The Same Customer, Three Different Devices, Three Different Analytics Profiles

A landscaping company's owner was reviewing a customer's history before a callback and was confused to see two separate, seemingly unrelated visitor profiles in the analytics platform — one that had browsed the services page on a phone during a lunch break, and one that had filled out the contact form from a laptop that same evening. It was obviously the same person continuing the same research across two devices. To the analytics platform, with no way to connect the two sessions, they were two unrelated strangers.

Why the same person looks like two people to most analytics tools

Most web analytics identifies a visitor through a browser cookie, which is tied to one specific browser on one specific device. The moment a person switches from their phone to a laptop, or from a work computer to a personal one, that cookie-based identity resets, and the analytics platform has no inherent way to know it's the same underlying person unless something else connects the two sessions. This isn't a bug — it's simply the limit of what a cookie was ever designed to track.

Deterministic matching: certain, but limited in reach

Deterministic matching connects sessions across devices using a shared, confirmed identifier — most commonly, a login. If a customer logs into an account, fills out a form with their email, or is identified through a loyalty program on both their phone and their laptop, the business can confidently link those sessions as the same person, because the identifier is exact and confirmed, not inferred. The tradeoff is coverage: deterministic matching only works for the portion of traffic that actually authenticates or identifies itself in some way, which for most small business sites without a login-based product is a fairly small slice of total visits.

Probabilistic matching: broader coverage, built on inference

Probabilistic matching attempts to connect sessions across devices using signals that suggest, without confirming, that two sessions belong to the same person — shared network characteristics, similar behavioral patterns, or device fingerprinting signals combined statistically to produce a confidence score rather than a certainty. Large advertising platforms use versions of this at massive scale, feeding on far more signal than a small business site alone could generate. For an individual small business, running its own probabilistic matching independently is rarely practical or worth building — the technique matters mainly as something to understand conceptually, since it's part of how the ad platforms a business already uses attribute conversions behind the scenes.

• Deterministic: a login, a submitted email, a loyalty ID — exact and confirmed, but limited to visitors who actually identify themselves
• Probabilistic: inferred from shared signals and behavioral similarity — broader reach, but a matter of statistical confidence, not certainty
• Most small businesses will only ever meaningfully use deterministic matching directly; probabilistic matching is mostly relevant as background knowledge about how ad platforms attribute results

You don't need to solve cross-device identity yourself. You need to know it's happening, so you don't mistake fragmented sessions for fragmented customers.

The practical takeaway for a small business

Building genuine probabilistic cross-device matching is out of reach for the large majority of small businesses, and it isn't the useful takeaway here anyway. The useful takeaway is recognizing that session-level analytics data understates real customer engagement whenever a purchase decision spans more than one device, which is common for anything involving research or comparison — a wedding venue, a major home renovation, a used car. Knowing that this gap exists changes how a business should interpret analytics: a low session-to-conversion rate on mobile doesn't necessarily mean mobile visitors aren't converting, it may mean they're converting later, on a different device, in a session the analytics platform can't connect back.

1. Wherever possible, capture a confirmed identifier early in the funnel (an email address at a lower-commitment step, like a newsletter signup or a saved-favorites feature) to enable deterministic matching for at least a portion of traffic
2. Treat cross-device gaps as a known limitation when reading device-segmented reports, rather than assuming the data fully reflects reality
3. For any paid advertising platform reporting device-level attribution, understand that the platform's own probabilistic modeling is already partially compensating for this, and its reported numbers may differ from a simpler on-site analytics count for exactly this reason

Why this matters even without solving it directly

Cross-device identity is one of the areas where a business doesn't need to build a solution, but does need to avoid making decisions on a false premise. A campaign that looks weak in device-segmented reporting purely because of session fragmentation, not because it's actually underperforming, is a mistake worth avoiding before cutting budget. Understanding the mechanism, even without the ability to fully close the gap, is often enough to catch that mistake before it's made.

← Back to all articles