We Sent 2,120 Cold Emails to Shopify Brands. 1.56% Replied. Here Is Every Number.

I run Seed, a Shopify app for creator gifting, and I also cold email Shopify brand founders to get them to try it. Between June and September 2026 we sent 2,120 of those emails from a self-hosted outreach engine that logs every send, open, click, reply, and bounce to our own database. Nobody in this category publishes real cold email numbers for DTC brands, so here is every one we have, with the count behind each one, as of 2026-09-14.

This is not a "cold email benchmarks across all industries" report pulled from a sequencer vendor's aggregate dashboard. It is one operator's actual send log, for one specific audience: Shopify DTC brand founders and operators, contacted about one specific product. Treat it as a real data point for that audience, not a universal law.

Key takeaways

• 1.56% of 2,120 cold emails sent to Shopify brand founders got a human reply (33 replies).
• 87.9% of those 33 replies came from the very first email in the sequence, not a follow-up.
• From follow-up 3 onward, reply rate was 0% across 476 emails sent.
• Median time from send to reply was 0.28 days, about 6.7 hours, across the 33 replies.
• A warm list of 85 past-gifted creators replied at 20.0%, about 25x the 0.79% rate on 2,035 genuinely cold sends.
• 1.23% of all 2,120 sends bounced (26 bounces).
• 67.0% of sends (1,420 of 2,120) showed an "opened" pixel fire, a number known to be inflated by Apple Mail's pre-fetch, not a trustworthy open rate.
• Click tracking recorded a flat 0% across all 2,120 sends in this period, a broken metric on our side, not a real 0% click rate.
• Reply rate varied by source: 0.36% on a 1,677-email cold prospecting list versus 20.0% on an 85-email warm list, a 56x spread.
• We emailed 742 distinct people across 639 distinct company domains.
• A burner-mailbox probe run on 2026-09-09, before any address reached a real sender, found 8 of 88 fresh addresses (9.1%) were already dead mailboxes.
• Only 18 of 742 contacted people (2.4%) unsubscribed.
• A 16-day test of a "Re:" prefix on the first email's subject line saw a 0.23% reply rate (n=428), versus 3.55% (n=141) without it in the rest of the period, though the subject change shipped bundled with a body-copy rewrite, so it is not a clean subject-only test.
• Monthly reply rate ranged from 0.71% in June (n=1,269 sent) to 6.82% in August (n=44 sent, a small sample).

What reply rate should you expect from cold email to Shopify brands?

Across 2,120 cold emails we sent to Shopify brand founders and operators between June and September 2026, 33 got a human reply, a 1.56% reply rate. Looked at by person instead of by email, 3.91% of the 742 distinct people we emailed eventually replied, across up to eight touches each, and 87.9% of those replies came from the first email alone. Both numbers are correct, they just answer slightly different questions.

The two numbers that are not trustworthy here are opens and clicks. Apple Mail's privacy protection pre-fetches the tracking pixel on every email the moment it arrives, whether or not a person ever opens it, so our 67.0% open rate mostly measures how many recipients use Apple Mail, not how many people read the email. Clicks sat at exactly zero across all 2,120 sends in this period, which is a tracking gap on our end, not a real 0% click-through rate. Replies and bounces are the two numbers we trust, because both require a real mail server on the other end to act.

Do follow-ups work, or is it all in the first email?

Mostly the first email, in our data. 87.9% of our 33 replies came from email 1. Follow-up 1 added another 9.1% of replies, follow-up 2 added 3.0%, and every touch from follow-up 3 onward, 476 emails sent across five more follow-up stages, produced zero replies. If your first email does not land, the data here says more follow-ups are unlikely to save it.

Caveat worth stating plainly: each stage's "sent" count shrinks because a reply, bounce, or unsubscribe stops a contact's sequence, so later stages are a self-selected group of people who did not respond to anything earlier. That is normal for a cold sequence and is not a flaw in the data, but it means stage 3 onward is, by definition, the hardest-to-reach remainder of the list, not a fresh sample. Median time from send to reply across all 33 replies was 0.28 days, about 6.7 hours, which tells you most replies that happen, happen fast, not after days of thinking it over.

Does a "Re:" subject line change anything?

For a 16-day window, 2026-06-23 to 2026-07-09, we put a literal "Re:" prefix on email 1's subject line to look like a continuing thread. Reply rate in that window was 0.23% (1 of 428 email-1 sends). Outside that window, with a plain subject and no "Re:", reply rate on email 1 was 3.55% (5 of 141 sends). In our data, the "Re:" version did noticeably worse, not better.

The honest caveat: this is not a clean subject-line-only test. The "Re:" prefix shipped in the same change as a rewrite of the email body copy, so we cannot separate how much of the drop came from the subject versus the body. We are publishing the number because it is the real before-and-after from our send log, not because it proves the subject line alone was the cause. If you want to test a fake-reply subject line, test it in isolation, which we did not do here.

How many cold emails bounce if you do not verify?

1.23% of our 2,120 sends bounced overall (26 bounces), but that undercounts the real risk of an unverified list. A burner-mailbox probe we ran against a fresh list on 2026-09-09, before any address reached a real sender, found 8 of 88 addresses (9.1%) were already dead mailboxes. A simple MX-record check would have missed all 8, because catch-all domains accept every address at the mail-server level and only bounce after they have swallowed the message.

Those two rows are not measuring the same thing and should not be subtracted from each other. The 1.23% is bounces from our real sender across the whole period, most of which predates the burner-probe gate. The 9.1% is what a dedicated, disposable mailbox found when it deliberately tried every address on one fresh list first, before any address was allowed to reach a real sender, so it catches dead mailboxes behind catch-all domains that a bounce alone would not flag until after the damage was done. We only have 11 real sends since the probe gate went live on 2026-09-09, with zero bounces among them, which is too small a sample to claim the gate dropped our bounce rate to zero. The 9.1% dead-mailbox figure from the probe pilot is the stronger evidence that verifying before you send from a real address matters.

Warm list vs cold list: how big is the gap?

We define warm as people who already have a real-world connection to us: in this case, 85 creators who had previously been gifted product by a brand using Seed, re-contacted directly. Cold is everyone else, 2,035 sends across a purchased contact dump, a prospected list of DTC founders, and a small hand-picked list, none of whom had any prior relationship with us. Warm replied at 20.0%. Cold replied at 0.79%. That is roughly a 25x gap.

Caveat: every individual source here except the cold prospecting list is a few hundred sends or fewer, small enough that one or two extra replies would move the rate noticeably. The hand-picked list's 0.00% on 76 sends, for example, could easily have landed at 1 to 2% with a slightly different list. The 25x warm-to-cold gap is the most load-bearing number in this section because it is built from our two largest, cleanest buckets, 85 warm sends against 2,035 cold sends, not from the smallest ones.

What did the replies actually say?

We cannot tell you, beyond the raw count. We log whether and when someone replied, not the text of what they said. The 33 reply timestamps live in our database; the actual reply content lives only in the mailbox, which we did not mine for this post. What we can say is that the 33 counted here already exclude out-of-office replies, auto-acknowledgements, and unsubscribe confirmations, which our reply poller filters out before a message is ever marked as a real reply. So 33 is a floor on genuine human responses, not a number padded by automated noise, but we are not going to guess at a positive or negative split we cannot actually measure.

How we collected this

This runs on a self-hosted outreach engine we built, not a third-party sequencer, sending from a real mailbox and logging every event to our own database. A few operational facts matter for reading these numbers correctly.

• Sender identity changed mid-period. On 2026-07-09 we moved sending off a university mailbox that had been suspended and onto a dedicated domain mailbox, warmed up gradually afterward. Sends from the first few days after that switch carry extra warm-up-ramp throttling that later months do not.
• The cold merchant pitch paused on 2026-07-11.Fresh cold outreach to brand founders (the pitch behind most of this post) was switched off that date while we reworked the approach. Warm app-install onboarding nudges and the creator re-engagement sends kept running throughout, which is part of why July and August's blended reply rates skew higher than June's, a higher share of warm sends in the mix, not a sudden improvement in cold performance.
• A burner-mailbox verification gate went live on 2026-09-09. Before that date, a new address only had its domain's mail records checked before a real sender touched it. After, a disposable mailbox we do not care about sends to it first and we only promote it to a real sender if that live delivery survives. See the bounce section above for what this caught on its first run.
• Opens are deliberately excluded as a trust signal.Apple Mail's Mail Privacy Protection pre-fetches every tracking pixel on arrival, inflating open rates for any list with a meaningful share of Apple Mail users. We report it here for completeness because it is a real number in our log, but we do not treat it as evidence of engagement anywhere in this post.
• Exclusions. All numbers above are sends with a recorded `sent_at` timestamp (queued and failed sends are excluded). Reply classification beyond a raw count is not available, for the reason stated above. The "Re:" subject-line comparison conflates a subject change with a simultaneous body rewrite, stated as a caveat in that section.

Full per-stage, per-source, and per-month detail, plus what we could not compute and why, is documented alongside the raw pull. This post reflects a pull run 2026-09-14 against our own production database.

Download the data

Two CSVs, both aggregate, no names or emails in either file: monthly funnel numbers (sent, opened, clicked, replied, bounced, unsubscribed, reply rate by month) and reply rate by sequence stage (email 1 through follow-up 7). Both are licensed CC BY 4.0, free to use with attribution to Seed (seedinfluencers.com); definitions and licence detail are in the accompanying README.