afG
ads for Good
← Back to the blog
Data & TrackingSep 3, 2026

Identity resolution: how one customer stops looking like five

Here's a problem hiding inside almost every marketing report. One person browses your store on their phone during lunch, comes back that evening on their laptop, opens your email a few days later, and finally buys. To you, that's one customer on one journey. To your analytics, it's often four different "users" who have nothing to do with each other.

The work of recognizing that those four are actually one person is called identity resolution. It's the least visible part of marketing measurement and arguably the most important, because everything downstream — attribution, journeys, lifetime value, even basic conversion counts — falls apart if you're counting one person as several. Let's unpack how it works and why it's so hard.

What identity resolution actually is

Identity resolution is the process of connecting all the scattered records that belong to the same person into a single, unified profile. Every time someone interacts with you, they leave a trace: a browser cookie, a device ID, an email address, a customer account, a phone number. Identity resolution is the work of looking at all those traces and figuring out which ones belong together — that the anonymous phone session and the logged-in laptop purchase are the same human.

Done well, it turns a pile of disconnected events into a person with a history. Done poorly — or not at all — you're left with what most analytics tools give you by default: a crowd of strangers who are secretly the same handful of customers, counted over and over.

The identity graph: the map of who's who

The structure that makes this possible is called an identity graph. Think of it as a constantly-updating map that links every identifier back to the real person behind it. This email, that device, this cookie, that phone number — all connected to one node that represents an individual.

The graph is what lets a new piece of information light up an old journey. Someone browses anonymously for three weeks, then finally enters their email at checkout. The moment they do, a good identity graph can connect that email back to those three weeks of anonymous browsing — and suddenly a customer who "appeared from nowhere" has a full backstory. Without the graph, that history is lost; the sale looks like it came from a stranger.

Deterministic vs. probabilistic: two ways to match

There are two broad approaches to deciding whether two records are the same person.

Deterministic matching uses definite, shared identifiers — the same email, the same customer ID, the same phone number appearing in two places. It's high-confidence: if two records share a hashed email, you can be genuinely sure it's one person. The tradeoff is that it only works when you have those shared identifiers to begin with.

Probabilistic matching makes educated guesses from circumstantial signals — the same device type, location, and behavior patterns suggesting two sessions are probably the same person. It fills gaps deterministic matching can't, but it's exactly what it sounds like: a probability, not a certainty. The strongest identity resolution leans on deterministic matching wherever possible and treats probabilistic signals as a supplement, not a foundation — because building your measurement on guesses means building it on sand.

Cross-device: the problem that makes all of this necessary

The reason identity resolution matters so much more now than a decade ago is that people stopped using one device. A single customer moves between a phone, a laptop, a tablet, sometimes a work computer — and each device looks like a different person to tracking that relies on cookies or device IDs.

Cross-device tracking is the specific challenge of following one person across all their screens. It's genuinely hard, because the thing that used to stitch devices together — third-party cookies and cross-site tracking — is exactly what privacy changes have been dismantling. Which is why modern identity resolution leans on first-party, consent-based identifiers you collect directly, rather than the third-party trails that are disappearing.

Why users "disappear" in your analytics

Put all this together and you can finally explain one of the most confusing things in analytics: users who vanish and reappear as new people. Someone clears their cookies — new user. Switches from phone to laptop — new user. Comes back a month later after the cookie expired — new user. Your returning-customer count looks low, your new-user count looks inflated, and your loyal buyers are scattered across a dozen "anonymous" records.

They didn't disappear. Your tools just lost the thread connecting their visits. Every one of those "new users" is often someone you already know, wearing a disguise your analytics can't see through. This is why anonymous-vs-known user counts are so misleading by default: the line between anonymous and known isn't about the customer, it's about whether your system managed to hold onto their identity.

Why this is the foundation of everything else

Notice that every other measurement problem depends on this one. You can't attribute a sale correctly if the touches leading to it are split across five "users." You can't calculate lifetime value if each purchase looks like a first-time buyer. You can't map a journey if the journey is shattered into device-sized fragments. Identity resolution isn't one feature among many — it's the floor the whole building stands on.

That's precisely why we built Chapter around identity first. It resolves each customer to a single, durable identity across every device and touch — using deterministic, consent-based identifiers as the backbone — so that everything downstream (attribution, journeys, lifetime value) is finally counting people instead of fragments. Get the person right, and the rest of your measurement stops lying to you.

Frequently asked questions

What is identity resolution?
It's the process of connecting all the scattered records that belong to the same person — browser cookies, device IDs, email addresses, customer accounts, phone numbers — into one unified profile, so an anonymous phone session and a logged-in laptop purchase are recognized as the same human instead of counted as separate users.
What is an identity graph?
An identity graph is a constantly-updating map that links every identifier — email, device, cookie, phone number — back to the real person behind it. It's what lets a new signal, like an email entered at checkout, light up weeks of earlier anonymous browsing and give a customer a full history instead of appearing out of nowhere.
What's the difference between deterministic and probabilistic matching?
Deterministic matching uses definite shared identifiers — the same email or customer ID appearing in two places — and is high-confidence. Probabilistic matching infers a match from circumstantial signals like device type, location, and behavior — a good guess, not a certainty. Strong identity resolution leans on deterministic matches and treats probabilistic ones as a supplement, never the foundation.
Why do my returning customers show up as new users?
Because your tools lose the thread that connects their visits. Clearing cookies, switching from phone to laptop, or returning after a cookie expires each makes the same person look brand-new. Most of those new users are people you already know — the line between anonymous and known is about whether your system held onto their identity, not about the customer.
Why is cross-device tracking so much harder now?
A single customer moves between phone, laptop, and tablet, and each device looks like a different person. The third-party cookies and cross-site tracking that used to stitch devices together are exactly what privacy changes are dismantling — so modern identity resolution relies on first-party, consent-based identifiers you collect directly.
Why does identity resolution matter for attribution and lifetime value?
Every other measurement depends on it. If the touches before a sale are split across five users, attribution can't credit them correctly; if each purchase looks like a first-time buyer, lifetime value is wrong; if a journey is fragmented across devices, you can't map it. Get identity right and everything downstream finally counts people instead of fragments.
← Back to the blog