Conversion rate is the percentage of interactions that end in an action an advertiser has decided to count: a purchase, a form submission, an app install, a phone call. Divide conversions by the interactions that preceded them and multiply by 100, and the result is the rate. Google's Ads help centre gives the standard example: 50 conversions from 1,000 interactions is a conversion rate of 5%. It exists because traffic is a cost and outcomes are the point: click-through rate shows attention, conversion rate shows whether attention became business.

The definition looks settled. It is not. Each platform picks its own numerator, denominator, time window and rules for conversions it cannot observe, so two systems can report different rates for the same campaign and both be correct.

One formula, several denominators

The numerator is whatever an advertiser has configured as a conversion action. The denominator is where the variants begin.

Per interaction. Google Ads divides conversions by "the number of total ad interactions that can be tracked to a conversion during the same time period," according to its help centre. Because an advertiser can track several conversion actions, or count every conversion rather than one per click, the rate can exceed 100%.

Per session. Google Analytics 4 (GA4) reports a session key event rate: sessions with at least one key event, divided by all sessions. PPC Land's explainer on key events uses the example of 600 sessions out of 20,000 containing a purchase or sign-up, a rate of 3% regardless of how many key events those 600 sessions contain.

Per user. GA4's user key event rate counts each person once, however many sessions they needed. Dynamic Yield's retail benchmark also divides purchases by unique users, according to Smart Insights.

Per impression. App stores sit at the far end of the funnel. Apple's App Store Connect defines conversion rate as total downloads and pre-orders divided by unique device impressions, where an impression is a view of more than one second on the Today, Games, Apps or Search tabs.

A worked case shows why this matters. A shop records 100 orders from 4,000 sessions by 2,500 users, with 3,000 paid clicks. The session rate is 2.5%, the user rate 4% and a per-click rate, if every order were credited to an ad, 3.3%. Same shop, same week, three answers.

How the platforms count

Google Ads separates two columns. Conversions includes only actions set as primary, and those feed bidding; All conversions adds secondary actions and other sources, according to Google's documentation on primary and secondary actions. Conv. rate is calculated from the first. Google Ads also dates a conversion to the click that produced it, not to the purchase, according to a legacy Google comparison of Analytics and Ads counting. A Monday click that leads to a Friday purchase raises Monday's rate, so recent days always look weaker until lagging conversions arrive.

GA4 keeps its terminology separate. Since March 2024, valuable events in Analytics have been called key events; "conversion" is reserved for the Google Ads object used for bidding. Key event rate metrics reached the user and traffic acquisition reports in June 2024.

Meta does not lead with a single "conversion rate" column. Ads Manager reports results and cost per result against the campaign objective, and since April 2019 has shown a conversion rate ranking, comparing an ad's expected conversion rate with ads chasing the same audience and optimisation goal, according to agency PMG.

On Amazon the inputs are clicks and attributed orders, and the window differs by account type: 7 days after a click for Sponsored Products bought by sellers, 14 days for vendors and for Sponsored Brands and Sponsored Display, according to analytics vendor Intentwise.

Windows, attribution and modelling

A conversion counts only if it falls inside a conversion window. Google Ads defaults to 30 days after a click, configurable up to 90 days depending on the source, and 1 day after an impression for view-through conversions, according to its help centre. GA4 replaced its six preset click windows with any value from 1 to 90 days on August 11, 2026, and opened the engaged-view window, fixed at three days, to 1 to 30 days. A longer window mechanically lifts the rate.

Meta removed 7-day and 28-day view-through windows from its Ads Insights API on January 12, 2026, leaving 1-day, 7-day and 28-day click, 1-day view and 1-day engaged view. In March 2026 it narrowed click-through attribution to link clicks, moving conversions after likes, shares and saves into a separate engage-through category.

The numerator also contains estimates. "In the 'Conversions' column, Google reports both modeled and observed conversions," according to Google's help page on modelled conversions. Modelling fills gaps left by declined consent, Safari and Firefox cookie limits and Apple's App Tracking Transparency, and can take up to five days to settle. Cross-device conversions have been part of Google's attribution picture since 2019, and of the Conversions column by default since a rollout that began on September 6, 2016, according to Google.

From converted clicks to key events

Google's conversion tracking dates to 2001, according to a 2016 post by AdWords product managers Tal Akabas and Frederik Kemner. AdWords long offered "1-per-click" and "many-per-click" counts, which became converted clicks and conversions around February 2014, according to Practical Ecommerce. Google retired converted clicks in September 2016, saying the metric could not handle non-last-click attribution, cross-device conversions or store visits.

Universal Analytics counted each goal once per session, according to Google's legacy documentation. GA4, which replaced it, moved to event-based counting and, in March 2024, to the key event label.

Typical values

Benchmarks depend on the denominator and the vertical. WordStream's 2026 search report, covering 13,474 US campaigns from April 2025 to March 2026, put the average Google Ads conversion rate at 8.18%, from 2.64% in finance and insurance to 16.22% in animals and pets. Its Meta edition, published on September 14, 2026, reported a median of 8.54% for lead campaigns, ranging from 4.15% to 15.87%.

Retail sits lower because the denominator is broader. Dynamic Yield's 2024 data put the overall retail rate at 2.9%, with 3.2% on desktop and 2.8% on mobile, according to Smart Insights. Contentsquare, analysing 99 billion sessions on more than 6,500 websites, reported that conversion rates fell 5.1% between the fourth quarters of 2024 and 2025, while average order values rose 6%.

Lead generation, e-commerce and app installs are not comparable. A lead form asks for an email address; a checkout asks for money. A 10% rate is unremarkable for the first and exceptional for the second.

Why the industry watches it

Conversion rate is the hinge between price and outcome. Cost per acquisition (CPA) equals cost per click (CPC) divided by conversion rate, so a 2-dollar CPC at 5% means a 40-dollar acquisition, and at 2.5% an 80-dollar one. Automated bidding prices auctions on predicted conversion probability, making the rate both a report and an input.

It has also become the main evidence in the argument over AI search. Microsoft Clarity found that referrals from AI assistants converted to sign-ups at 1.66% against 0.15% for search across more than 1,200 publisher sites. A study of 973 e-commerce sites by Maximilian Kaiser and Christian Schulze reached the opposite conclusion for retail, finding ChatGPT traffic converted below every channel except paid social.

Where the metric fails

Counting rules inflate it. View-through conversions credit an ad that was seen but not clicked, which shows that exposure preceded an action, not that it produced it. A rate that rises because a window was lengthened has not improved anything.

Consent breaks the numerator. In one case reported in April 2026, a misconfigured Consent Mode V2 setup cut a client's Google Ads conversions by 90% overnight; behavioural modelling later recovered about 40% of the attribution data, and the rest was lost. Since June 15, 2026, the ad_storage consent signal alone governs what Google Ads can collect from linked Analytics properties. Modelled conversions, meanwhile, are blended into the same column as observed ones, so the share of a reported rate that is estimated is not visible in the standard view.

Benchmarks rarely survive scrutiny. WordStream's Meta figures of a 1.80-dollar CPC and 8.54% conversion rate imply a cost per lead near 21 dollars, not the 27.39 dollars reported, because each median is calculated separately.

And the rate measures correlation, not cause. Only experiments such as conversion lift studies separate incremental conversions from those that would have happened anyway.

Not the same as

Click-through rate divides clicks by impressions. CTR stops at the click; conversion rate starts there.

Conversion value is the money attached to conversions. A campaign can double its conversion rate while halving its value per conversion.

CVR is the common abbreviation for conversion rate. Inside bidding models, predicted conversion rate (pCVR) is an estimate, not a reported result.

Key event rate is GA4's session or user-based measure, counted separately from Google Ads' interaction-based conversion rate.

Recent developments

The AI referral dispute has sharpened. Criteo said on May 5, 2026 that AI-referred traffic converted at close to twice the rate of traditional search in three retail categories, a figure it reported itself. Brainlabs claimed in October 2026 that AI referrals converted at roughly 1.5 times organic search, a ratio PPC Land could not reproduce from Brainlabs' own table, which pointed to about 1.3: key event rates of 13.4% against 10.4%.

New platforms are writing their own rules. In August 2026 OpenAI added view-through conversions with a fixed one-day window to ChatGPT Ads, in a separate column that leaves post-click conversion rate, bidding and billing untouched. Each new surface adds another definition, and another reason to ask what exactly was divided by what.

Timeline

  • 2001: Google introduces conversion tracking, according to a 2016 Google Ads blog post
  • September 30, 2009: AdWords begins counting view-through conversions on the content network, within 30 days of an ad view
  • October 1, 2013: AdWords adds Estimated Total Conversions, estimating cross-device conversions
  • February 2014 (approximately): AdWords replaces 1-per-click and many-per-click with converted clicks and conversions
  • July 19, 2016: Google announces that cross-device conversions will be included in the Conversions column automatically
  • September 6, 2016: Cross-device conversions rollout into the Conversions column begins
  • September 2016: Google retires the converted clicks metric
  • March 2019: Facebook announces conversion rate ranking, replacing the relevance score by the end of April 2019
  • May 1, 2019: Google Ads adds cross-device activity to all attribution reports
  • September 2020: Google introduces modelled conversions in Display & Video 360
  • July 1, 2023: Standard Universal Analytics properties stop processing data
  • March 2024: GA4 renames conversions to key events; Consent Mode V2 enforcement begins in the European Economic Area
  • June 13, 2024: GA4 adds user key event rate and session key event rate to acquisition reports
  • July 21, 2025: Google begins further enforcement of its EU consent requirements
  • October 24, 2025: Kaiser and Schulze's study of 973 e-commerce sites finds ChatGPT traffic converts below most channels
  • December 18, 2025: Microsoft Clarity publishes AI referral conversion data for more than 1,200 publisher sites
  • January 12, 2026: Meta removes 7-day and 28-day view-through windows from the Ads Insights API
  • February 25, 2026: Contentsquare reports a 5.1% year-on-year decline in conversion rates
  • March 2026: Meta limits click-through attribution to link clicks and introduces engage-through attribution
  • May 5, 2026: Criteo reports AI-referred conversion rates close to twice traditional search in three retail categories
  • May 19, 2026: WordStream reports an 8.18% average Google Ads search conversion rate
  • June 15, 2026: Consent Mode becomes the single control for advertising data from Google Analytics to Google Ads
  • August 11, 2026: GA4 makes click-through windows any value from 1 to 90 days and engaged-view windows 1 to 30 days
  • August 18-19, 2026: OpenAI documents view-through conversions for ChatGPT Ads with a fixed one-day window
  • September 14, 2026: WordStream reports an 8.54% median conversion rate for Meta lead campaigns
  • October 7, 2026: Brainlabs publishes key event rates for AI referrals and organic search

Summary

Who: Advertisers, agencies, e-commerce operators and app developers read conversion rate; Google Ads, Google Analytics, Meta, Amazon Ads, Apple's App Store Connect and newer platforms such as ChatGPT Ads each define and report their own version; benchmark publishers including WordStream, Dynamic Yield and Contentsquare, and analysts such as Microsoft Clarity, Criteo and Brainlabs, publish comparative figures.

What: The percentage of interactions that end in a defined conversion, calculated as conversions divided by a denominator that may be ad interactions, sessions, users or impressions. Platform choices on counting, conversion windows, attribution and modelling mean the same activity can produce several valid rates.

When: Google conversion tracking dates to 2001. AdWords split converted clicks from conversions around 2014 and retired converted clicks in September 2016, when cross-device conversions entered the Conversions column. GA4 renamed conversions to key events in March 2024, and 2026 brought changes to Meta and GA4 attribution windows, a consent overhaul on June 15 and view-through reporting in ChatGPT Ads.

Where: In ad platform interfaces and APIs, web analytics tools, retail media consoles, app store analytics and, since 2026, advertising inside AI assistants.

Why: Conversion rate connects what traffic costs with what it produces, converting cost per click into cost per acquisition and feeding automated bidding. Its weaknesses are inconsistent denominators, adjustable windows, view-through and modelled counting, consent gaps, unreconciled benchmarks, and the fact that it measures attributed outcomes rather than incremental ones.