Two vantage points
A platform's AttributionAssigning conversion credit to touchpoints. starts from an ad interaction. Google Ads, for example, documents that its conversion columns count conversions that followed interactions with its own ads, that view-throughA conversion credited to an ad that was shown but not clicked, within a set view window. conversions are reported in a separate column for impressions that were not clicked, and that a conversion is reported against the date of the ad interaction rather than the date it happened. Meta's reporting follows a similar logic with its own click and view windows, covered in the Meta attribution settings guide. Both may also add modelled conversionsConversions a platform estimates statistically for activity it could not observe directly, such as unconsented or cross-device journeys. for journeys they could not observe directly. Each platform is answering the question "which of our ads can be connected to this conversion?" and nothing about other platforms enters the answer.
Analytics attribution starts from the site or app. GA4Google's event-based analytics platform. records sessions with a source, medium and campaign, stitches them into a conversion path for each user it can recognise, and applies the property's reporting model to share credit across every channel on the path, including organic search, email, referral and direct. It reports the key event on the day it occurred. What it cannot see is anything that did not produce a recorded arrival: an impression that was never clicked, a click from a person who declined consent, a visit on a device it could not connect to the converting one.
Neither vantage point is more honest. The platform's view is complete about its own ads and blind to everything else; the analytics view is complete about arrivals and blind to exposure. The platform versus analytics discrepancies guide catalogues every reason the two diverge; the Google Ads versus GA4 comparison works through the most common pair in detail. This page stays on the attribution logic and on what to do with each number.
The two systems side by side
Ad-platform attribution vs Analytics attribution
| Row label | Ad-platform attribution | Analytics attribution |
|---|---|---|
| What it observes | Its own impressions, clicks and engaged views, plus conversions sent to it by tag, API or import | Arrivals on the site or app from any channel it can classify, plus on-site behaviour |
| Scope of credit | Only the platform's own ads; other channels on the path do not exist to it | Every channel that produced a recorded session, including organic, email, referral and direct |
| View-through | Counted, usually in a separate column, within a view window | Not observed; an impression that was not clicked leaves no trace |
| Window | Set per platform, sometimes per conversion action or ad set; click and view windows differ | Set per property as a lookback window for key events |
| Model | The platform's own, often data-driven or last click, applied within its scope | The property's reporting model, applied across channels |
| Conversion source | Platform tag, server-side API, offline import or a linked analytics key event | Key events defined in the property, from tags or measurement protocol |
| Date reported against | The ad interaction, so figures for a past period keep rising as conversions land | The day the conversion happened |
| Unobserved conversions | Modelled in, where consent or browser restrictions hide them | Modelled in some reports, blank in others; behaviour documented per report |
| Overlap with other platforms | Ignored; the same conversion can be claimed by several platforms | Impossible by construction; each conversion is credited once |
| What it drives | Automated bidding and in-platform optimisation | Cross-channel reporting, on-site analysis, audience building |
| Who controls the rules | The platform, with settings exposed to the advertiser | The property owner, within the documented options |
| Best question it answers | "Which of our campaigns and ads are associated with conversions?" | "Across all channels, how did converting visits arrive?" |
Double counting across platforms
Suppose a person sees a video ad on one platform, clicks a social ad on a second, searches the brand on a third and buys. Each platform looks for its own interaction within its own window and finds one, so each counts the order. The video platform records a view-through conversion, the social platform a click-through conversion and the search platform a click-through conversion. Three conversions have been reported for one order. The analytics property, applying one model to the path it can see, credits the order once, split between paid social and brand search; the unclicked video never appears.
This is not an error in any platform. It is the inevitable result of each platform answering its own question honestly. The sum of platform conversions is therefore a figure with no meaning, and the ratio of that sum to the number of real orders rises with every platform added to the mix and with every view window widened. Teams that compute return on spend per platform and then total it are adding overlapping claims, which is why a set of individually profitable platforms can coexist with a business that is not growing.
Analytics attribution removes the overlap but introduces its own bias. Because it can only credit a recorded arrival, channels whose influence arrives as a view, as a word to a friend or as a later brand search are systematically under-credited, and direct or brand search are over-credited. Platform figures overstate; analytics figures under-credit upper-funnel work. The honest position is to hold both, label both, and reach for an IncrementalityConversions caused by marketing. test when a decision depends on which is closer to the truth.
Three platforms, one month of orders
An online cookware brand closes a month with 880 orders in its order system. Its search ads platform reports 420 conversions, its social platform 380 (of which 140 are view-through) and a video platform 150, almost all view-through. Added together the platforms claim 950 conversions, more than the orders that exist. GA4, on paid and organic last click, shows 900 purchase key events, close to the order count once a handful of test orders and refunds are allowed for; it credits 310 to paid search, 190 to paid social, 220 to organic and brand search, 110 to email and 70 to direct, and credits nothing to video. The channel lead builds a reconciliation table with a row per channel and three columns: platform figure, GA4 figure, gap explained. The paid search gap of 110 is mostly brand-search clicks that GA4 assigns to organic brand terms and conversions landing on earlier click dates; the social gap of 190 is view-through plus modelled conversions; the video gap is entirely view-through. Spend decisions for video go to a geo holdout rather than to either figure.
Because analytics cannot see an unclicked impression, the view-through versus click-through conversions comparison recommends reporting the two kinds of platform conversion on separate lines.
Which to use for which decision
The two systems are tools for different jobs, and the job decides the number.
| Decision | Use | Because |
|---|---|---|
| Choosing bids, budgets and creatives inside one platform | The platform's attribution, left on its default model | Automated bidding optimises towards it; judging it on another number is incoherent |
| Comparing channels against each other | Analytics attribution under one model, held constant | Each conversion counted once, same rules for every channel |
| Reporting marketing results to the business | Orders or revenue from the system of record, with the analytics split as detail | Neither platform nor analytics is the ledger |
| Explaining why a platform and analytics differ | A reconciliation table, one row per channel, gap explained | Differences are structural and recur every month |
| Deciding whether a platform's spend is worth keeping | An incrementality test, informed by both figures | Neither attribution system observes the counterfactual |
| Judging conversion rate (CVRShare of visits or clicks that convert.) by channel | Analytics sessions as the denominator, platform clicks noted separately | The two denominators are different units |
The working habit that holds this together is a reconciliation table that appears in the same report every period, with the platform figure, the analytics figure and the explained gap for each channel. The author's paid media analytics expertise page describes that reporting routine, and the attribution hub sets out the models and the experiments it relies on.
When the analytics column of that table needs a rule the interface no longer offers, or conversions defined with data the interface lacks, the tutorial on building an attribution model in SQL rebuilds rule-based credit from the BigQuery export.
For a cause-by-cause diagnosis of the over-count that appears when platform conversions are summed in a monthly report, see why channel conversions exceed real sales.
Common mistakes
- Adding platform conversions together. The total counts the same orders several times.
- Expecting analytics to validate a platform. Different vantage points, different questions.
- Judging a platform on analytics last click alone. Views and assisted touches vanish.
- Comparing a closed month in analytics with an open month in the platform. Platform figures keep rising.
- Changing a platform's window and treating the jump as growth. The rule changed, not the business.
- Reporting a channel's return without naming the attribution source. The number is meaningless without it.
- Letting view-through windows widen quietly. Each widening lets the platform claim more of the same orders.
The Meta attribution settings guide and the Google Ads versus GA4 comparison show how these differences play out for the two platforms most teams reconcile first.
