A HubSpot auto renewal date is either a cliff or a project, and the only thing that decides which is whether the calendar runs backward from it. Run forward, from the day someone decides to leave, it’s a list of things to build and a date that keeps getting closer. Run backward, from the day the contract ends, it’s a short list of things that can’t be done afterward, a longer list that can, and a cutover placed early enough that the old portal is still there if something breaks.
This guide is the backward version, and it comes from one exit: a small B2B software company with a lean marketing team, paying HubSpot $14,000 a year, with a contract end date that was fixed before we opened the portal. Every date below is given as T minus, counted back from the day the contract ended, because the shape of the calendar is the point and the dates themselves are the client’s.
What actually happens on the HubSpot auto renewal date if you leave
The short answer is less than most people fear, plus one thing most people don’t plan for.
HubSpot doesn’t disappear. The portal drops to the free tier with the contacts, companies, deals and tickets still in it, and the API stays reachable for whatever the free tier still includes. Nothing in a well-built exit requires the company to renew or reactivate a paid subscription, and on this engagement that was written into the signed scope as a plain statement of fact.
Every paid feature ends with the contract, and the one an exit feels first is Marketing Hub email sending. When the contract ends, the automated chain stops, and every link in every email the portal ever sent, which on this portal was routed through a HubSpot-hosted sending subdomain, stops resolving the moment that subdomain is pointed elsewhere.
That asymmetry is the whole calendar. The CRM side of an exit has a soft landing: the data is still there three months later if you missed something. The email side has a hard one. If the new sender isn’t carrying the program when the contract ends, the program is off, and the reputation a new sending domain needs takes weeks to earn and can’t be bought faster. Every date in this guide is placed by that one fact.
The dividing line is not hard versus easy
Most exit plans sort the work by difficulty, and that’s the wrong axis. The architecture memo on this engagement put it in one line we’ve used on every exit since: the dividing line isn’t hard versus easy, it’s dies-with-the-portal versus doesn’t.
On this one, the list of what dies with the portal held the suppression and bounce history, including the addresses that belonged to deleted contacts; the spam and territory rules that existed only inside two workflows’ enrollment criteria; the pipeline stage labels and their order; a custom object holding 98 contract records and the 50 renewal dates someone had typed in by hand; every file attachment, which is in no CSV export; the HTML of 442 marketing emails; the definitions of 90 workflows; and a restore of the archive that a named person has actually opened. 1 See also 1 See Dies With the Portal: Seven Things a HubSpot Export Will Not Give You, for what’s on the extraction list and why each item is unrecoverable afterward.Read Dies With the Portal: Seven Things a HubSpot Export Will Not Give You That list became the first phase of the signed scope, and it carried a note beside its date that no other phase carried: cannot be done after the contract ends.
What doesn’t die includes the four reports anyone actually looked at, which can be rebuilt once someone asks for them; the long tail of website content and its redirects; the bridges to other systems; and the re-permissioning of contacts whose consent is unclear. All of it can follow the contract end, and on this engagement most of it was left outside the fixed scope to do exactly that.
A plan sorted by difficulty front-loads the website because it’s the biggest piece. A plan sorted by mortality front-loads a week of extraction that produces nothing a CEO can click on, and that week is the one that can’t be bought back.
Week zero is extraction, not build
We wrote the architecture for this exit 46 days before the contract ended. It scheduled the week that followed as week zero: extractions and applications, and no build work. Day one was the contact-level suppression export. Day two was the rules inside the workflows and the pipeline stage labels. Nothing in that week needed a decision from the client or a line of code.
Week zero didn’t run. Our own notes, written a week later, say so in plain words: nothing in the record showed the suppression export, the rules files or the stage labels had been pulled, and the week’s work was all audit refinement. The plan was still correct, because the first contracted phase already held every one of those items. What had changed was that there was no buffer left in front of it.
A plan sorted by difficulty front-loads the website. A plan sorted by mortality front-loads a week of extraction that produces nothing anyone can click on, and that week can’t be bought back.
The lesson is about sequence. Extraction is the only work in an exit that gets no easier with time and can’t be rescheduled past the contract end. Everything else can. So it goes first, before the proposal is signed if at all possible, because the day the extraction is done is the day the calendar stops being a cliff.
Why does the email warm-up have to start inside the contract?
It has to start inside the contract because a sending domain earns its reputation at a rate nobody controls.
The company’s marketing domain had been quiet since 2023. A domain that quiet has no recent reputation with mailbox providers, so it warms up like a new one under Google’s bulk-sender rules, in force since February 2024, and the usual planning figure is four to eight weeks of widening sends to reach production volume. 2 Source 2 Google Workspace Admin Help, “Email sender guidelines,” in force for bulk senders since February 2024. Senders of 5,000 or more messages a day to Gmail must authenticate with SPF, DKIM and DMARC, support one-click unsubscribe, and keep reported spam rates under 0.1%, never above 0.3%. Our plan put the first warm-up send exactly four weeks before the contract ended, which is the floor of the window with nothing beneath it. 3 Editor's note 3 The four-to-eight-week ramp is planning guidance from email platforms and deliverability vendors, not a measurement.
The ramp doesn’t need to finish before the contract ends, and it won’t. It needs to start early enough that a reputation problem surfaces while HubSpot is still the fallback. If the new domain lands in spam in its second week, you find out at T minus two weeks, with the old sender still paid for and still able to carry the program. If the ramp starts after the contract ends, the same problem surfaces with no sender at all.
Two dependencies sit in front of the first send. The first belongs to the customer rather than the agency: the new sending subdomains need their authentication records published, and on this engagement the DNS lived in the client’s own cloud account next to the records its product runs on, so nothing could be published until someone there was introduced. Our ask to the client said it without softening: there’s no default if this goes unanswered, and every day it slips comes out of the warm-up window. The second dependency is the suppression list, which loads into the new platform before a single subscriber, so the new sender never re-learns years of bounces by sending to them. 4 Definition 4 Sending-domain setup, in order: a subdomain for marketing mail and a separate one for transactional mail, never the apex; SPF, DKIM and DMARC published before the first send; the suppression list loaded before any subscriber; the ramp ordered by engagement, widening weekly.
One design decision followed from the same risk, and it cost nothing. Transactional mail (form confirmations, support replies) was split from marketing mail onto its own subdomain, so a reputation problem on the marketing side can never stop a form confirmation arriving. At this company’s volume, about two inbound leads in five months, every one of those confirmations matters.
Never cancel early
The temptation is obvious. The new platform is live at T minus three weeks, the old one costs $14,000 a year, and a prorated month looks like free money.
The signed scope on this engagement listed, among the things the client provides, no early cancellation of the HubSpot contract, with the reason attached: it’s already paid for through the contract end, and cancelling early would end marketing email sending before the new platform is carrying it. Email sending is the paid feature you most need running until the ramp has widened to full volume and the first real send has landed.
There’s a second reason, and it shapes the rest of the calendar. The old portal is the only fallback the project has, it’s paid for, and it’s the one system in the plan known to work. Everything about where the cutover sits follows from refusing to throw that away a day earlier than the contract does.
Cut over before the last day
We placed the cutover at T minus nine days, with a window to T minus seven for monitoring and the final three days before the contract end held as a sweep. The decision record gives the grounds in one sentence: HubSpot is the only fallback and it’s already paid for, so a plan that needs the last day to go right has no fallback at all.
That sentence was written against two of the three candidate designs, which had cut over on or near the deadline. One put the DNS flip at T minus three days, with a self-hosted CRM stood up in 46 days and zero slack, and the review called it the weakest. The design that won wasn’t the most elegant. It was the one with a week of slack in it.
The plan put the website live at T minus eighteen days, nine days ahead of the CRM cutover, because the site is the riskiest visible piece and deserves a real bake while the old one is still reachable. In the event it went live at T minus nine, inside the slack that existed for exactly that purpose, and the CRM and email cutover was still open behind it when this was written, inside what the slack left. That’s what slack is for. A plan with none would have had the same slippage and nowhere to put it.
The final sweep, T minus three to the contract end, is short and mechanical. It starts with a last delta import of anything changed since the cutover. The suppression export runs again and is diffed against the new platform, so opt-outs taken in the last three days carry across. Everything gets a final checksummed export, and someone opens a restore read. The losses schedule is signed in writing, and every external account, ours included, is revoked. Then the contract ends, and the portal becomes a free-tier archive nobody needs to log into.
A plan that needs the last day to go right has no fallback at all. The design that won wasn’t the most elegant. It was the one with a week of slack in it.
Two clocks that do not stop on the date
Most of the calendar counts down. Two items run on other clocks, and both are easy to miss because neither appears on a renewal invoice.
The first counts down from the day a workflow was deleted. A deleted workflow can be restored from the interface for 90 days. 5 Editor's note 5 HubSpot Knowledge Base, “Delete and restore workflows,” read 16 November 2026. A restored workflow comes back as a new, inactive workflow without its history or past versions. On this portal, four workflows that had unsubscribed, demoted or deleted contacts, 55,422 enrollments between them, had been deleted in the two months before we read the portal. For the pair deleted earliest, the restore window closed within days of the contract end. We chose not to restore them, because every suppression they’d caused was already captured in the export, and the only thing a restore could have saved was the rules each one ran. Had those rules mattered, the real deadline would have fallen days before the contract end, and nobody would have known.
The second counts forward from the last email the portal sends. The FTC’s CAN-SPAM compliance guide says an opt-out mechanism must be able to process requests for at least 30 days after a message is sent, and that it may require nothing more of the recipient than an email address or a visit to a single page. 6 Source 6 Federal Trade Commission, “CAN-SPAM Act: A Compliance Guide for Business.” Requests must be honored within 10 business days, and a sender may not require any information beyond an email address or any step beyond a reply or visiting a single page. Every one of the 16 live emails we exported carried an unsubscribe link and a preferences link pointing at HubSpot system paths on a HubSpot-hosted subdomain, and people were using them: seven opt-outs in the two weeks between our first read of the portal and the export, about one every two days, each through the link in the email.
On this engagement the last possible HubSpot marketing send was a Tuesday at T minus three days, the nurture chain’s final scheduled run. The legal floor therefore ran to about T plus 27 days, well after the portal could honor its own links. Our answer was a one-field unsubscribe page on the new site, with the old HubSpot paths redirected to it, kept indefinitely because old emails sit in inboxes for years and the page costs nothing. Underneath it sat one deliberate choice: the email subdomain stayed pointed at HubSpot until the contract’s last day, so the opt-out links, and every tracked content link in the last three nurture sends to up to about 1,063 enrolled contacts, kept working natively to the end. Moving platforms doesn’t move the obligation. It only moves who has to meet it.
A late signature compresses the work, not the deadline
The proposal assumed signature on a given day, and it arrived three days later. Every phase date stayed where it was, because the contract end doesn’t care when the proposal was signed, and the three days came out of two places: the first phase, which was left with six working days, and the deliverability ramp, whose first send moved from four weeks before the contract end to about three and a half.
That’s the arithmetic the renewal conversation never shows the buyer. Every day between the decision and the start of extraction comes out of the one budget with no slack in it: the weeks a mailbox provider needs before it trusts a new sender. Those can’t be bought back at any rate.
Our notes also recorded what to do if the signature had landed much later: not compress the ramp, but cut over later inside the remaining slack and hold the last days as the sweep, as a deliberate decision rather than a drift. The honest move when time is short is to spend the buffer on purpose and say so. The dishonest one is to keep every date and delete the week that made them safe without saying so.
What you gain when the date becomes a project
The calendar turns a renewal date from something that happens to you into something you schedule, and that changes what the date is worth.
With the calendar in hand, the renewal negotiation has a floor under it. You know what dies on the date and what doesn’t, so the vendor’s “you will lose everything” has a measured answer: the paid features, email sending first among them, and here’s the week its replacement goes live. You know the warm-up takes four to eight weeks, so you know the last responsible day to decide, and it can fall before your contract’s notice deadline. You can say no with a plan, which is a different conversation from saying no with a hope.
The email program comes out the other side in better shape than it went in. A sender that starts with years of bounces and complaints already suppressed, on a subdomain that can’t take the company’s transactional mail down with it, is a stronger deliverability posture than the one being paid for, and it’s in place on day one rather than recovered over a quarter.
The fallback is free. For the length of the contract the old portal is a paid-for safety net under every step of the migration, and keeping it there costs nothing extra. Placing the cutover early is what converts that from a sunk cost into insurance.
And the company ends with a record. Every phase of this exit ends in a test that passes or doesn’t, written into the scope before the work began, and the losses schedule is signed at handover. A year from now nobody argues about what was agreed, because it’s written down with dates on it. To a CFO that’s worth more than the month of fees a quick cancellation would have saved.
The worksheet
Work backward from your own contract end and put the dates in yourself. The intervals come from one exit, and the order is the part that transfers. The sheet leaves out two things: a cancellation deadline, because your notice terms are whatever your contract says and you should read them this week, and any calendar date, because the shape is the point.
| When | What has to be true | Why it can’t wait |
|---|---|---|
| T minus 7 weeks, or as early as the decision allows | Decision made. Extraction begins: suppression history (the contact-level export and the portal-level timeline), the rules inside workflow criteria, pipeline stage labels, custom objects, file attachments, email HTML, workflow definitions. Any deleted workflow you need is restored now. | Everything in this row dies with the portal. The ramp can’t start until the suppression list is in hand. The 90-day restore window runs on HubSpot’s clock. |
| T minus 6 weeks | New sending subdomains registered, one for marketing and one for transactional, with authentication published. DNS access from the customer is the gate. | There’s no default if DNS access goes unanswered. Every day it slips comes out of the warm-up. |
| T minus 5 to 4 weeks | Suppression loaded into the new platform, then subscribers. First warm-up send, engagement-ordered, widening weekly. | Four to eight weeks to production volume, and four is the floor. A problem has to surface while the old sender still works. |
| T minus 24 days | First phase accepted: a written inventory of what was extracted, someone at the company reads a record back from it, both subdomains authenticated with a test send from each. | A backup nobody has opened is a claim. This phase can’t be done after the date. |
| T minus 18 days | Website live on the new platform, redirects in place, old tracking removed. | The riskiest visible piece gets a bake while the old site is still reachable. |
| T minus 13 days | Email program rebuilt and test-sent through its full handoff. Full rehearsal: every form submitted end to end, then every leg deliberately broken to confirm the failure is visible. | A failure nobody is watching still has to announce itself before the fallback is gone. |
| T minus 9 days | CRM and email cutover. Work frozen in the old portal. New workflows switched on. | The old portal is the only fallback and it’s already paid for. A plan that needs the last day has none. |
| T minus 7 days | Cutover complete. Monitor. | Slack. If the cutover slips, this is where it slips to. |
| T minus 3 days | Last possible marketing send from HubSpot. Final delta import. Suppression re-pulled and diffed against the new platform. | Opt-outs taken in the final days must carry across. The 30-day opt-out clock starts here. |
| T minus 2 days | Final checksummed export. Restore read. Losses schedule signed in writing. | After the date, the archive is the only record. |
| T minus 1 day | Every external account revoked, the agency’s included. The contract isn’t cancelled early. | Marketing sending must be carried by the new platform before the old one goes. |
| The date | Portal drops to the free tier. Data stays. Marketing sending ends. Email subdomain repointed. | Nothing here requires renewing or reactivating. |
| T plus 27 days, 30 days after the last send | Old unsubscribe links still land on a working one-field page. Keep it indefinitely. | The legal floor runs from the last send, not from the contract. |
If you’re still deciding whether to leave, the six-factor framework comes first, and the calculator will tell you what the date is worth in dollars. If you’ve decided, the list of what dies with the portal is the row-one detail this calendar assumes. The calendar itself is the thing between those two: the reason a renewal date can be a project with slack in it instead of a cliff, and the reason the slack has to be put there on purpose.
The research is free. So is the diagnosis. 60 seconds →