Attribution When Shoppers Switch Devices Before Buying
Research on a phone, purchase on a laptop is a normal journey that breaks normal attribution. How to read cross device journeys without fooling yourself.
A shopper sees an ad on a phone on Tuesday, reads a review on a tablet that night, and buys on a work laptop on Friday. Most attribution setups see fragments of that story and credit whichever fragment touched the purchase session, which is usually the least interesting part.
Cross device behaviour is not an edge case to be corrected. For considered purchases it is the typical shape of the journey. This guide covers reading it honestly with the tools most teams actually have.
Accept what each data source can and cannot see
Platform attribution sees what happens inside its own logged in world and guesses at the rest. Analytics sees sessions and devices as separate until a login connects them. Orders see the end. None of them sees the whole journey, and a model that pretends otherwise produces precise, false confidence.
Write down, per source, where its vision stops. That short document prevents most cross device arguments, because the team stops asking a source to answer questions its data cannot reach.
Use logged in and post purchase signals to size the gap
Where shoppers log in on both devices, the joined journeys give a sample of true cross device behaviour: how often it happens, how long it takes, and which channels start journeys that end elsewhere. That sample, handled carefully, sizes the blind spot in the rest of the data.
A short post purchase question about where the shopper first heard of the product, asked consistently, adds a second estimate. Neither source is the truth alone. Where they roughly agree, the team can plan. Where they disagree wildly, measurement needs work before budgets move.
Judge channels on their role in the whole journey
Discovery channels will always look weak in last session reporting for cross device purchases, because their work happened on another device days earlier. Cutting them on that evidence removes the start of journeys whose ends the business then misses a month later.
Report channels with a role note beside the numbers: mainly starts journeys, mainly assists, mainly closes. The note, agreed quarterly from the joined sample and survey, keeps budget discussions anchored in how buying actually happens.
Set decision rules that tolerate uncertainty
Cross device measurement will stay approximate. Set rules accordingly: move budgets in smaller steps, hold changes for a full typical journey length before judging, and require two signals to agree before a major cut to a discovery channel.
Record major attribution decisions and the evidence used at the time. When results arrive, the record shows whether the evidence or the decision was wrong, which is how a teams judgement improves instead of just its dashboards.
Related reading
If this topic is useful, these earlier guides go deeper on connected questions: Planning Incrementality Tests You Can Actually Run After Peak, Attribution When Gift Cards Pay: Who Gets Credit for the Sale?, and Server Side Tracking Fixed Consent Loss and Created New Gaps.
Frequently asked questions
Why does last click undervalue discovery channels?
Because discovery often happens on a different device, days before purchase. Last click only sees the purchase session, so it credits the closing touch and misses the start of the journey entirely.
Can cross device journeys be measured exactly?
No. Logged in samples and post purchase surveys estimate them well enough for budget decisions, but no standard setup sees every device switch. Planning should assume a range, not a point value.
How long should a channel change be held before judging?
At least one full typical journey length for that product, often two to four weeks for considered purchases. Judging a discovery change inside a week mostly measures journeys that started before the change.
What is the simplest cross device improvement most teams can make?
Add a consistent post purchase discovery question and review joined, logged in journeys quarterly. Together they size the blind spot cheaply and stop the worst misreadings of channel value.