Hi friends,

Happy Saturday, Elvis here.

In the last issue, Carolyn discussed the psychological elements behind the PressPulse cold email that achieved a 15% reply rate. If you enjoyed her breakdown, she started a newsletter to do more copywriting analyses like the one you saw, subscribe here for more.

If you missed her breakdown and the original email copy, you can find it here.

This week, we’re moving beyond individual cold emails to explore creating high-impact, hyper-targeted, and hyper-personalized cold emails at scale.

This is probably my favorite topic to write about, and it's the key secret behind achieving a 15% reply rate on my cold emails.

Why Personalization Matters?

Here's why you should personalize each email:

1. Trust is non-existent in cold emails

Think about it. You're a stranger sliding into someone's inbox. They don't know you. They don't trust you.

Personalization changes that. It shows you've done your homework. You care. You're not just another spammer.

Generic emails scream "I don't really care about you."

Personalized emails whisper "I get you. I understand your challenges."

2. Inbox fatigue is real.

Your prospect probably gets 100+ emails daily. Most are junk.

A personalized email stands out like a red Ferrari in a parking lot of gray sedans.

3. Effort matters.

Mass emails are easy. Personalization takes work.

People appreciate effort. It shows you value their time.

Bottom line: Personalization isn't just nice-to-have. It's essential if you want results.

But the biggest challenge is that personalization is hard to scale. It takes a lot of time and can be very costly, often requiring the hiring of numerous virtual assistants.

But that’s no longer true in 2024 - now the answer is in your hands:

With AI and the right approach, you can write cold emails so good that’s virtually indistinguishably for most people.

For example, Carolyn said she was shocked when I told her the email she received was written by AI. Out of the 20k people I contacted for PressPulse, only 3 people asked me if they are AI-driven. (Hint: they also send cold emails themselves)

Getting the personalization right is the toughest part of the puzzle for most people, but the reward is well-worth it.

Entrepreneurs like Alex Hormozi who realized the impact AI made in the outbound space are literally printing money as we speak:

Prerequisites

Before you go further, this is what you should have by now:

• A hyper-targeted list. (Learn from my 3 examples)

• A validated offer (From talking to early users)

• A copy for one person

To craft the initial copy, pick one person from your list. Let's call her Sarah.

What keeps Sarah up at night? What's her biggest win this year?

Now, write your pitch as if you're sitting across from Sarah at a coffee shop. By focusing on one specific person, you can create the most relevant and tailored pitch. This will serve as the foundation for our next steps.

Checked the list? Now let’s scale.

Secret ingredient to AI personalization

When discussing personalization, I'm not talking to BS like "my cousin-in-law went to the same college as you." Instead, personalization should be directly relevant to the problem you're solving and the solution you're offering.

What's the secret ingredient to effective, relevant personalization?

Data.

Just as a Michelin chef can't prepare a meal without basic ingredients, even the best copywriter can't craft a personalized email without adequate data.

If all you know about a lead is their name and company, there's little to personalize.

But if you know details like their company's size, competitors, current OKRs, key clients, current ad campaigns, biggest marketing challenges, and upcoming initiatives…

Personalization becomes straightforward. (You can probably already imagine the impact a targeted email with this level of detail can carry)

Collecting Data - Preparation for personalization

To create personalizations that drive results, you first need to collect data that’s relevant.

Step 1 - Data baked into your list

If you are building a hyper-targeted list like you should be, there's already a lot of data included in the list.

For example, if you found them through a product hunt launch:

• Just launched a new product

• Trying to market the new product

• Level of traction (# of upvote)

• People's reaction to their product (comments)

If you found them on google maps:

• Full address

• All their customer reviews (a gold mine into their business)

If you found them through a HARO feature like I did, then you know:

• They are already investing in PR

• Already a HARO user

• The topic they are interested in

• Their website

• The article they are featured in

• The quote they provided in the article

• (I ended up deciding not to use anything related to the article so I don't give away my strategy. You’ll also need to decide on how much you allow others to reverse-engineer your list-building approach)

Step 2 - Data you can collect (aka data enrichments)

This is where the fun begins.

You can dive deep into your customer's operation and uncover a wealth of public data from different data providers, and use them in your email copy.

For example:

• Serper enables Google searches.

• Snowflake helps you find competitors.

• BuiltWith lets you explore technology stack.

• LeadMagic helps you locate LinkedIn page.

• Apify allows you to crawl any ad library.

• Claygent allows you to research their website.

Individually, these tools might not seem very useful, but when combined, they allow for unprecedented and unique personalization.

Let me give you an example:

• Got website? Find their facebook page.

• Got facebook page? Find their FB ads.

• Got their ad library? Grab their VSL.

• Got their VSL? Analyze the tech-stack they use.

• ...

or:

• Got website? Find blog article.

• Got blog article? Find their search engine ranking.

• Got keyword ranking? Track their SEO performance vs competitors.

• …

The opportunities are limitless. With tools and AI available for almost anything you can imagine. The only limit is your imagination.

This is why I’m a huge fan of Clay. Unlike traditional tools like Apollo where you have a limited amount of data, your lead list is infinitely extensible inside Clay. With the combination of web scraper, AI agents, and access to almost all the enrichment sources on the market, you can easily collect all the data you can think of, then collected even more data based on those new data.

Turn data into personalizations

Now that you have gathered all this valuable data, it's time to weave it into your email copy to create a highly personalized message. Here's how to do it effectively:

The biggest power with AI is its ability to act on and transform unstructured data, and you can use that to turn huge amount of data into very precise email copies.

First, let's be clear on what NOT to do:

Here's the information about a lead:
{{someDetail}}

write a cold email to pitch my PR app.

The output email will sound very robotic because you can't control what exactly the output is, and you'll have no way to improve it because you don’t even know if the result is any good or not. (Unless you want to read all 10k emails each time you change the prompt)

To avoid the AI completely going off-track, you need to give it some guard rails to stay within.

Here’s how to do that (and 3 golden prompts).

How to create personalization with precise control

The best way to instruct the AI is to turn your copy for one person into a template, then instruct the AI to work within the templates.

For example, this is the exact template I used for PressPulse. (You'll also see that this email can't be created without the having precise data about each lead)

So rather than having the AI write the entire emails, I provide it with tons of useful data, then ask it to complete a pre-written template instead.

In the prompts I give it very specific requirements to control the variance in the output. Here are 3 strategies to use in your prompts.

Strategy 1 - Complete my sentence

This is the single most predictable way of using AI to write copies.

Rather than allowing the AI to say anything, you give it half of a sentence and ask it to complete the other half.

For example, by feeding the AI a lead’s entire website and ask it to complete a sentence, we get the first line used in my PressPulse email:

For example for Carolyn, the AI did its research and generated this output: “Journalists from places like Forbes would love to hear from you about your expertise in optimizing web and email marketing ecosystems for 7, 8, and 9-figure businesses”

Strategy 2 - Multiple choice

The second way of getting reliable output is simply give AI choices, and ask it to choose instead of generate.

For example, this is how I dialed in the copy on which media outlet the lead would dream to be featured on.

With this approach you are guaranteed to have one of the results you are happy with, 100% of the time.

Strategy 3 - Hyper-segmentation

The third approach is not exactly generating copies. But rather you can use AI-agents to segment a lead into very specific categories.

For example, I recently helped a graphic design agency owner reach some local print shops owners. We used Claygent to study their website and find out if they are offering graphic design as a service.

Then we created 2 different copies:

• If no, then our value prop is getting them more customers.

• If yes, then our value prop is a flexible designer with lower cost.

Bonus: Prompting FAQ

How much does this cost?

Generating 1000 personalized emails with GPT4o cost me $7.41.

What model should I use?

GPT4o or similar level models as they are most capable in following instructions.

The model doesn’t follow my instructions, what do I do?

Usually this is because you have too many instructions and AI gets confused. In this case, remove the instructions that can be cleaned up programmatically.

For example, if you ask the model to end on a period, you can easily clean it up with a formula, since Clay supports running any Javascript code. Sometimes you can also chain LLM calls to achieve the desired formatting.

That’s a wrap. I hope this gives you enough resources to scale your cold emails with confidence.

I have a few more drafts written around everything that happens after getting a reply, no set publishing date yet but subscribe if you want to get it when they go live.

Until next time,

Elvis Sun

p.s. If you need personalized help building a hyper-targeted outbound engine that brings quality leads on auto-pilot, I offer paid consultation here.