View-through attribution is the easiest way to make a channel look better than it is. An impression nobody clicked gets credit for an install, and suddenly the channel's ROAS improves on paper while your blended numbers quietly stop adding up.
That does not make view-through wrong. For some formats it is the only honest way to measure assist. But counting it indiscriminately is how teams end up over-investing in channels that are mostly taking credit for installs that would have happened anyway.
This post defines view-through attribution, explains how it inflates paid numbers, and gives a clear rule for when to count it and when to discount it.
What Is View-Through Attribution?
View-through attribution: a method that credits an app install or conversion to an ad impression the user saw but did not click, within a defined view-through window, as distinct from click-through attribution which requires an actual tap.
The contrast matters:
-
Click-through credits an install to a tap. The intent signal is strong: the user did something.
-
View-through credits an install to an impression. The intent signal is weak: the user may have glanced past a video for two seconds.
View-through exists for a real reason. Some formats rarely get clicked, like skippable video, connected TV, and certain social placements. If a click is the only thing you count, those formats look worthless even when they genuinely move people. Whether that assist deserves last-touch credit is the question this post answers.
How View-Through Credit Actually Gets Assigned
The mechanics are simple, and the simplicity is where the trouble hides.
-
The ad network logs an impression served to a device.
-
An install arrives within the view-through window, a lookback period that is usually much shorter than the click window.
-
The network matches the install to the impression and claims credit.
A few things make this messier than it sounds:
-
View-through windows are short by design (often 24 hours or one day), but even a short window catches a lot of organic on a high-impression channel.
-
Self-attributing networks apply view-through differently from one another, so Meta, Google, and TikTok will each count it on their own terms. This is part of why platform numbers and MMP numbers diverge, as covered in why your Meta ROAS and MMP ROAS don't match.
-
The same install can sit inside two networks' view-through windows at once, so both claim it unless an MMP de-duplicates.
For the underlying logic on window lengths, our attribution windows guide is the reference.
Why View-Through Inflates Paid Numbers
The inflation is structural, not accidental.
-
It credits installs that organic would have produced. A user who was going to install anyway happened to see an impression first, and view-through hands the credit to paid.
-
It double counts across networks. Two networks both served an impression, both count the view-through, and your summed paid total now exceeds reality.
-
It breaks blended reconciliation. The clearest red flag: when paid plus organic no longer adds up to your actual installs, view-through credit is usually part of the gap.
This is the same dynamic behind paid installs cannibalising organic growth. Across multi-channel audits, the most common cause of a channel that looks great in isolation but does not move blended growth is a long view-through window on a network already running heavy retargeting. The window quietly absorbs installs that organic would have produced, and the channel reports a flattering number that does not survive contact with a holdout test.
When View-Through Is Worth Counting
There are genuine cases. Count view-through when:
-
The format rarely gets clicked but clearly drives behaviour. Video, connected TV, and some upper-funnel social placements are measured for assist, not last touch. A click is the wrong success signal here.
-
You are explicitly measuring assist, not last touch. Brand and upper-funnel campaigns exist to influence, and a tightly windowed view-through can be a fair proxy for that influence.
-
An incrementality test has confirmed lift. This is the strongest case. If a holdout or geo test shows the channel produces real incremental installs, its view-through credit is earned. See our incrementality testing guide for how to run one.
When to Discount or Exclude View-Through
Discount or strip it out when:
-
The window is long and the network is retargeting-heavy. This is the classic inflation setup. A retargeting network with a generous view-through window will claim large swathes of organic.
-
Your organic baseline is strong. When organic is healthy, overlap is likely, and view-through credit mostly relabels organic as paid.
-
You are reporting to finance. The defensible cut for a P&L is click-through plus only the view-through that an incrementality test has confirmed is incremental. Everything else is a marketing-internal view, not a finance number.
Choosing which model to trust per decision is the same muscle as picking between multi-touch and single-touch attribution: match the model to the decision, not to the channel that looks best.
How to Configure and Reconcile View-Through in an MMP
A workable setup:
-
Set view-through windows deliberately. Do not accept network defaults. Pick a window that reflects how the format actually influences behaviour, and keep it short unless you have evidence for longer.
-
Separate the two credits. Report click-through and view-through as distinct cuts so you can always see how much of a channel's performance rests on impressions nobody clicked.
-
Validate before you trust. Run a holdout or geo test before treating any view-through credit as real. If the lift is not there, discount it.
How to validate this in your MMP
-
Pull each channel's installs split by click-through and view-through, and look for any channel where view-through is the majority.
-
Re-run your blended reconciliation with view-through excluded and see how much the paid total moves.
-
For any channel that depends heavily on view-through, queue an incrementality test before the next budget cycle.
Tools like Linkrunner expose click-through and view-through as separate reporting cuts with windows you set, so you can see how much of a channel's reported performance depends on view-through before you trust it.
FAQ
What is the difference between view-through and click-through attribution?
Click-through credits an install to a tap on an ad. View-through credits an install to an ad impression the user saw but did not click, within a view-through window. View-through carries a much weaker intent signal.
How long should a view-through window be?
Shorter than your click window, and short enough to reflect genuine influence. Many teams use 24 hours. The longer the window, the more organic it absorbs, so lengthen it only with evidence.
Why does view-through make my paid numbers look inflated?
Because it credits paid for installs that organic would have produced, and because multiple networks can claim the same impression-driven install unless your MMP de-duplicates.
Should I count view-through conversions when reporting to finance?
Report click-through plus only the view-through that an incrementality test has shown to be incremental. That is the cut a finance team can defend.
How do I know if view-through credit is actually incremental?
Run a holdout or geo-lift test. If the channel produces measurably more installs when it is on than when it is off, its view-through credit is earned. If not, discount it.
The Takeaway
View-through attribution is a tool, not a verdict. Count it for formats that earn assist credit and that pass an incrementality test. Discount it for long windows on retargeting-heavy networks and strip it from finance reporting unless the lift is proven. Above all, keep click-through and view-through as separate cuts so the inflation can never hide inside a blended number.
If you want to see view-through and click-through reported side by side with windows you control, request a demo from Linkrunner. Start by pulling one channel's view-through share this week. If it is the majority of that channel's credit, you have found your next incrementality test.