Was this email forwarded to you? Get all the tips and tricks for direct mail dominance beamed down straight to your very own inbox - HERE. 

So,

Here's to looking at (people like) you kid. 

Name that movie! 

Before the end of last year we sent out a beacon for customer and prospect feedback.

We got a variety of responses (We even had one of the recipients (shout out to Joe Cunningham⚡️) do a review of our request email) from both sides of the aisle.

We sweetened the deal with the opportunity to win $500 buckaroos, and the winner (who we selected randomly; we have video evidence) chose to donate the proceeds to his local animal shelter! I mean, 'c'mon! That's incredibly heartwarming. ❤️

You're a legend, Carlos!

But aside from learning that Carlos is a gem 💎 of a human, we also learned this...

Overwhelmingly (and I mean overWHELMINGLY) Marketers are prioritizing prospecting in 2025 over customer retention.

90% of respondents said they were focusing on prospecting in 2025.

I suppose it makes sense since we also heard that 62% of respondents are planning on increasing their budget in 2025 to theoretically fund the search for those expensive, shiney, new customers.

🚨 POP QUIZ 🚨

Besides money, what is fundamental to running successful prospecting campaigns?

Gosh, you're brilliant! YES - lookalike audiences. ✨

And since we're data-driven marketers, we're going to go into detail about how to build powerful and performant lookalike audiences to help you crush your prospecting goals this year.

Let's go.

--------------

What are lookalike audiences?

Here's how our data science team defines a lookalike audience:

If you didn't know, now ya do.

But what does a lookalike audience actually look like?

It can vary wildly.

How to craft a lookalike audience?

Here's the table-stakes approach.

Step 1:

Define some common features that can be used to zero in on your ICP.

For example:

• Demographics: Age, income, education level, occupation, family size
• Geographic data: Location, type of neighborhood, climate (for service-based companies)
• Behavioral data: Past purchase history, website browsing patterns, app usage
• Psychographic data: Interests, hobbies, values, lifestyle choices (for brands like outdoor companies selling sports equipment)
• Technographic data: Devices used, technology adoption rate
• Financial data: Credit score, investment behavior, spending patterns (for luxury brands with high pricepoints)
• Social data: Social media activity, influence score, types of content shared

Step 2:

Mix and match your key audience features and apply it to a prospecting model.

Step 3:

Launch an ad campaign and wait for results that help you optimize over time.

Sending broad-based direct mail campaigns to locate responders

is inefficient and expensive.

Obviously, this is oversimplified, but the framework rings true.

But let's be honest; if it were that easy, we wouldn't be having this conversation.

The real threat to prospecting campaign success is data access and quality.

Not everyone has access to high-fidelity data sets, and many prospecting models are built off of limited feature sets that only scratch the potential of what's possible with today's advanced data science.

Moving beyond basic lookalike audeinces

Here's my amigo, Billy Hayes of the Postie Data Science Team briefly explaining how we approach LALs a bit differently.

“Instead of just seeing prospects as a random collection of people, we use machine learning to separate them into a ranked list.”

Postie’s lookalike audiences rank prospects 

from most likely to buy to least likely to buy.

When you add a ranking system (among other things) to your prospect audience, you've officially entered advanced lookalike territory.

To get there, you need 3 things:

1. First party data
2. Third-party data from providers like Epsilon, Experian, or Acxiom (our top pics)
3. Sophisticated machine learning models monitored by expert data scientists (like Billy ❤️)

Now, we're not naive; like I mentioned earlier not every brand has access to high-fidelity data sets or an in-house data science team.

But, you can partner with one! 💡

That's not a plug, just a fact.

When you do, your first-party data is going to be instrumental in the construction of your LAL models.

So here's 3 things we look for when we're combing through customer's 1PD here at Posite:

Casting a wide(er) net

If the answer is yes to everything above you get a gold star ⭐️ and we can move on to the next step.

If the answer is no, then stop and fix that problem first.

Assuming we are working with reliable customer data we then select our seed audience. Usually, it's a group of individuals who have high LTV, or maybe it's a group of people who have purchased your premium products, or—or—or, the list can go on.

We then use our friends Acxiom, Experian, or Epsilon to construct a comprehensive picture of every household in the U.S. by supplementing your CRM customer data with thousands of other attributes that can be used to train an ML model for your campaign targeting.

Machine Learning Has Entered the Chat

This is where we throw gasoline on the fire. 🔥

With ML models we can go beyond the basic filters I mentioned above.

The table stakes approach is simple enough for a human to analyze, but it's often rigid and can miss new prospects that don't fit neatly in our pre-defined buckets.

Instead, we delegate the analysis of vast data sets and pattern recognition to a robot that allows us to peer into the nuances of customer behavior that was previously impossible for us mere mortals.

The robots can take that job; we don't want it anyway. 🤖

Exactly how the ML models work is a topic we've already covered, and if you are interested, check out this article to jog your memory.

Beyond the pattern recognition here are two reasons our ML models are the bee's knees for building prospecting audiences:

1. Bias and insufficiency removal: We don't just take data at face value. Our process involves carefully removing biased and insufficient data that could skew your results. 
2. Data normalization: To make sure our machine learning models can process all types of data effectively, we standardize it for an apples-to-apples comparison across different data sources and types.

After an ML model like Postie's chews on your data for a bit, out pops a LAL audience that represents individuals (households) across the United States that reflect the nuanced habits and attributes of your seed audience.

This turbo-charges the chances that each recipient will come running at you with their money after you serve them that oh-so-sweet ad campaign you've been cooking up on the back burner while you've been building out your prospecting audience.

Now, in theory, you can complete all these aspects of enhanced lookalike audiences on your own, but if you're even slightly curious to see how easy we've made it, shoot me a note, and I'd love to show ya.

See you next month. ✌️

Ryan 👨🏻‍💻 & Billy 🐶

Marketing Demand Manager

Postie Inc.