What We Learned From Running a Cold Email Competition

Messaging Madness 2025 has concluded. For a month, we voted on what makes a good cold email, comparing the incomparable and trying to capture the essence of relevant messaging.

We created four scenarios with 16 entries per scenario. We selected sixty-four entries to face off in a scenario-based bracket. 130,638 votes were cast. Four skilled competitors won their scenarios and were granted a yearly Hunter subscription. The one and only Anna Sahakyan emerged victorious, winning an additional $1,000.

What have we learned from this frenzy of email copywriting?

First, that circumstances had a powerful impact on the outcomes. What played an essential role in Round 1, where voters needed to decide on 32 different matchups, wasn’t as much of a factor when there were only eight, four, or two emails to compare.

Second, we learned that a cold email can be 50 words or 200 words, and its length doesn't necessarily matter as long as it’s relevant.

Let’s dive into the details.

Methodology

To understand what happened in this tournament, we took a look back at the 64 emails and the votes they received in each round, starting with four scenario-specific rounds and finishing with two rounds played out between the four scenario winners:

• Round 1 (64 emails)
• Round 2 (32 emails)
• Super 16
• Open Rate 8
• Follow-up 4 
• Championship game

We also analyzed how our internal scores played out in reality. 

Before the bracket went live, we manually rated all entries based on five factors, rating each email on a scale from 5 to 15. Based on these ratings, the top 16 of each scenario entered the tournament. 

Now that we have our winners, we examined the predictive power of the scores assigned to these emails.

Additionally, we performed feature engineering by adding multiple quantitative (e.g., word count, paragraph count) and qualitative metrics (e.g., type of CTA) that describe these emails, which helped us understand how they differ.

Needless to say, we shouldn’t blanket apply the results of this tournament to all cold emails. This wasn’t a scientific study, and participants were free to ask anyone from their LinkedIn followers to their moms for votes. 

That being said, some best practices proved relevant despite the unusual format.

Findings

Our predictions were both right and wrong.

Before the tournament, we ranked each entry in five categories, awarding 1-3 points in each:

• Clarity/transparency
• Personalization
• Relevance
• Trustworthiness
• Readability

This was our method for selecting the best entries, but the public votes didn’t always align.

Two of our favorites were eliminated in Round 1. But as the tournament progressed, our scores better predicted the outcomes. We rated the four top spots 11, 15, 13, and 11 points, respectively.

Our explanation here is that as Messaging Madness progressed and there were fewer matchups to vote for, voters paid closer attention to the qualitative factors that we had prioritized in our internal scoring. 

In other words, when faced with rating 32 different matchups, the visual appearance of the email was more important to voters than intricate personalization.

Can we learn anything about cold emails from this? Perhaps the parallel is that if your inbox is cluttered with external emails (and 37% of decision-makers get 10+ cold emails/week), you’re less likely to judge them on their merit, and more likely to look at their superficial qualities.

Personalization and relevance matter

Our belief that relevance is key to writing a good cold email and that personalization is crucial to achieve that was baked into our internal scoring.

This was largely because Messaging Madness launched after our State of Cold Email research, which found that irrelevance was the #1 reason decision-makers ignore cold emails.

Round 1 showed that, when faced with a ton of emails to review, voters didn’t pay as much attention to the content of the emails as they did later down the line. However, in consecutive rounds, personalization and relevance-related factors played a major role.

With the help of AI, we analyzed all emails to check if they contained specific references to the target persona. 

We asked GPT-4 to look for things like mentioning a news item or quote, referring to a specific product they use, or citing public info like a hiring post, industry trend, or review.

Not surprisingly, these indicators of relevant messaging were associated with an average of more votes.

We also checked how many typical personalization tokens (e.g., {Recipient_Name}, {Company_Name}, or simply a made-up name) these emails contained. 

Compared to our internal personalization scoring, which was based on more than just using tokens, using these placeholders didn’t matter as much.

In fact, in Round 1, using placeholders damaged your vote count. Our explanation is that these placeholders look “templatey” and that’s why they lost votes. In a real scenario, as long as the custom attributes are implemented properly, this wouldn’t have a negative impact.

The lessons here are:

• (Obviously) make sure the custom attributes in your emails work so you don’t open with the legendary “Hi name.”
• Deep personalization isn’t a silver bullet. You first need to grab the reader’s attention with superficial qualities—pattern interrupt, a pleasant layout of the paragraphs, relative brevity, etc.—and then show relevance to keep the attention and generate interest.

Brevity and conciseness are two different things.

The tournament revealed that word count has a complex relationship with the success of cold emails.

Especially in later rounds, we found that longer emails prevailed. However, in Round 1, shorter emails performed slightly better.

The #1 and #2 emails were 141 and 236 words long, respectively. They wouldn’t pass a cold email quality check on LinkedIn, but thousands of people still voted for these longer messages.

There’s a structure component to email length, too. Using too many paragraphs correlates negatively with the vote count.

We didn’t find a significant difference in performance between emails with and without bullet points or ordered lists, so formatting had a minimal overall impact. 

Readability, quantified using the Flesch-Kincaid readability test, also didn’t consistently correlate with votes. Flesch-Kincaid grade level measures word and sentence length to determine how difficult something is to read.

The average Flesch-Kincaid score for all entrants was around 9. The four top emails had respective Flesch-Kincaid scores of 10, 10, 8, and 2.

Follow-up Four was diverse.

The final lesson came from taking a closer look at the four emails that came out on top. 

Just like the four scenarios entailed different value propositions and personas, the four winners created significantly different emails.

Conner Eagleton won the Growth Manager scenario with a short email, using just 52 sharp words with a Flesch-Kincaid score of 2. He stood out with his clever phrasing that let him paint quite a full picture with just three lines of text. He even managed to squeeze in some social proof. This is an excellent example of a quick cold email that's focused on probing your recipient's needs.

On the other hand, Felipe Moraes used a long email that's more conversational.

Ryan Rogers, on the other hand, won the Pre-revenue Founder scenario using 236 words, meaning that his email was about 354% longer. His Flesch-Kincaid score was 10! Ryan showed how a friendly approach works well when there's a live event involved. His email is also on the longer end, which helped him provide extensive social proof and multiple arguments in favor of his solution.

Finally, Anna Sahakyan won with 141 words that paid an extensive compliment to the recipient. Her tone was warm, and it's clear that she didn't sacrifice her personal touch for the sake of brevity.

Cold email lessons from the competition

1. There's no one-size-fits-all approach to cold email. Your audience and the goal of your campaign must dictate how you structure your email. SEO outreach must be smooth and can be short because it's a mutually beneficial partnership, and you only need to show yourself as an easy-going, trustworthy partner. But if you're reaching out to COOs in larger companies, you'll need hard facts to convince them that you're worth their time.
2. Since you can never be sure a campaign will work until you try it, you need to look at the right metrics and repeatedly tweak your approach. This also means that you shouldn't burn through thousands of leads with a single send, and instead take a slow approach to give yourself time to pivot.
3. Superficial features (like paragraph count or using an image) can positively influence how someone feels about the message you send. However, decisions are always made based on deeper factors—relevance, arguments, and your offer.

The problem is that it's hard to scale relevance without multiplying the manual effort. And the solution is our upcoming AI Writing Assistant—join the waitlist now.