's too soon and they remember the no. Over 18 the contact has probably moved and the conversation is cold enough to be a new deal.
3. Skip the recently touched. Any contact with activity in the last 90 days is in someone's hands already. Dedupe on company so two lost deals at one account don't become two emails.
4. Check the contact is still there. If they left, search for whoever holds the title now. Found: that's your new contact, and the loss reason is effectively "champion left" whatever the CRM says. Not found: skip until next week.
5. Research by loss reason, not by company. Lost on budget: look for funding, an acquisition, new locations. Lost to a competitor: a leadership change, or when their contract started. Champion left: who replaced them. Every finding needs a date and a URL. No URL, no finding.
6. Grade against the table. New website, an award, a hiring post for a receptionist: not material, logged, skipped. The grade is a lookup, not a judgment call, which is what makes it safe to run unattended.
7. Write from the pair. Ten openers, one per (reason, change) combination that passes. The opener names the month you talked, the reason they gave, and the change, in that order. Your product is in the second paragraph.
8. Send small, reopen on reply. One send per account per quarter, three steps over two weeks. A reply moves the deal to a "reactivated" stage with the old deal linked, so the rep sees the history. Monthly, report reactivation rate by loss reason; after a quarter you'll know which reasons are worth chasing and which to drop.

The mock

Same made-up company as the first two plays: Zebrawood, scheduling software for multi-location businesses. Sixty fake closed-lost deals across dental groups, vet groups, freight brokers, staffing firms, and the like, with loss reasons spread across the five values, plus a file of what the research step "found" at 34 of them. Run through the real script.

The grading table

Lost on	Answered by	Not answered by
Budget	Funding, acquisition, new locations	New leader, new website, hiring
Timing	New locations, new ops leader	Funding, award, hiring
Competitor	New ops leader, their contract coming up	Funding, new locations, hiring
Champion left	A named replacement in the role	Anything else, because there's nobody to write to
No decision	New leader, new locations, funding	New website, award, hiring

Two accounts, one email

Out of 60 fake deals, 23 got an email. Eight had a change that didn't answer the loss reason, eight were too recent, seven had been touched in the last 90 days, seven had no change found, six were too old, and one had a contact who'd left with no replacement found yet. Twenty-three of sixty is high because the fake research file is generous. On a real book, a third of accounts having a material change inside a year would be a very good quarter; one in six is normal.

The files:

• deals.csv, 60 fake closed-lost deals as they'd come out of the CRM
• changes.csv, what the research step found, with a source for each
• reactivate.py, the window, the skips, the grading table, the ten openers

python3 reactivate.py 2026-10-27
# 60 closed-lost deals, 23 get an email
# skipped: 8 change not material to the loss reason, 8 too recent,
#          7 touched in the last 90 days, 7 no change found, 6 too old,
#          1 contact gone, no replacement found

-> Juniper Ridge Staffing  lost 2026-01-15 (budget), owner Dana, to Nora Alvarez
   change: funding on 2026-10-09  [https://news.example/juniper-ridge-growth-round]
   "When we talked in January 2026, budget was the blocker. Since then you took a
    growth investment to add 10 branches over two years, which usually changes
    that conversation."

-> Foxglove Fitness  lost 2026-03-01 (champion left), owner Dana, to Hannah Osei
   change: new_person_in_role on 2026-10-06  [https://people.example/foxglove-regional-manager]
   "Your predecessor and I talked about scheduling in March 2026, before they left.
    Nothing was decided. Worth 20 minutes to show you where it ended up?"
...

What breaks

The loss reason is a lie. Reps pick "timing" when they mean "they didn't like me" and "budget" when they mean "I never found the decision maker." The play can't fix that, but the monthly report by loss reason exposes it: a reason that never reactivates is a reason that was never real. Rename or retire it.

The research invents a change. Any step that asks a model "what changed at this company" will, some weeks, get a confident answer about nothing. The rule is a source URL or no finding, and the grading step drops anything without one. The mock's changes file has a source column for this reason; in production, a human spot-checks ten a week.

The contact left and you wrote to them anyway. An email to someone who left eight months ago, about a conversation they had before they left, forwarded to their replacement with "lol," is the worst outcome on this page. Step 4 exists to prevent it, and the mock skips the one account where it couldn't find a replacement.

Credits. Research costs money per account. The window, the 90-day skip, and the dedupe all run before the research step, and the research only runs on what's left. Sixty deals became 39 research calls in the mock, not 60.

Too many emails. If the loop sends more than 10 percent of the window in a week, something in the grading is too loose. Cap it at one send per account per quarter and read the first month's output yourself.

Copy it

Take the three files, export your own closed-lost deals in the same columns, and run the script with an empty changes file to see how many accounts even make the window. Then wire the research step to Apollo or Clay with one prompt per loss reason, and move the script into n8n on a Monday schedule. If you'd rather have the whole thing built, with the sequences and the HubSpot stages, it fits inside the Outbound Engine install as the second campaign.