CTV measurement and attribution: an honest guide for app marketers
CTV has no click. Attribution therefore works differently from mobile: installs are connected to TV exposure through device graphs and incrementality, not click chains. This guide explains what the methodology can prove, what it cannot, and how to set up measurement that survives scrutiny from your finance team.
The no-click problem
In-app performance marketing is built on a click chain: impression, click, install, event. Television breaks the first link — viewers cannot click a TV. Everything about CTV measurement is a workaround for this fact, and understanding the workarounds tells you exactly how much confidence to place in the numbers.
The good news: the workaround stack is mature. Device graphs, incrementality testing and deterministic response signals together produce measurement solid enough for serious budget decisions — as long as you read the output the way the methodology intends.
Device graphs: linking TV to phone
A device graph connects devices belonging to the same household — typically through shared IP addresses and observed co-location patterns. When your ad plays on a television, the graph identifies the phones in that household; installs from those phones within an attribution window are credited to the CTV exposure.
Two consequences follow. Attribution is household-level and probabilistic — modeled, not observed. And the attribution window is a choice, not a fact: longer windows credit more installs with less certainty. Ask any partner which graph and which windows they use before comparing their numbers to anything else.
Incrementality: the stronger proof
Attribution answers 'which installs followed exposure'. Incrementality answers the better question: 'which installs happened because of exposure'. Lift methodologies compare exposed households or regions against matched unexposed controls and measure the difference in install rates, branded search and store visits.
For CTV this is the gold standard, because it sidesteps the click problem entirely. A clean geo-lift test — CTV spend in some regions, none in matched others — produces evidence your finance team cannot argue with, even when it cannot attribute a single specific install with certainty.
Deterministic signals: QR and search
Some CTV response paths are exactly measurable. A QR code in the creative produces scans that track to store visits and installs with click-level precision. Branded store search spikes after flights are observable directly. Vanity paths and deep links add more.
Build creative to generate these signals deliberately — persistent QR, clear spoken call to action — and they become the calibration layer for the modeled numbers: if attributed installs move but QR and search do not, something in the model deserves a hard look.
Setting up your MMP for CTV
Your mobile measurement partner needs to receive CTV attribution data from the buying side — integrations exist between major MMPs and CTV measurement providers, but they must be configured before launch, with agreed attribution windows and event mappings. Installs attributed to CTV should land in your MMP as a distinct channel, comparable against in-app sources.
Define the billable event before spend: the in-app action that represents value — registration, first purchase, subscription start. CTV's contribution should be measured all the way to that event, not stopped at the install.
Reading results honestly
Expect ranges, not point values. A healthy CTV report states attributed installs with a confidence interval, lift metrics against controls, and deterministic counts for QR and search. Be suspicious of dashboards presenting CTV installs with click-level precision — that precision does not exist in the methodology.
Compare channels on cost per incremental billable event, not cost per attributed install. CTV will look worse on naive attribution and better on incrementality; only the second comparison is honest.
The mistakes that corrupt CTV measurement
Four recurring ones. Reading attributed installs as exact counts and cutting campaigns that are actually incremental. Attribution windows set so long that credit becomes meaningless. Measuring CTV only at the install while in-app channels are measured to revenue events — an unfair fight CTV will lose. And changing creative, audience and budget simultaneously, so no result can be explained.
The discipline is the same as any channel: one variable at a time, pre-agreed methodology, and decisions made on incrementality rather than modeled attribution alone.
How measurement is agreed with Mobadly
Measurement approach — attribution windows, device-graph methodology, lift testing, MMP integration — is part of the media plan, prepared individually per campaign and confirmed in writing before spend. Pricing and budgets are likewise agreed per campaign, dependent on audience, GEOs and season.
Register as an advertiser with your app and goals, and the planning conversation starts with what you need to prove — because a CTV test designed around your finance team's questions is the only kind worth running.