Identity resolution is the unglamorous plumbing that decides whether "the same customer" across email, web, app, and store is actually one person in your data, or five different ghosts. Get it wrong and every downstream number is off.

What it actually is

Identity resolution stitches scattered identifiers, an email, a phone number, a login, a device id, into one persistent view of a person. The map of how those identifiers connect is the identity graph.

Think of it as answering one question at scale: are these two events the same human. Everything from deduplicated customer counts to cross-device attribution depends on that answer.

Deterministic vs probabilistic

There are two ways to make the match, and the difference is confidence:

  • Deterministic matching uses shared known identifiers, like a hashed email that appears in both your CRM and your checkout. High confidence, lower reach.
  • Probabilistic matching infers a link from signals like device, IP, and behavior. Higher reach, but it guesses, and guesses are sometimes wrong.

Most real systems blend the two, leaning deterministic where they can.

Why marketers should care

Identity is the layer under most things you actually want:

  • Deduplicating customers so your counts and LTV aren't inflated.
  • Following a customer across devices without double-counting them.
  • Activating first-party audiences and matching conversions back through CAPI.

Weak identity quietly corrupts all of them at once.

The privacy angle

As third-party cookies fade in Safari and Firefox, identity leans harder on first-party identifiers, especially the hashed email. That's the durable key that works across email, site, CRM, and ad platforms.

It also raises the stakes on consent: an identity graph is powerful, and it has to respect what people agreed to, or it becomes a liability.

Where it goes wrong

The classic failure is over-merging, where a shared family device or a recycled email fuses two people into one profile. Under-merging is the opposite, splitting one loyal customer into several and undercounting your best relationships.

Both distort the numbers, so identity work is never "set once", it needs monitoring like any other pipeline.

Identity resolution is the plumbing nobody demos and everybody depends on, because it decides whether one customer reads as one person. Get the graph right, lean on first-party keys, respect consent, and every metric built on top of it gets more honest.

Want a stronger data analyst role or a raise? Grab the FREE Product Analyst Playbook and get the exact roadmap to your next offer.