Customer acquisition isn’t guesswork.

If you’re running a DTC brand, every visitor to your website represents a probability — not a certainty — of purchase.

Logistic regression is one of the most powerful and practical tools you can use to estimate that probability.

It doesn’t just tell you what happened.
It tells you how likely something is to happen next.

Let’s break it down.

The Business Question

Suppose you want to estimate:

 What is the probability that a website visitor makes a purchase?

We might build a model using:

• TIME = Time on site (measured in 10-minute increments)
• FIRST = 1 if first-time visitor, 0 if returning
• ORGANIC = 1 if traffic came from organic search, 0 if not
• EMAIL = 1 if traffic came from email, 0 if not

 

Besides email and organic search, paid social is another source of web traffic, but it’s not shown. Why?

Because in logistic regression, one category must serve as the baseline. In this case, Paid Social is the reference category. Every other traffic source is compared against it.

The Model

The logistic regression equation looks like this:

Let’s make this intuitive.

• The intercept (-1.5) is the starting point.
• Each coefficient shifts the likelihood up or down.
• The model does not output probability directly.
• It outputs something called the logit.

What Is a Logit?

A logit is simply the log of the odds.

Odds and probability are related but different:

• Probability ranges from 0 to 1.
• Odds are probability divided by (1 – probability).

 

The logit transforms that into a number that can range from negative infinity to positive infinity — which makes the math work cleanly.

To convert the logit into probability, we use the logistic function:

You don’t need to memorize this.

You just need to know:

The higher the logit, the higher the probability.

Interpreting the Coefficients (In Plain English)

 TIME (0.2)

For every additional 10 minutes spent on the site, the log-odds increase by 0.2.

Translation:

More time on site increases the likelihood of purchase.

And because it’s scaled in 10-minute increments, it’s business-friendly. A coefficient of 0.2 per 10 minutes is meaningful. A coefficient of 0.02 per minute would feel abstract.

FIRST (-0.7)

If the visitor is a first-time visitor, the log-odds decrease by 0.7.

Translation:

Returning visitors are significantly more likely to convert.

Usually, a first-time visitor is less familiar with the brand and may still be doing research, while a returning visitor is more familiar and likelier to be closer to purchase.

ORGANIC (0.8)

Visitors from organic search perform better than Paid Social (the baseline).

EMAIL (1.2)

Email traffic performs even better relative to Paid Social.

This aligns with what most DTC brands observe:

• Email = warm audience
• Organic = intent-driven
• Paid social = broader acquisition net

Why Paid Social Doesn’t Appear in the Equation

Because it’s the baseline.

If we changed the baseline to Organic, the equation would change — but the predictions would not.

The model always compares categories relative to one reference group.

This matters because interpretation changes based on your chosen benchmark — but the underlying math stays consistent.

A Practical Example

Suppose a visitor:

• Spends 20 minutes on the site (TIME = 2)
• Is a returning customer (FIRST = 0)
• Came from Email (EMAIL = 1, ORGANIC = 0)

Plugging into the model:

Now convert logit to probability:

So this visitor has roughly a 52.5% probability of purchasing.

That’s actionable.

Why This Matters for DTC Brands

Logistic regression allows you to:

• Score visitors in real time
• Prioritize remarketing spend
• Trigger personalized offers
• Allocate acquisition dollars intelligently
• Compare channel performance on conversion likelihood

 

Instead of asking:

“Which channel converts better?”

 

You ask:

“Given everything we know about this visitor, what is their probability of conversion?”

That’s a much more powerful question.

This Isn’t Just a Math Exercise

Logistic regression isn’t about equations.

It’s about decision-making under uncertainty.

For DTC brands, that means:

• Lower CAC
• Smarter budget allocation
• Higher conversion efficiency
• Better personalization

 

And most importantly:

It transforms raw behavioral data into a predictive asset.

Conclusion

Customer acquisition will always involve risk. But it doesn’t have to involve guesswork.

Logistic regression gives you a structured, measurable way to estimate purchase probability — and once you can estimate probability, you can optimize it. The brands that win in DTC are not the ones that spend the most. They’re the ones that understand their data well enough to deploy capital intelligently. If you can predict which visitors are most likely to convert, you’re no longer just marketing — you’re allocating investment with precision.