When to sunset cold subscribers, without losing anyone who matters

A subscriber who never opens or clicks isn’t neutral. Mailbox providers rate your sending domain as one pool of reputation, not one score per recipient, so mail nobody engages with drags down delivery for the people who do want it. The fix isn’t a blanket unsubscribe. It’s defining cold correctly, giving those contacts one honest chance to come back, and removing the rest.

Key Takeaways

• Domain reputation is shared across every recipient, so cold contacts cost the subscribers who actually want your mail, not just themselves.
• A fixed number of days is a poor definition of cold on its own. Judge it against your own send frequency, not a universal number.
• Opens have been an unreliable signal since 2021, when Apple Mail started prefetching remote content on delivery. Weight clicks more heavily.
• Removing a cold contact from your active list is not the same as recording an unsubscribe, suppressing the address, or erasing the person. Each does something different and reverses (or doesn’t) in its own way.

Why a cold subscriber costs the subscribers who want your mail

Google’s own explanation of Gmail reputation is that it’s “a rating of the quality of domains and IP addresses used to send email,” and that “messages from senders with a high reputation are more likely to be delivered to Gmail’s Inbox instead of to Spam.” That rating sits on the domain, not on any one recipient. A spam complaint from someone who forgot they ever signed up counts the same as one from someone who actively dislikes your product, and it lands on the same score that decides whether your engaged subscriber’s next email reaches the inbox at all. (For the wider checklist, see why emails go to spam and, if Gmail or Yahoo specifically are the problem, their bulk sender requirements.)

Google’s own sender guidelines are direct about the fix: “Consider unsubscribing recipients who don’t open or read your messages.” The same page says plainly that “Google doesn’t track open rates” itself, so that advice isn’t about a metric Google is watching over your shoulder. It’s about the outcome dead weight produces anyway. A contact who hasn’t opened anything in a year doesn’t remember signing up, and when your mail finally annoys them enough to act, marking it as spam is easier than hunting for an unsubscribe link. For any account sending 5,000 or more messages a day to Gmail, that page also sets a hard number: keep the spam rate reported in Postmaster Tools “below 0.30%,” the same threshold Gmail, Yahoo and Microsoft’s bulk sender rules covers for all three providers. Cold contacts are exactly the group most likely to push an otherwise healthy list toward that line.

What “cold” actually means

A fixed day count is the common definition, and it’s a poor one on its own, because it ignores how often you send. A weekly newsletter’s subscriber with no opens in 90 days has ignored roughly a dozen emails in a row: a clear signal. A subscriber on a quarterly product update, at the same 90 days, has had one chance to open, maybe two. The same number means something different depending on your cadence.

Define cold against how often you actually mail, not the calendar: a contact is cold once they’ve missed several consecutive sends, not once a fixed number of days has passed. Monthly senders can reasonably use a 90-day window, since that’s three misses. Someone who sends twice a year needs a different question entirely: how many of the last three or four sends did this person ignore, whatever span of months that covers.

Segmenting by last engagement

Build the segment from two things: contact status and how long since the contact last engaged. A workable rule set is status is subscribed, and no open in the last N days, where N comes from the cadence math above.

Weight clicks in, not just opens. A contact who reads without clicking still looks cold if you only rule on opens, and opens alone are a soft signal. See build a segment with the rule builder for how a rule like this reads in practice.

One practical limit is worth knowing before you pick N: most platforms that track engagement per contact only retain that history for a bounded window. On EmailFig that window is 90 days, which caps how far back a single engagement rule can look (why engagement rules only look back 90 days). A sender who genuinely needs a longer memory, like the twice-a-year case above, needs a tag written by hand or by a scheduled process instead: a rule based on a day count resets its own meaning the moment the calendar moves past what the platform retains.

Trust clicks, not opens

Open-based engagement has been an unreliable signal since 2021. Apple’s own documentation for Mail Privacy Protection says remote content, tracking pixel included, “is privately downloaded in the background when you receive a message” rather than when the recipient views it. Every message that reaches an Apple Mail inbox with the feature turned on registers as opened whether or not anyone ever reads it, and Apple doesn’t publish what share of its users have it enabled, so there’s no reliable discount to apply to your own numbers.

A click doesn’t have that problem. It only fires when a person actually follows a link, and there’s no equivalent background fetch for that. Where your platform reports both, lean on click-based rules for anything that decides whether to remove someone, and treat your open rate as useful only against your own past campaigns, never as an absolute figure or a number to compare against anyone else’s list. See why open rates are inflated for the mechanics. The same engagement signal decides more than your complaint math: it’s also what Gmail weighs for Promotions tab placement.

The re-engagement sequence

Give the cold segment one real chance before removing anyone, and make it a short sequence rather than a single email, since different people notice different sends.

1. One honest email to the cold segment. Ask directly whether they still want to hear from you, with one clear action: a link to click, or a preference center where they can choose fewer emails instead of none.
2. A wait of one to two weeks.
3. A second, shorter email with the same ask, and a plain statement that this is the last one before sending stops.

That shape maps directly onto a wait-then-branch automation: a tag marks someone cold, the first email sends, a wait step holds for the interval, and a yes/no split checks whether they clicked before deciding what happens next. Anyone who engaged rejoins the regular list; anyone who didn’t reaches the second email before the sequence ends. See branch with a condition step for how that split reads, and automation triggers for what starts the flow off a tag.

A frequency offer belongs in that first email alongside the straight ask to stay. Some of the people this segment catches aren’t disengaged from your content, they’re overwhelmed by how often it arrives, and a link to a preference center where they can drop to monthly instead of weekly recovers a subscriber that a binary “still want this?” would have lost. See customize the unsubscribe experience for building that choice into the page. It only works as a genuine option, not a delay tactic: a contact who picks a lower frequency and still doesn’t engage at that pace is answering the same question, just more slowly.

Don’t reach for a recovery-rate number here. How a list was built, how long contacts have gone quiet, and what the email actually asks for all move that number by a lot, and any figure you find quoted online belongs to somebody else’s list, not yours. Run the sequence, then read your own results before deciding what counts as working for you.

When to stop

One re-engagement attempt is enough. If a contact doesn’t respond to a direct, honest ask across two emails with a wait between them, a third or fourth attempt won’t change the answer. It will just look like you can’t take no for an answer. Whoever is still in the segment once the sequence finishes is the group to remove.

Unsubscribe, suppress, delete: three different actions

These get used interchangeably, and they aren’t the same thing.

A cold contact who never took any action of their own didn’t unsubscribe, so recording one on their behalf misstates what happened. Removing them from your active list is the honest description: it stops the mail without inventing a consent event that never occurred. Save suppression for someone who told you, directly or by marking a message as spam, that they don’t want your mail at all, and save erasure for an actual right-to-be-forgotten request. See suppressions: why an address can’t be re-added and erase a contact for exactly what each one changes.

What’s left

A cleaned list is smaller, but that was never the point. What matters is that mail to the subscribers who are still there keeps landing, because the domain carrying it isn’t paying for a group of contacts who were never going to open it again.

List hygiene is one piece of the wider picture; see email deliverability for people who aren’t experts for how it fits alongside authentication, bulk sender rules, and bounce and complaint thresholds.