View-Through vs. Click-Through Attribution

In March 2026, Meta retired engage-view attribution and replaced it with engage-through. Most advertisers found out three weeks later when their numbers shifted and nobody had a clean explanation. That was the third time in four years the goalposts moved on impression-based credit. And every time, the same comparison gets revisited: view-through versus click-through, as if the only question is which model gives you a more complete picture.

That framing is the problem. View-through and click-through are not two equally valid lenses on the same truth. One of them is built on a data source that is significantly dirtier than the other, and almost nobody says so directly. This article will. It will also explain when view-through is the right tool to use and when it will actively mislead you.

I've spent years in data pipelines and attribution setups. The view-through debate is usually framed as a philosophy question about credit. It is actually a data quality question about contamination. Get that distinction wrong and you'll budget toward campaigns that were never working.

Quick answers

What is the difference between view-through and click-through attribution?

Click-through attribution credits a conversion to an ad the user actually clicked. View-through attribution credits a conversion to an ad the user was served but did not click, provided they convert within a defined lookback window. Click-through requires a deliberate action. View-through requires only that an ad slot rendered somewhere in the user's session history. Those are not the same quality of evidence.

Does view-through attribution inflate conversion numbers?

Yes, structurally. It assigns credit on the weakest possible signal, an impression, so it will always report more attributed conversions than click-through for the same campaign. Some of that extra credit represents genuine assisted influence, particularly for brand and display campaigns. A meaningful fraction of it is coincidence and contaminated data dressed up as influence. The question is not whether it inflates, but by how much and why.

What is a view-through attribution window?

The lookback period after an impression during which a subsequent conversion still gets credited to that view. Meta historically used a one-day view window by default. Google Display defaults vary by campaign type. The shorter the window, the less inflation, because you are extending credit less generously to impressions that may have had nothing to do with the conversion. A 28-day view window will attribute far more conversions than a one-day window, and far fewer of those attributions will reflect genuine influence.

Is view-through attribution accurate?

Less accurate than most practitioners assume. The conversion event itself can be reliable, particularly with server-side collection. The causal link back to a view is not reliable, because the view pool includes bot impressions, fraudulent placements, and ads that rendered below the fold and were never actually seen by a human. You have an accurate conversion event connected to an unreliable cause.

When should you use view-through attribution?

For upper-funnel and brand awareness campaigns where clicks are rare by design, view-through is often the only signal you have. Use it directionally to understand reach impact. Never use it as a primary ROAS input for performance campaigns, and never let it drive bidding decisions without also examining click-through data in parallel. It is a trend indicator, not a budget signal.

How does Meta measure view-through conversions?

Meta logs an impression server-side, then matches a later conversion event to that impression using its own identity graph. As of March 2026, this sits under the engage-through attribution framework, which folds qualifying engagements and views into one credited category. The match relies entirely on Meta's proprietary impression and identity data. You cannot audit it from the outside, which means you are trusting Meta's categorization of what counts as an impression that deserves credit.

What is engage-through attribution versus view-through attribution?

Engage-through is Meta's 2026 successor to engage-view. It broadens the creditable pool beyond passive views to include defined engagement actions. In practical terms, it makes the credited pool larger and harder to benchmark against historical view-through numbers. A relabeling exercise that also moved the threshold. If your view-through numbers jumped in late March 2026 without a corresponding lift in real business outcomes, this is likely why.

Can you turn off view-through attribution in Google Ads?

You cannot fully disable it, but you control the window length and how you read the column. Set view-through windows as short as the campaign type allows. Report click-through and view-through conversions in separate columns, never blended. Never let a blended attribution number drive a smart bidding decision. Separation and skepticism are the only controls you have on the Google side.

Why view impressions are dirtier than click data

Every guide on this topic treats view-through as a generous model and click-through as a conservative one, as if the only difference is how much credit you are willing to extend. That misses the actual issue.

View-through is not just more generous. It is built on a fundamentally worse data source than click-through, and the gap is large enough to change how you interpret the numbers entirely.

Start with what gets blocked. Analytics and pixel scripts are blocked for 25 to 35 percent of real users by ad blockers, privacy browsers, and tracking protection features. That punches holes in click data. But click-through has a rough self-correcting property: when a user's browser blocks the tracking script, it usually cannot fire the click event either. The missing click and the missing conversion tend to disappear together. The model stays internally consistent, just smaller.

View-through has no such symmetry. The impression is logged server-side by the ad platform, independent of whether the user's browser would have permitted a tracking script. So the impression pool retains every served ad, including ones from sessions where the conversion side is invisible. You end up with a credit source that is fuller than the evidence underneath it.

Now add the contamination layer. Of the traffic that does get measured, global invalid traffic runs at 20.64 percent across digital channels (Fraudlogix 2026). On Meta's own properties, the average IVT rate sits at 8.20 percent. Instagram runs at 38 percent. Meta's Audience Network reaches 67 percent IVT (Fraudlogix 2026).

Click fraud gets the attention in most discussions, but bots generate impressions far more cheaply and far more frequently than they generate clicks. A bot does not need to click an ad to create a view. It needs only the ad slot to render. That means the impression pool feeding view-through attribution is proportionally more contaminated by invalid traffic than the click pool, not less.

Stack those two facts. View-through credit is assigned from a pool that is inflated by impressions from unmeasurable sessions and contaminated by bot impressions at a rate that exceeds one in five globally, and reaches two in three on the worst placements. Then a real human converts. The platform finds an impression in their attribution window and assigns credit. Sometimes that impression genuinely influenced them. Sometimes it was a bot-driven render on a low-quality placement that happened to fall inside the lookback window of a person who would have converted organically.

Here is a concrete illustration. An AI startup ran a signup honeypot expecting moderate fraud. What they recorded was 3,000 signups, 77 percent of them fraudulent, with 650 of those accounts traced back to a single device fingerprint. One machine, 650 identities. Before that machine ever hit the signup form, it loaded pages. It rendered ad slots. It generated impressions. If any of those impressions sat inside a view-through attribution window, the reporting credited those ads with influencing conversions that were never human and never real customers.

Multiply one honeypot across every fraud operation targeting your funnel and you see why view-through numbers drift away from business reality. The attribution model is not the primary problem. The data feeding it is.

This error compounds. Bot-contaminated impressions inflate view-through credit. That inflated credit signals that a campaign is performing. You shift budget toward it. The campaign continues buying cheap, bot-heavy placements that generate the fake impressions in the first place. The measurement error does not just sit in a column. It steers money. For a deeper look at how attribution model choice interacts with data quality, the root issue is almost always upstream of the model itself.

The engage-through shift changed the numbers again

Meta's March 2026 transition from engage-view to engage-through attribution is worth treating as its own event, not just a definitional footnote.

Engage-view credited an ad for a conversion if the user had previously engaged with or viewed the ad within a window. Engage-through expands the definition of what counts as a creditable engagement, broadening the pool. For advertisers who benchmark current performance against historical data, this created a discontinuity that looks like a performance improvement but is partly a definitional change.

The practical consequence: if you are comparing view-through attribution numbers from before and after March 2026 on Meta campaigns, you are not comparing the same thing. The credit rules changed. Any trend analysis that crosses that boundary needs to account for the methodology shift, or it will misread the signal.

This is a good example of the broader problem with impression-based attribution. The platform controls the definition of what counts as a view, what counts as an engagement, and what lookback window applies. You cannot audit those definitions from your own data. You are reading a number that Meta produced using rules Meta wrote, applied to impression logs Meta holds. The shadow analytics problem runs through every platform-reported attribution metric, and view-through sits at the most opaque end of that spectrum.

How click-through attribution handles the same data problems

Click-through attribution has its own weaknesses, but they are different in character and, generally, more tractable.

The main issue with click-through is undercounting. Ad blockers, ITP restrictions, and browser privacy features suppress tracking on click events. A user clicks an ad on iOS Safari, the click fires, but ITP shortens the cookie lifetime to seven days, compared to 90 to 400 days with a proper first-party setup. If the user converts on day 12, the click-through model misses the attribution entirely.

First-party tracking architectures fix most of this. Run your analytics and conversion tracking on your own subdomain, with server-side collection and a proper first-party cookie, and you recover most of the sessions that pixel-based tracking loses. That is the first-party analytics approach: survive ad blockers, ITP, and Brave Shields because the tracking runs on your domain, not a third-party script.

Click-through is also an honest signal about intent. A user who clicked was interested enough to act. That intent signal is real. View-through carries no equivalent evidence of intent: the user may have been shown the ad while scrolling past at speed, may have had it rendered below the fold, or may have been a bot. Click-through is conservative. It will undercount. But the conversions it does count are connected to a real action.

The combination that produces reliable attribution is: first-party click tracking with server-side collection for the conversion signal, bot filtering at the data layer before that data reaches any ad platform, and view-through used only as a supplemental directional signal for upper-funnel campaigns. That combination is what the conversion API setup is designed to support. Clean the pipe first, then decide on the model.

What bot filtering does to view-through numbers specifically

Most CAPI and server-side tracking tools forward all events to the ad platform and let the platform decide what to do with them. That means bot-generated conversion signals reach Meta and Google and get used to train the bidding algorithm.

Meta's average IVT rate of 8.20 percent (Fraudlogix 2026) means roughly one in twelve conversion events you send through a standard CAPI setup was generated by invalid traffic. On Instagram placements it is closer to four in ten. On Audience Network it approaches two in three.

When bot conversions reach Meta's system, Meta trains its Lookalike Audiences and bidding algorithms on those signals. The algorithm optimizes toward the behavior pattern of bots, not customers. Your custom audiences reflect the contamination. Your ROAS numbers improve in reporting while your actual customer acquisition cost climbs, because the platform is finding more of whatever generated the bot conversions, not more of your real buyers.

Filtering bot traffic before it reaches the CAPI endpoint changes this. A 361 billion IP database that classifies datacenter IPs, residential proxies, VPN exits, and known fraud networks can flag invalid sessions before the conversion event is sent. The result is that your CAPI sends cleaner signals, your EMQ scores improve, and the algorithm trains on real customer behavior. The fraud traffic validation layer is upstream of the attribution model. Fix the input and both view-through and click-through numbers become more meaningful.

The click fraud problem in Google Ads is the same issue on the click side. The principle is identical: filter before the signal reaches the platform, not after.

View-through versus click-through by use case

Performance campaigns with short purchase cycles. Use click-through as your primary signal. View-through should be reported separately but should not drive bidding. The purchase decision is fast enough that click-to-conversion paths are trackable. View-through will add apparent volume that does not connect to incremental revenue.

Brand and display campaigns. View-through is often the only meaningful signal, because clicks are rare by design. Use it directionally to understand reach impact and frequency effects. Set a short window, one day maximum, to limit coincidental attribution. Do not use it to report ROAS.

Retargeting campaigns. Users in a retargeting pool have already shown intent by visiting your site. Both click-through and view-through will look good because you are fishing in a qualified pool. The danger is that view-through inflates the apparent lift of retargeting by crediting impressions served to people who would have converted anyway. Run holdout tests periodically to measure incremental lift. Attribution numbers alone will not tell you this.

B2B with long sales cycles. View-through attribution is particularly unreliable in B2B because the lookback windows are too short for the actual sales cycle. A lead that saw a LinkedIn ad in January and signed a contract in April will not be connected by a 28-day view window. The LinkedIn Insight CAPI setup is more valuable here than attribution window management. Get the offline conversion data back to the platform, then look at assisted-touch analysis rather than last-view credit.

Ecommerce with multi-platform spend. The most common scenario where view-through misleads. You are running Meta, Google, and TikTok simultaneously. Each platform reports view-through credit independently. The sum of their claimed conversions will exceed your actual conversion count because each platform is crediting the same real purchase. Deduplicated, server-side conversion data sent via CAPI is the only way to avoid this. See the cross-channel attribution setup for how to structure the deduplication.

Feature comparison: attribution signal quality by data architecture

The quality of view-through and click-through data depends heavily on the infrastructure collecting it. This table shows how different architectures affect the reliability of both signals.

The last row describes an architecture that filters invalid traffic before conversion events reach the ad platform. The signal quality improvement feeds both click-through (fewer bot clicks inflating the click pool) and view-through (fewer bot sessions generating fraudulent impression matches).

The data-driven attribution interaction

Google's data-driven attribution model and Meta's Advantage+ learning algorithm both consume conversion signals to weight credit across the full customer journey. View-through events feed into that weighting.

If bot impressions inflate the view-through pool, and that pool feeds the algorithm weighting, the system assigns more credit to impression touchpoints that were never real. That shifts budget toward display and video placements that generate cheap, bot-heavy impressions. The data-driven attribution for smart bidding interaction means the contamination does not stay in one column. It propagates into the bidding strategy.

The fix is the same as everywhere else: clean the signal before it reaches the algorithm. Data-driven attribution on clean data produces useful channel weighting. Data-driven attribution on bot-contaminated data will consistently overweight cheap impression-heavy placements and underweight the channels that generate genuine intent signals.

Last-touch versus view-through: the credit question

Last-touch attribution assigns 100 percent of credit to the final touchpoint before conversion. View-through adds impression touches to the eligible pool for that final credit. The combination of last-touch plus view-through is particularly prone to misattribution.

Consider the path: a user searches Google, clicks a search ad, visits your site, does not convert. Three days later they see a display ad. Seven days after that they search your brand name directly, click a search ad, and convert. Last-touch assigns credit to the brand search click. But if the display impression sits in a view-through window, it may appear as an assist that inflates display ROAS.

None of those models are necessarily wrong as approximations. All of them are unreliable when the impression data contains invalid traffic. The marketing attribution models overview covers the credit philosophy in depth. The data quality question is prior to the model question.

Practical configuration recommendations

Set view-through windows to one day for display and social. Longer windows increase coincidental attribution. A seven-day view window attributes conversions to impressions that had no plausible influence on the decision. One-day windows are still generous. Start there and evaluate incrementally.

Report click-through and view-through conversions in separate columns. Never report a blended number as your primary metric. The separate columns let you see the view-through increment and evaluate whether it correlates with real business outcomes.

Run holdout experiments on view-through-heavy campaigns. The only way to measure incremental lift is to serve no ads to a control group and compare conversion rates. If your holdout group converts at nearly the same rate as the exposed group, the view-through credit is mostly coincidental.

Use the Facebook attribution window optimization framework for Meta campaigns specifically. The default settings favor Meta's revenue, not your measurement accuracy.

Filter bots before sending CAPI events. The view-through problem is partly about the impression pool (which you cannot directly clean) and partly about the conversion pool (which you can). Bot-filtered CAPI events improve EMQ, which improves how Meta matches those conversions back to impressions. A cleaner conversion signal produces cleaner view-through attribution even when the impression pool itself is not fully auditable.

When not to use DataCops

DataCops is the right fit for businesses that need multi-platform CAPI, bot filtering before events reach the ad platform, and a bundled consent solution. CAPI starts at the Business plan at $49 per month. It is not the right fit in every scenario.

If you run Shopify exclusively and your order volume justifies it, Elevar's order-level attribution fidelity and native Shopify integration may be worth the $200 to $950 per month premium. Elevar has years of Shopify-specific engineering that DataCops does not replicate.

If you have an in-house GTM engineer who wants full container control, Stape at $17 per month gives you the sGTM hosting infrastructure to build whatever you need. DataCops is a managed outcome. Stape is infrastructure. Engineers who want to own the stack should use Stape.

If you need SOC 2 Type II certification today for enterprise procurement, DataCops has that certification in progress but not complete. If your procurement process requires it immediately, wait for completion or evaluate Datahash, which covers larger enterprise compliance requirements at custom pricing.

If you spend only on Meta and your volume is below 50,000 sessions per month, Meta's free one-click CAPI (launched April 2026) is a reasonable starting point. It does not filter bots and it only covers Meta, but for a single-platform basic setup it costs nothing.

If your primary analytics need is post-purchase attribution modeling and media mix modeling, Triple Whale at $179 per month annual or Northbeam at $1,500 per month entry are built for that use case. DataCops cleans the signal going into the ad platforms. Those tools analyze what comes out of them. They are complementary, not competing.

The actual question

Every view-through conversion your reporting shows represents an impression that the ad platform says influenced a purchase. Before you trust that credit, ask one question: of the impressions that fed that attribution model last month, what percentage were served to real humans who could plausibly have been influenced by what they saw?

If you cannot answer that with a number, your view-through attribution is measuring the output of an unaudited pool that includes bot renders, below-the-fold placements, and sessions your own analytics never recorded. The model is not the problem. The data feeding it is.

The conversions you attributed to view-through last month, how many can you prove were real humans who actually saw the ad?