Your Ad Platforms have a definition of “New Customer”...and it’s not yours.

A practical guide to defining what “new” actually means for your business, before someone else’s default does it for you.

On my mind this week

I’ve been thinking a lot about new vs. existing customer definitions lately, and this is something that has come up quite often on my radar in the past 12 months from different brands.

What I mean is, some brands are running acquisition campaigns (which, ofc, might have new customers goals) and these campaigns are in fact retargeting their own existing customers.

The campaign looks fine on the surface, CPAs within range, conversions coming in, platform happy. Then someone pulls a more granular report, or a finance person starts asking questions about orders, and it turns out a meaningful chunk of the people being “acquired” already have a purchase history with the brand, some of them are even loyal customers.

And when this comes about, the conversation always pivots to tracking first.

“What did the data folks mess up now?”

Maybe it’s a floodlight tag acting up or the audience list didn’t upload properly… it must be something to do with the tracking. And sure, fair, your data collection mechanism will collect whatever you tell it to… but in my experience tracking is just a small part of it all. Or in fact is a consequence of randomness.

When you pull the thread, the campaign was always going to do this, because nobody had formally agreed on what “new customer” meant before anything was built.

The exclusion logic was a best guess. The suppression list came from a system that defines purchase history differently to the CRM, which defines it differently again to finance. The audience meant to exclude existing customers was perhaps built off 90-day email opens, which misses anyone who bought 2 years ago and never opened an email since.

And this is what I mean with a consequence of randomness, the campaign or tracking wasn’t broken, but was doing exactly what it was set up to do. The setup just had a hole in it that nobody had looked at directly… or in fact, looked at it for what it is.

A business-level problem.

So, in this edition of my newsletter, this is what we are going to discuss.

This newsletter edition is written in partnership with Stape.

The default is always someone else’s definition

When a brand doesn’t define “new customer” explicitly (and most don’t), the definition gets inherited from whatever system is doing the measuring.

Usually, that’s the ad platform.

Google’s new customer acquisition goals, Meta’s customer list exclusions, most CRM systems, they all default to some version of: someone we have no prior purchase or identifier record for.

It could be by first transaction or a cookie, or first known email. This is a data definition, not a business definition. And the difference matters a lot, because data can only see what it has been given access to, and what it has been told to look for.

Say you’re Nike.

I have been buying Nike shoes from Amazon, eMAG (back when I was in Romania), different retail stores and so on, for decades. I owned A LOT of pairs over the years, I follow Nike on Instagram, I’m in Reddit sneakerhead threads and, in general, I know about product drops before they happen.

By any reasonable definition of “customer”, I’m about as existing as it gets.

Now let’s say I see an ad on Insta, click, visit Nike.com for the first time and buy a pair directly.

DING DING DING! Every system on that site will flag me as a new customer.

I’ll enter an acquisition flow, and might get a first purchase 10% welcome discount. I will probably count toward the new customer KPI that someone’s bonus depends on. And at some point I will probably appear in a retargeting campaign designed to nudge cold prospects back into consideration.

For Nike, or any other brand, none of that is useful. All of it costs money, and it’s happening because the system was never told what “new” actually means, so it guessed 🙂

“New customer” is not one thing. It’s at least 4.

And I do understand why this is hard. There’s a lot at stake in how “new” gets defined, different teams have genuinely competing interests in the answer.

An acquisition team wants a wide definition because it makes their addressable audience larger and their numbers look better. A retention team wants a narrow one because it makes their base look more valuable. Finance wants whatever reports cleanest to CAC. Nobody is wrong exactly, but they’re just pulling in different directions, and the definition ends up being nobody’s, or everybody’s at once.

On top of that, the analytics industry has, for a LONG time, pushed us towards treating this as a session-level question by default.

Every major analytics tool, the ad platforms, ships with “new vs. returning” already defined for you. Cookie-based, device-scoped, session-level. It’s right there in the dashboard, it has a number next to it, and it looks like an answer.

So most brands inherit it without ever consciously choosing it. The tool answered the question before anyone thought to ask it at a business level.

Because when we see a name with some numbers attached to it, we don’t rush to question it much, do we?

Classic representation of New vs. Returning Customers in tools.

The result is that “new customer” isn’t a single concept but a cluster of at least 4 different things, and most businesses have collapsed them all into one bucket without realizing they made that choice.

What is actually considered new?

1. New to the brand?

   Someone who has no real prior awareness of or relationship with you. For a niche DTC brand just starting out, this is most of your addressable market. For Nike, for IKEA, for Coca-Cola? This population is genuinely small. And yet, acquisition budgets are often sized as if it were large. The result is that you spend a lot of money convincing people who already love you, that you exist.
   

2. New to the category?

   Someone who has never bought this type of product before. This is genuinely different from a category veteran who’s switching brands. A first-time sneaker buyer needs different messaging, different education, different proof points than someone who owns 20 pairs and knows exactly what they want. Their lifetime value expectations are also different.
   

3. New to the channel?

   An existing customer of the brand, but one who has never transacted through this specific touchpoint: the website, the app, a specific retailer partner. This is probably the most commonly misidentified “new customer” in the wild, and it’s almost entirely a tracking problem. (More on this below)
   

4. New to the purchase relationship?

   Someone who knows the brand, maybe well, has followed it, been gifted a product, researched it for months, but has never personally transacted. They’re not cold, they are not new in any meaningful strategic sense, they just haven’t bought yet.

Each of these profiles has a completely different strategic implication.
Each one needs a different conversation, a different offer, and a different way of measuring success.

A quick, but important detour: a new customer is NOT the same a new user

So, when I started writing this article, I pinged Jim Kultgen (Amplitude) on Slack to ask more about how Amplitude measures new vs. existing customers. And while chatting with him, he brought up a good point that a customer doesn’t mean an user.

So let’s pause a bit on that.

In a tool like Amplitude, and many others for that matter, there is an important distinction to be made about what shows up as a user and a customer. This is an issue that I guess most analytics platforms didn't consciously think about.

First, what is a user?

In product analytics, a user is an entity that triggers events, someone interacts with your software or digital product.

Since we are using this example, Amplitude tracks users through a combination of signals: a device ID (generated by the browser or device), a user ID (which you configure and typically assign at login or signup), and an Amplitude ID that it generates to reconcile the two.

The key thing to understand is that without a user ID being set (meaning, without an authenticated identity being passed to the platform) a “new user” in Amplitude is, at its core, a new device.

Someone clears their cookies? New user. Opens the app on a second device? New user. Same person, 2 browsers? 2 users.

This is not a flaw specific to Amplitude. It’s how all event-based tracking works when there’s no stable identifier to anchor to.

But it might be that the analytics tool is designed with the expectation that you’ll instrument a user ID for logged-in users, which lets it reconcile events across devices and sessions under a single profile. When it’s set up correctly, it’s meaningfully more robust than pure session level web analytics. But even then, it’s still tracking users, not customers.

What’s the difference, Juliana?

A user interacts with the product. A customer has a commercial relationship with the business. In B2C consumer contexts, which is mostly what we’ve been talking about in this article, these 2 things tend to overlap fairly neatly. The person using the app is usually also the person who paid for it or is being acquired as one. So the distinction feels almost academic, lol, but I will give it a try.

In B2B SaaS, it immediately becomes very real. One customer, say, a company that bought 50 seats of your software might have 50 users, none of whom are individually the “customer” from a commercial standpoint. The customer is the account. The users are the people inside it. A “new user” joining that account is not a new customer at all, it’s seat expansion within an existing relationship. Counting them as acquisition is not a good idea.

But even in B2C and consumer SaaS, the edge cases are more common than people want to admit:

• Someone who was brand-aware, researched your product, but never signed up, they’re not a user yet, but they’re not cold either. When they eventually do create an account, every system will count them as a new user. Whether they’re a new customer depends on what your business means by that.

• Someone who trialed your software a year ago, deleted their account, went quiet, and is now back to buy, is technically new, because the account is gone. But they know your product. They made a deliberate decision to return. Are they new? In what sense?

• Someone who had a personal account, let it lapse, and is now signing up again with a new email address, previously a heavy user, showing up as a brand new one.

In any product with meaningful trial or freemium history, or any consumer brand with account churn, these patterns make up a real portion of what gets counted as new user acquisition.

Why this matters for advertising

The retargeting and acquisition problem is the same, just wearing different clothes.
A SaaS company running paid acquisition campaigns wants to reach people who have never used their product, or who lapsed and might be re-engaged. But if their “new user” signal is device-scoped or tied to a recently created account, they’ll end up showing acquisition creative to people who know exactly what the product does, already made a decision about it, and either chose not to convert or actively churned.

And just like with consumer brands, the platform won’t tell you this is happening. The signal going to Meta or Google says “new user.” The bidding strategy optimizes accordingly. The report says you’re growing. Nobody flags that a meaningful chunk of those “new users” had an account fourteen months ago and left because of a pricing issue that’s since been fixed, and probably just needed a very different message, not an acquisition one.

So, to wrap up the detour, even for B2B + users, the fix is the same: define what “new” means at the business level, then instrument that definition in your tracking, not the other way around.

What where someone buys tells you about how they buy

Not all customers are driven by the same motivation and intent. I recently had a conversation at work with a colleague from the CRM team, where we were talking about another brand, that I’m a fan off, and just like in the case with Nike, while I LOVE the brand, I’m not a die hard fan to buy from the website every new drop, or exclusive collections, I will always resort to Amazon to get their shoes, for the right price. That doesn’t make me less brand aware or diminishes my likeness for the product quality or the look, but I’m not motivated enough to pay higher to get exclusives.

This made me reflect a lot about whether, besides attribution and MMM, are we looking enough at the channel a customer chooses to transact through as a signal for their purchasing motivation and intent?

I think the channel someone picks is not random, but in fact it reflects their relationship with the brand, their price sensitivity, and probably a lot about their likely future behavior.

These are some scenarios I thought of:

1. Brand website / Direct

This customer made a brand-led decision. They knew the URL, typed it in, or navigated directly. They are not in comparison mode, or if they were, that decision is already made. They’re willing to pay full price and engage with an experience that’s built around the brand rather than optimized for price discovery.

This customer has high brand affinity. They are, often, the least “new” customer a brand could have, even if it’s their first purchase on the website. Their LTV tends to be higher, their return rate tends to be better, and their brand relationship is already warm.

2. Marketplace buyers (Amazon, and similar)

These customers made a category or product decision first, and then selected from a set of options likely sorted by price, reviews, or delivery speed. Brand loyalty may exist, but it wasn’t the primary driver of the transaction.

I actually fall into this bucket with a lot of brands, including sneaker brands. I buy from Amazon because it’s cheaper and faster, not because I feel a deep connection to the brand’s website experience. And I will absolutely switch if something else comes up in the comparison. So, people like me are a different kind of customer, with a different churn risk, and they should probably be on a completely different journey than the direct buyer.

PS: I will say that the only brand I purchase from the website for, because of my high love for the brand and because I do want the new exclusive drops is Dr Martens 🙂 And always happy to pay the higher price.

3. Physical retail only

So here, depending on the category, this can actually be the most loyal segment of all: habitual and local. Or it can be a fully convenience-driven purchase with no brand intent whatsoever (they were in the store, the product was on the shelf, and the alternative was out of stock). The catch is that retail only customers are often completely invisible to brand tracking systems. Of course, there are ways, offline + online, sure thing. But if there is no offline+online stitching, it means when they eventually do interact digitally, they show up as brand new, despite potentially years of purchase history.

4. Multi-channel customers

Finally, these are people who buy across both direct and retail, or who research on one channel and purchase on another, are almost universally the highest-value segment. They are also the most likely to be misidentified as “new” in any single-channel view, because each channel-specific system only sees its own slice of the relationship. ROPO ftw, right?

Now, the point I am making here isn’t that one channel is better than another (if anything the channel mix requires a separate newsletter and deep-dive that I do not have enough information to write) but that the channel, besides performance data and attribution, has a more meaningful, strategic signal about who this person is and why they showed up. And it requires a bit more qualitative research and thought put into it 🙂

I will also add here something I heard recently from Eddie Aguilar, to make sure AI is mentioned in this article, haha.

He said something along the lines of: “As AI reshapes how brands engage customers across channels, organized identity orchestration becomes the foundation every downstream model depends on. Yet most companies have regressed to last-touch attribution, but because their data integrity is too fragile to support the multi-touch models they know they need. Until brands treat identity resolution as infrastructure rather than a nice-to-have, AI will only accelerate bad decisions faster.”

Aaaaaand yes, why not, let’s address the “session-level” in the room

The technical infrastructure most brands rely on to define customer identity was built for session measurement. Cookies, pixels, and client-side tracking were designed to answer one question: did this user visit this site before? Useful! But it’s not the same question as: is this person new to my brand?

The gap between those two questions is enormous.

A customer who clears their cookies is new again. Switch from mobile to desktop? Bought in-store and visited the website for the first time? Uses a shared household account merged with their partner? Sadly still new, new, new.
Has an ad blocker? possibly never recognized at all.

This just creates a mismatch between what the technology was designed to measure and what the business actually needs to know.

BTW, I was recording a podcast episode today with Simo Ahava, for Standard Deviation Podcast and we were talking about this topic, and he remembered a talk he did in 2015 about the n ways a session can be defined…

I would love to say things have dramatically changed in the last decade but… I cannot seem to be able to, lol!

Simo’s deck url

How does the lack of customer definition looks like in your tracking data

OK so this is the part where I put my analyst hat on for a second. If any of the above resonates and you want to sense-check whether this is happening at your brand, here’s my quick TLDR of things I’d look at first.

1. Suspiciously high new customer acquisition numbers. If the volume doesn’t match what you know about your brand’s actual market reach, you’re almost certainly counting existing customers on new devices, post cookie clearing, or first time on this channel as new.

2. Remarketing audiences full of recent buyers. A standard pixel-based remarketing audience with no suppression logic will retarget people who just bought from you, because the pixel didn’t see the conversion, or it happened on a different device, or nobody set the window correctly.

3. New customer CPA looks fine, LTV is flat. If the cost per acquisition looks healthy but long-term value isn’t growing, a meaningful chunk of your “new” customers were probably already yours, people who would have come back for free.

4. Your CRM and your ad platform disagree on customer counts.

5. Audience exclusions based on a CSV uploaded three months ago. You uploaded a suppression list once, maybe quarterly, and existing customers are still seeing acquisition creative because the list hasn’t been refreshed since then.

Client-side or server-side?

The 5 symptoms above are mostly client-side problems; they happen because browser-based tracking (pixels, cookies, client-side tags) can only see what the browser sees. It doesn’t know your CRM, it doesn’t know this person bought 3 times last year on a different device. It can’t enforce a definition it was never given.

So it guesses, and the guess is almost always “new user.”

I will add a very important caveat to this: There a lot of things you can do with a client-side tracking to improve your targeting, and this is something Simo will go into the podcast episode we recently recorded, will link it back here in the article when the episode goes live.
Server-side is not a silver bullet for better targeting! And the scope of this article is not to argue for that.

So, I asked Dan Murovtsev, Product Manager at Stape, to go into the practical tracking layer of this, specifically, what you should be looking for in your setup to identify where the definition problem is showing up technically, and what a server-side implementation looks like once the business logic is actually in place.

This is what he shared:

There’s only so much that client-side can give you regardless of how solid the definition is. Typically new/existing distinction in the browser comes from 2 places or the combination of those:

Your most straightforward way is of course a cookie. It gets set on first visit/conversion and then is expected to still be there on a returning one. The problem with that, is of course the fact that we live in 2026 with things like Safari ITP, Mozilla ETP, built-in limitations to emerging AI browsers and endless stream of extensions that intentionally (and not so much) either prevent cookies from being set in the first place, limit their lifetimes or clean them up regularly.

When you are running server-side tagging, your cookies gain resistance to a lot of those hurdles and as such even this most rudimentary way becomes more reliable, although this is still far from what we’re looking for

The second source would be your dataLayer and/or your login state. The quality (and sometimes even the presence) of a good dataLater depends on what CMS you are using and whether it exposes such things to the front-end reliably. Don’t even get me started on custom-built sites, cause in my experience a dataLayer in those is hardly ever complete.

Now when you are running server-side, anything with an endpoint can become your “dataLayer”. Literally, if there is an HTTP endpoint - it is available. And in a secure way at that, since no credentials are exposed in the browser. This enables you to compensate for lack of data in the browser as well as pull it from a source (or a combination of multiple) that fits your deterministic definition.

Section powered by Stape

But, who owns this definition? (Usually nobody)

In most organizations, the definition of “new customer” is nobody’s job.

The brand team thinks about new customers in terms of awareness and market penetration. The CRM team thinks in terms of first transactions and email acquisition. The media team thinks in terms of platform audience exclusions. Finance thinks in terms of CAC. And while there is nothing wrong about how each teams looks at what “new” means, the problem is that none of them talk to each other at the definition level.

This results in a business that can simultaneously be running an acquisition campaign to people already in its CRM, a retention campaign to lapsed prospects who never actually converted, and a win-back campaign to customers who never left, they just switched channels.

And there is also a political incentive problem worth naming here…

Acquisition teams want a wide definition of “new customer” because it makes their addressable audience larger and their numbers look better. A tighter definition means fewer people to reach, more pressure to perform, and a smaller number to report on.

Nobody is rushing to shrink their own metrics.

Let’s talk about solutions. What can good look like?

While I do not hold all the answers, as most of you probably guessed by now, the fix is not primarily technical. It starts with a cross-functional conversation that most businesses have genuinely never had:

What do we mean, strategically, when we say new customer?

A useful answer should cover at least these questions:

• What is the minimum prior relationship that makes someone “not new”? One purchase anywhere? One purchase in our own channel? Just awareness? Being in a household with an existing customer?

• Does channel matter? Is a first-time website buyer who has been buying at retail for years a new customer, or an existing customer migrating channels?

• What is the lookback window? A customer who bought once, 7 years ago, are they new again? After what point does someone lapse back into a new-customer-like state?

• What is the unit of identity? Individual, household, device, email address? (This one is harder than it sounds, especially for retailers.) AND SO IMPORTANT.

These questions need to be answered by strategy and brand, not by the data team.

The data team’s job is to encode the definition once it exists. The analytics team builds the audience logic and the tracking infrastructure surfaces the right signals to make that logic work.

This is the optimal approach:
Strategy defines » analytics encodes » tracking implements.

What usually happens instead is tracking defaults » analytics inherits » strategy never looks at it.

My POV on the optimal approach for how to think about tracking a new customer definition.

Once you have the definition, here’s what server-side actually unlocks

Let’s say you’ve done the work and came out the other side with a real definition. You know what “new customer” means for your business. You have the lookback window, the identity unit, the channel logic, all of it.

So, I asked Dan again, what’s next?

If you’re still relying on client-side tracking, your definition logic almost certainly can’t be covered by a browser, security/access concerns will prevent you in most cases.

Server-side unlocks just that. It gives you means to securely pull required data from any sources you have access to and weave that into your tracking, real-time. In practice, it could look like:

• A purchase event that knows who the buyer is. When a conversion fires server-side, you can pass a customer_status parameter, new or existing based on your CRM logic, not the browser’s memory. Google’s new customer acquisition bidding, Meta’s value optimization both can use this signal. And now they’re optimizing for your definition of new, not theirs.

• Audience suppression that stays current. Because server-side events carry first-party identifiers, hashed email, customer ID you can build and maintain suppression audiences that update in near real-time. Someone converts today; they’re in the existing customer suppression list tomorrow. Not after the next quarterly CSV upload.

• Cross-device identity that actually holds. A customer who browsed on mobile and converts on desktop is, to a pixel, two different people. To a server-side setup enriched with authenticated identity, they’re the same person, and if they’re in your CRM, they’re flagged correctly. The “new customer” signal reflects reality.

• Cleaner signal to the platforms, better model performance. The ad platforms are running machine learning models to find more people like your best new customers. If you’re feeding them a signal that’s 30% polluted with existing customers mislabeled as new, those models are learning from bad data. A cleaner signal means the lookalike logic, the bidding strategy, the creative targeting all of it performs better, because the ground truth it’s learning from is actually true.

This is why the order of operations matters so much. Server-side infrastructure without a business definition is just a more expensive pipe that carries the same wrong signal faster.

But server-side infrastructure built on top of a real definition? That’s where the whole thing starts to work the way everyone hoped it would when they set up tracking in the first place.

Section powered by Stape

Finally, please, remember this is a brand problem before it’s a data problem.

It’s tempting to treat this as something that better technology will eventually solve. Cleaner identity graphs, better cross-device matching, server-side enrichment, tighter CRM integration all of these help, and some of them help a lot.

But as I said in the beginning of this article, none of these methods are a silver bullet that can solve your problems. Especially if you skip the internal alignment and skip approaching this from a business POV.

Technology can only implement a definition that exists and cannot create one for you.

A brand that has never decided what a new customer means will not gain strategic clarity from upgrading its tag management infrastructure…

The companies that get this right tend to have a few things in common. Someone owns the definition across channels, not just within their own team’s reporting.
The definition gets revisited as the brand scales, because what “new” means at 50k customers is different from what it means at 500k. And the definition is treated as a business asset, with real implications for budget allocation. (ALSO, cannot stress this enough, this is not a set and forget type of thing, you need to constantly go back to it and make sure it still stands true for your business, growth)

That definition, built intentionally and maintained with some discipline, is what makes the difference between an acquisition budget that actually grows a brand, and one that endlessly recirculates spend within the audience that already knows and loves you.

And that, my friends, is a problem worth solving before you touch a single pixel. 🙂

Until next time,

X

Juliana

PS: I wanted to give a shout to my talented friend Caroline Vidal, who just started a new blog. Her first article is about: Faster Data Prep & Pipelines with Gemini in BigQuery. Make sure you check it out! It’s really good.