A conversion path is the ordered list of marketing interactions that a measurement system can link to one person before that person completes a valuable action, such as a purchase, a lead form or an app install. A typical path might read: YouTube view, then generic search click, then email visit, then brand search click, then purchase. The concept exists because single-interaction reporting hides most of that sequence. Last-click reports show only the step that closed the sale, and the channels that opened or sustained the journey disappear from view. Path reports put them back, and they supply the raw material that every multi-touch attribution model, whether rules-based or algorithmic, needs to divide credit.

How a path is assembled

Building a path takes three things: interactions, an identifier and a window.

Interactions are the events a platform records. Google Ads counts clicks and YouTube engaged views, and adds video impressions from YouTube and Demand Gen campaigns when a reporting toggle is switched on, according to its help page on attribution reports. Campaign Manager 360 (CM360), Google's ad server, records every impression and click on an advertiser's ads and calls each one an exposure. GA4 (Google Analytics 4) records site visits plus, where linked to Google Ads, impressions, engaged views, emails, downloads and app usage.

Identifiers tie those events to one person. Cookies, click IDs such as Google's gclid, logged-in accounts and hashed emails all serve. Where the identifier breaks, the path breaks. A person who clicks an ad on a phone and buys on a laptop appears as two paths unless the platform can perform identity resolution across the devices.

Windows bound the sequence. Google Ads attribution reports look back 30, 60 or 90 days before a conversion, with most reports opening at 30, and the setting is separate from the conversion window used for bidding. CM360's Path to Conversion report accepts a window of up to 90 days for both clicks and impressions, and a custom value can only narrow the advertiser's Floodlight window, not extend it. Amazon's reporting uses a fixed 30-day window. In GA4, the lookback window became configurable to any value from 1 to 90 days for clicks on August 11, 2026.

Once events are joined and ordered, the system cuts the sequence at the conversion. Every path therefore has a length (the number of touchpoints) and a duration (the time from first touch to conversion). Google Ads reports both as path metrics: average interactions to conversion, and average days or hours to conversion, measured from either the first or the last interaction.

Where paths appear in the main tools

Google Ads groups the reports under Goals, then Attribution. The Conversion paths report offers two views, "Paths" and "Transition paths", and can be split by device. Path metrics, assisted conversions and a model comparison sit alongside it. Some exclusions matter: model comparison leaves out the Search Partner Network, Gmail and App campaigns, and offline conversions uploaded more than seven days after the event are bypassed by attribution reports.

GA4 calls its version the Key event attribution paths report, a name that reflects GA4's March 2024 switch from "conversions" to key events. Trade guides, including a Search Engine Journal piece by Ameet Khabra from November 5, 2025, describe it as "formerly" the conversion paths report; Google's help page does not record a rename. The report shows "all paths up to 20 touchpoints long" and holds data from June 14, 2021 onwards, according to Google. Its default dimension is the primary channel group, switchable to source, medium or campaign.

GA4 also slices each path into thirds of unequal size. Early is the first 25% of touchpoints, rounded to the nearest whole number; mid is the middle 50%; late is the last 25%. A one-touch path gives everything to late, and paths with fewer than three touchpoints have no mid segment. On an eight-step path, therefore, the first two touchpoints are early, the next four mid and the final two late.

CM360 requires a Floodlight configuration ID and allows up to 100 clicks and 100 impressions per path.

Amazon Ads reports the top five paths ranked by sales across Sponsored Products, Sponsored Brands, Sponsored Display, Sponsored TV and Amazon DSP. Amazon describes the data as "the actual ad events without interpretation", meaning no attribution model is applied.

A worked example

Amazon's launch material, as reported by PPC Land in January 2025, used a fictional brand, KitchenSmart. Its top path combined Streaming TV, video and Sponsored Brands and produced $706,821 from 5,543 purchases, 22% of conversion value. The second path, video plus Sponsored Brands plus Sponsored Display, generated $614,890 from 4,113 purchases, or 13%.

Path data, in other words, reveals that streaming TV appears upstream of a large share of sales, which a last-click report would never show. It does not reveal whether those sales would have happened without the TV spot. That distinction drives most of the disputes below.

Origin and evolution

Path reporting grew out of ad servers. DoubleClick for Advertisers, the predecessor of CM360, logged impressions and clicks against Floodlight conversions and offered path-to-conversion reporting before search platforms did. Google Analytics brought the idea to a mass audience when it launched Multi-Channel Funnels in August 2011, with Top Conversion Paths, Assisted Conversions, Time Lag and Path Length reports, according to practitioner write-ups published at the time.

The second phase was about credit. Google added attribution modelling to DoubleClick in September 2012, and AdWords offered six models in 2016. Paths became inputs to credit rules and, later, to data-driven models trained on converting and non-converting paths, a history covered in PPC Land's explainer on multi-touch attribution.

The third phase was erosion. Apple blocked all third-party cookies in Safari from March 24, 2020, and App Tracking Transparency arrived with iOS 14.5 in April 2021. Each change shortened observed paths, as unlinkable touchpoints fell off the front.

The fourth phase moved paths onto retail and platform dashboards. Google withdrew first-click, linear, time-decay and position-based models from Google Ads and GA4 in 2023, leaving data-driven and last-click options; GA4's own help says the older models were "no longer available as of November 2023". Amazon announced Conversion Path Reporting in beta on January 2, 2025, for North America, and expanded it worldwide on November 13, 2025, with historical data from September 15, 2025.

Why it matters for marketers

Paths decide which channels look useful. Upper-funnel media such as video, connected TV and display rarely close a sale, so they look weak under last click and stronger when the whole sequence is visible. Kochava, a mobile measurement firm, reported that connected TV came before search in 96% of installs that touched both, a vendor finding that illustrates the ordering question paths answer.

Paths also train bidding systems. Google's Journey Aware Bidding, announced in September 2025 for lead-generation advertisers, learns from non-biddable steps along the path rather than from the final conversion alone.

Limits and disputes

The first limitation is visibility. A path contains only what the reporting platform can see. Google's reports show Google interactions plus whatever GA4 collects; Amazon's show Amazon ads only. Television, radio, out-of-home, podcasts and word of mouth leave no trace, so the "first" touchpoint is the first observed, not the first in reality.

The second is truncation. GA4 caps the report at 20 touchpoints, Google Ads at 90 days, Amazon at 30 days and five paths. Business-to-business journeys that run for months are cut short. Privacy thresholds can also fold rare paths into "(Other)".

The third is causation. Path reports describe sequence, not influence. A retargeting ad that appears late in many paths may have reached people who were already going to buy. Incrementality experiments with holdout groups exist to test that, and Ron Berman of the Wharton School has argued that shared credit across paths creates a "team compensation problem" among the publishers being paid.

The fourth is self-reporting. Each platform builds paths from its own logs and grades its own contribution, so Google, Meta and Amazon path reports cannot be reconciled: each sees a different slice of the journey.

  • Customer journey is the full sequence of a person's experiences with a brand, online and offline. A conversion path is the measurable, ad-and-visit subset that a specific platform can tie to a conversion.
  • Conversion funnel is an aggregate model of stages (awareness, consideration, purchase) across many people. A path is one person's actual sequence.
  • Path exploration in GA4 traces page-to-page or event-to-event navigation within a site or app. It concerns on-site behaviour, not the marketing channels that brought the visitor.
  • Attribution model is the rule or algorithm that divides credit along a path. The path is the input; the model is the calculation.

A separate usage exists in website design and conversion-rate optimisation, where "conversion path" means the on-site route of landing page, offer, form and thank-you page.

Recent developments

In January 2026, GA4 added an attribution analysis report in beta that splits data-driven credit into early, mid and late stages, using the same 25/50/25 cut as the paths report. In October 2026, trade reports said the Google Ads Overview page had begun including video impressions in conversion paths and added a "Performance by stage" card. Google also said that from November 2026 Floodlight would use Google signed-in data to recover unattributed cross-device conversions in CM360, Display & Video 360 and Search Ads 360, which should lengthen paths that previously broke at a device switch. As of November 2026, Google had not said how many conversions that would add.

Timeline

  • August 2011 - Google Analytics launches Multi-Channel Funnels, including Top Conversion Paths and Path Length reports.
  • September 2012 - Google adds attribution modelling to DoubleClick for Advertisers.
  • May 2016 - AdWords offers six attribution models, applied to search paths.
  • March 24, 2020 - Safari blocks all third-party cookies.
  • April 26, 2021 - App Tracking Transparency ships with iOS 14.5.
  • June 14, 2021 - Earliest data available in GA4's attribution paths report.
  • November 2023 - First-click, linear, time-decay and position-based models no longer available in GA4.
  • March 2024 - GA4 renames conversions to key events.
  • January 2, 2025 - Amazon announces Conversion Path Reporting in beta for North America.
  • September 2025 - Google announces Journey Aware Bidding.
  • November 13, 2025 - Amazon expands conversion path reporting worldwide.
  • January 16, 2026 - GA4 attribution analysis report with early, mid and late stages enters beta.
  • August 11, 2026 - GA4 click lookback windows become editable from 1 to 90 days.
  • October 2026 - Google Ads Overview page reported to add video impressions to conversion paths.
  • November 2026 - Floodlight begins using Google signed-in data for cross-device conversions.

Summary

Who: Advertisers, agencies and analysts read conversion paths; Google (Google Ads, GA4, CM360), Amazon Ads and other platforms build and report them from their own logs.

What: The ordered sequence of ad interactions and visits a platform can link to one person before a conversion, bounded by a lookback window and used for reporting and attribution.

When: Ad servers logged paths in the 2000s; Google Analytics popularised path reports in August 2011; Amazon added top-five path reporting in January 2025 and took it worldwide in November 2025.

Where: In attribution reports such as Google Ads Goals > Attribution, GA4's Key event attribution paths, CM360's Path to Conversion and Amazon's Ad Console and API.

Why: Because most conversions follow several touchpoints, paths reveal channels that last-click reporting hides and feed attribution models and bidding; their limits are visibility, truncation and the gap between sequence and causation.