Every marketer has a version of the same problem. A customer taps an ad on their phone, installs on their tablet, and buys on their desktop, and the reporting shows three unique, disconnected interactions. New paid and organic AI channels have added even more places for those journeys to start and break. Identity resolution solves this by merging fragmented signals into unified device personas (when the necessary signals are available and their use is permitted), but the term means different things depending on who you ask.
Branch linking and measurement both run on the same identity foundation, Persona Graph. I sat down with Justin Furstoss from Branch’s product team to talk about what identity resolution means at the device level and why it matters for marketers running campaigns.
1. When people say “identity resolution,” what do they usually picture, and how does that differ from what Branch means by it?
Most people picture the marketing-database version: merging emails, customer relationship management (CRM) records, and loyalty profiles into one master customer record, mainly for personalization and CRM. In the linking and measurement world, identity resolution connects related interactions across devices, like a click in a mobile browser, an app install on a tablet, and a purchase on desktop — when an available identifier or permitted attribution method supports the connection. Same term, different identity layer. One cleans up known customer data, and the other connects touchpoints across devices and platforms.
2. And why does identity resolution matter for a marketer running campaigns across phones, tablets, and desktops?
Without identity resolution, each device looks like a different person. In practice, that fragmentation means your data represents a string of disconnected touchpoints instead of one customer journey. So you can end up overcounting users, undercounting conversions, and giving credit to the wrong channels, depending on the signals available and your configuration. Budget then flows to the channel that happened to engage the device, rather than the channel that actually drove the customer.
3. How does Branch actually resolve identities? Walk us through the process — where does deterministic matching vs. Branch’s proprietary models come in, and how you decide a match is confident enough to act on.
Branch uses deterministic matching when an available known identifier directly links a marketing interaction and a conversion, such as a permitted advertising identifier or a developer-provided user identifier.
When a deterministic signal isn’t available, our models can estimate whether a link click and later conversion came from the same device — where platform policies, customer configuration, and user consent status permit its use. Every match gets a confidence score, and the threshold to act depends on the use case. You might, for example, set a lower bar for a deep linking decision than an attribution decision, or require more scrutiny for a channel prone to fraud, like affiliates, than for a major social network.
4. People often misconstrue fingerprinting for probabilistic matching. What’s the difference, and what is Branch’s approach?
Fingerprinting builds a persistent device ID out of hardware and network attributes so you can track someone over time without consent. It’s brittle and it violates platform policy. Probabilistic matching is different in that it’s a one-time, confidence-scored prediction that determines whether two specific touchpoints belong to the same person. As a rule, we use proprietary probabilistic models designed to compare a limited set of signals from a link click and a later conversion for attribution, only where permitted.
5. We’ve talked about what challenges identity resolution addresses, but what use cases and user journeys does identity resolution actually help with?
Identity resolution touches practically everything we do at Branch. It lets deferred deep linking survive the install, so someone who taps a link before they have the app still lands on the right content after downloading it, and web-to-app, email-to-app, and QR code journeys. It underpins cross-device and cross-platform attribution by reducing duplicated user counts and improving lifetime value (LTV) calculations. And it powers reengagement, audience building, and identity-based fraud detection. Basically, it plays a part in any journey where the customer switches surfaces mid-funnel, and that’s most journeys now.
6. To make this more concrete, let’s imagine a customer who taps an ad on a phone, installs on a tablet, and buys on a desktop. What does the marketer see with and without resolved identity, and what does that gap cost them?
Without resolved identity, you’d see three anonymous, disconnected users: one ad click that never converted, another organic install, and a desktop purchase without a source. The ad channel then looks like a failure because it doesn’t appear to generate any revenue. With resolved identity, it’d look like one customer and one journey where the ad drove the purchase, so the ad channel gets credit for the value it actually provided. The cost of the gap is real money that can drain marketers’ budgets: You might cut spend on the channel that worked and put it into whatever channel happened to touch the customer last.
7. Marketers may never “see” the identity layer behind their linking and attribution. Why should they care about it?
Because it impacts every number they trust and rely on. Match rates, deferred deep link accuracy, and deduplicated return on ad spend (ROAS) all sit on top of it. Weak identity means links that drop people on a homepage and reporting that credits installs to the wrong campaign. Resolved identity gives you measurement you can reliably act on.
8. How does Branch’s approach to identity differ from how others in the market handle it?
Branch has more than a decade of experience building linking products and its graph is built on over 3 billion daily devices of observed linking activity. Because Branch runs the linking infrastructure itself, it sees the actual web-to-app handoff moments that others have to infer. Most of the legacy mobile measurement partners (MMPs) grew up as attribution SDKs, so their identity is mostly scoped to install attribution inside a single app’s context. Branch’s graph supports linking and measurement across eligible web and app touchpoints and is continuously refreshed by live link traffic.
9. As signal availability keeps shrinking and new channels like AI-driven search complicate user journeys, how does identity resolution hold up?
Signal loss kills approaches that depend on explicit identifiers, but it actually makes a model-based graph more valuable, because what matters is the quality of the deterministic ground truth you train on. As AI search and agentic surfaces fragment journeys even more, the core question of “is this the same customer?” only gets harder to answer. That makes identity resolution durable infrastructure.
Want to see what resolved identity looks like on your own campaigns? Talk to our team.
