Cold emails with an exclamation mark drew half the interested replies of those without, across 1.1 million first emails

Findings

1. Of 1,116,744 first cold emails sent from 6 October 2025 to 9 October 2026, 8.8% (98,668) carried at least one of 102 common spam trigger phrases in the body.
2. Emails with one or more trigger phrases drew human replies at 0.75% (741 replies), against 0.69% for emails with none (6,979 of 1,018,076), and 0.73% against 0.62% with the largest sender removed.
3. Emails with an exclamation mark drew human replies at 0.54% (191 of 35,192) and interested replies at 0.10% (34), against 0.70% and 0.20% for emails without one.
4. Emails with an urgency phrase such as "act now", "limited time" or "urgent" drew interested replies at 0.11% (14 of 12,933), against 0.20% without, and human replies at 0.60% against 0.69%.
5. Hype and guarantee words ("guaranteed", "100%", "amazing") went with more human replies, 0.80% (425 of 53,409) against 0.69%, and slightly fewer interested replies, 0.18% against 0.20%.
6. Emails with two trigger phrases drew 0.09% interested replies against 0.20% for emails with none, a result that rests on 5 interested replies (5,837 emails) and is directional only.
7. Pushy calls to action ("click here", "buy now") and scam-style phrases ("dear friend", "this is not spam") each appeared in about 0.1% of first emails, too few to report a rate.

How to read these numbers. A human reply is any reply from a person, refusals included, and an interested reply is the part of those that the sending platform labeled as possibly interested. Each email counts once: it is the first email one recipient address got from one sending workspace, and a reply counts if a human reply came back from that address. A reply rate mixes two things, whether the email reached the inbox and whether the reader wanted to answer, and these data cannot separate them.

Trigger words and the number of replies

hype and guarantee words0.8 per 100 emails

money and free-offer words0.75 per 100 emails

no trigger words0.69 per 100 emails

urgency words0.6 per 100 emails

Source: Carroway Research from Sales.co sending data, first emails sent 6 October 2025 to 9 October 2026. Groups with fewer than 2,000 emails, or more than 45% from one sender, are left out. Counts are in by-trigger-word-group.csv.

Most first emails in these programs carried no trigger phrase at all: 1,018,076 of 1,116,744 (91.2%). Of the rest, 92,476 carried one phrase, 5,837 two and 355 three or more. Emails with trigger phrases drew human replies at a slightly higher rate than emails without them, 0.75% against 0.69%, and the order held with the largest sending workspace removed (0.73% against 0.62%).

By group, money and free-offer phrases ("free trial", "for free", "free audit") appeared in 2.9% of first emails and drew human replies at 0.75%. Hype and guarantee phrases appeared in 4.8% and drew 0.80%. Urgency phrases were the one group to go with fewer human replies, 0.60% against 0.69%, and with the largest sender removed the order held at 0.60% against 0.63%.

The checker counts free offers and leaves out the bare word "free", because "feel free to reply" is ordinary writing in a sales email.

Interest, exclamation marks and capitals

none0.2 per 100 emails

10.2 per 100 emails

20.09 per 100 emails

Source: Carroway Research from Sales.co sending data. The two-word row rests on 5 interested replies; 3 or more words (355 emails) is left out. Counts are in by-trigger-word-count.csv.

Interested replies tell a different story from the reply count. Emails with an exclamation mark drew half the interested replies of emails without one, 0.10% (34 of 35,192) against 0.20% (2,157 of 1,081,552), and fewer human replies too, 0.54% against 0.70%. With the largest sender removed the human rates were 0.54% and 0.63%.

Emails with an urgency phrase drew interested replies at 0.11% (14 of 12,933) against 0.20%. Emails with one trigger phrase matched emails with none on interest (0.20% each), and emails with two drew 0.09%, a result that rests on 5 interested replies.

Emails with one or two all-caps words, counting only words of four or more letters and leaving out common business acronyms such as SAAS, HVAC and ASAP, drew human replies at 0.63% against 0.70% with none, and 0.62% against 0.63% with the largest sender removed, so the gap mostly came from one sender. The 2,201 first emails with a link drew 0.64% human replies against 0.69% without one.

Method

Source. Sales.co's sending database: the text of each cold email sent by 57 sending workspaces on its platform, and the replies its sending tool labeled. A workspace is one client's outbound program or Sales.co's own. The data were read on 24 October 2026 and cover first emails scheduled from 6 October 2025 to 9 October 2026; the newest 14 days are left out so replies to them can arrive. Sales.co supplies the data, and its email spam checker uses the same phrase list and the same counts.

Unit. One recipient address in one sending workspace, described by the first email that address got from that workspace (1,116,744). Each email body was read once, turned into counts and dropped. No email text, address or client is published.

Trigger phrases. 102 phrases in five groups, listed in trigger-word-list.csv, matched case-insensitive on whole words in the body after HTML is removed. Longer phrases are matched before the shorter ones inside them, so "100% free" counts once. The subject line is not part of this read.

Replies. A human reply is a reply from the emailed address that the sending tool did not label as an auto-reply, out-of-office notice, bounce, warm-up, test or spam. An interested reply is a human reply labeled "possibly interested". Replies from a colleague or another inbox are missed, so every rate is a floor.

Reporting rules. A row is reported only with 2,000 or more emails behind it and no single workspace above 45% of them. The largest workspace sent 16.1% of all first emails, so each rate is also given with it removed. Rows that fail the rules are in the CSVs with their counts and no rate.

Data

totals.csv holds the totals, by-trigger-word-count.csv and by-trigger-word-group.csv the cuts behind findings 1 to 7, caps-exclamation-link.csv the all-caps, exclamation and link rows, and trigger-word-list.csv the 102 phrases. The two chart files hold only the rows that pass the reporting rules.

• Totals CSV, 2 rows, 0.3 KBFirst emails, human and interested replies, sending workspaces and the largest sender's share, for all first emails and with the largest sender removed.
• Replies by number of trigger words CSV, 4 rows, 0.3 KBOne row per count of trigger phrases in the email body (none, 1, 2, 3 or more): first emails, human and interested replies and rates, the largest sender's share, the human rate without the largest sender, and whether the row passes the reporting rules.
• Replies by trigger-word group CSV, 10 rows, 0.8 KBFor each of five groups (money and free-offer, urgency, hype and guarantee, pushy calls to action, scam-style), emails that used the group and emails that did not, with the same columns.
• All-caps words, exclamation marks and links CSV, 7 rows, 0.6 KBThe same columns for all-caps words (acronyms excluded), an exclamation mark anywhere in the body, and a link in the first email.
• Chart data: human reply rate by group CSV, 4 rows, 0.1 KBEmails with no trigger words and each group that passes the reporting rules.
• Chart data: interested reply rate by word count CSV, 3 rows, 0.1 KBRows that pass the reporting rules.
• The 102 trigger phrases CSV, 102 rows, 3.6 KBOne row per phrase and its group, as matched (case-insensitive, whole words).