Cold emails with a percentage statistic get the fewest replies of any segment tested

A common piece of cold-email advice is to put a number in the first few lines: a percentage lift, a conversion rate, something concrete that supposedly signals credibility. This study tested that advice directly by classifying 287,790 cold emails sent between 2026-01-26 and 2026-08-14 into 6 content segments based on what kind of proof element, if any, appeared in the body, and then checking whether each segment replied at a different rate.

The segment built around a percentage statistic (79,293 emails in the sample) replied at 0.8%, against 99.2% that got no reply. That is the lowest reply rate of any of the six segments measured. For comparison, emails that named a specific client replied at 1%, emails built around a case study number replied at 1.5%, emails citing a timeframe stat replied at 1.8%, emails using general social proof replied at 1.1%, and emails with no proof element at all, nothing quantitative, no client name, no case study, replied at 1.3%. The plain no-proof emails outperformed the percentage-stat emails.

Across the whole dataset, regardless of segment, 3,368 of 287,790 emails (1.2%) got a reply and 284,422 (98.8%) did not. Reply rates in cold outreach are low everywhere, so the differences between segments are differences of a fraction of a percentage point on top of an already small base rate. That matters for how to read this finding: the percentage-stat segment is not failing badly in absolute terms, it is failing relatively, against other segments drawn from the same population of senders, subject lines, and target lists.

Why would a percentage make an email less likely to get a reply rather than more? One plausible mechanism is that a percentage statistic reads as a marketing claim, and recipients have learned to pattern-match marketing claims and discard them without reading further. "Increased X by a specific-sounding lift" is a sentence shape recipients have seen thousands of times in email subject lines and landing pages, and it may trigger the same skimming behavior regardless of whether the number behind it is true. A named client or a case study number, by contrast, points to something checkable, a company, a project, and may read as more like a fact and less like a pitch.

An alternative explanation is worth naming: it is possible that percentage stats are disproportionately used in emails that are otherwise weaker, for example templated bulk sends where a sender inserted a stock statistic without customizing the rest of the message. If that is true, the percentage itself might not be causing the lower reply rate, it could be a marker for a certain kind of low-effort email that would have underperformed anyway. This study's segmentation cannot separate the two, because it classifies emails by content, not by how much the rest of the email was personalized. Distinguishing them would require holding the surrounding email constant and varying only the presence of the percentage, something like a controlled subject-line or body test within a single sender's list.

What a company does with this on Monday: if a sales or growth team currently has a template that opens with a percentage improvement claim, this finding is a reason to test removing it, not a reason to assume the number is the problem. The variation across sending groups (reply rates for the "replied" class range from 0% to 4.5% across 33 groups) is wide enough that any single team's result could easily sit outside this pattern by chance. Treat the percentage-stat result as a reason to test, not a settled rule.