E-commerce funnel and commercial performance analysis (GA4)

01Business question

An online store typically knows its overall conversion rate, but not where visitors drop out between viewing a product and completing a purchase, or which devices and traffic sources bring revenue rather than just visits.

02Dataset & context

Dataset: Public Dataset. Google Analytics Sample E-commerce Dataset. This portfolio analysis uses Google's public GA4 Sample E-commerce dataset, available through BigQuery (bigquery-public-data.ga4_obfuscated_sample_ecommerce). It contains obfuscated event data from the Google Merchandise Store, covering 1 November 2020 to 31 January 2021. The analysis does not represent a PA Data Analytics client.

Dataset metrics used in the analysis include 270,154 unique users, 354,970 sessions, 5,692 purchases and $362,165 in recorded revenue, with an average order value of $63.63.

03Approach

Rebuild the purchase funnel from raw GA4 events using SQL in BigQuery, calculate conversion and drop-off at every step, then compare performance by device category and traffic source.

04Methodology

• Event-level funnel from first_visit through view_item, add_to_cart, begin_checkout, add_shipping_info and add_payment_info to purchase.
• Twelve BigQuery SQL queries covering KPIs, funnel steps, drop-offs, devices, traffic sources, countries and products.
• Validation in Python that revenue and purchase totals reconcile across tables.
• A Power BI dashboard for funnel monitoring.

05Modelling & analysis

The analysis reconstructed the e-commerce purchase funnel from GA4 event data, following users from product interaction through cart, checkout and purchase.

Dataset finding: the view-to-cart rate was approximately 20.5%, with 12,545 users reaching the add-to-cart stage out of 61,252 product viewers.

Dataset finding: a major leakage point occurred at the payment stage. Of 9,714 users who entered shipping details, 3,963 did not go on to add payment information, a drop-off of approximately 40.8%.

The analysis also compared purchases and revenue by device category and traffic source. In this dataset, mobile users purchased at a slightly higher rate per user than desktop users (2.16% vs 2.03%), while desktop generated more revenue ($208,815 vs $146,768). Google organic search was the largest traffic source by both users and revenue, and direct traffic was second by revenue.

06Results within the dataset

The analysis identified the payment stage as a major conversion leakage point, with approximately 40.8% of users dropping off at that stage. The funnel analysis also showed that only approximately 20.5% of product viewers progressed to the add-to-cart stage.

These findings highlight several areas for further investigation:

1. Payment and checkout friction. Investigate usability, errors, payment-method availability and other potential causes of payment-stage abandonment.
2. Product-to-cart conversion. Investigate product-page engagement, pricing, product information and calls to action to understand why a large proportion of product viewers do not add items to cart.
3. Channel quality. Evaluate acquisition channels using downstream funnel behaviour rather than relying solely on traffic volume.
4. Device experience. Compare funnel performance across device categories, including the gap between mobile's conversion rate and its lower revenue.

These are analytical opportunities identified in a public dataset, not changes that a business implemented.

07What this analysis demonstrates

How funnel analysis turns “our conversion rate is low” into specific places to act, and why event-level data in BigQuery gives a more complete view than standard GA4 reports.

08How a business could use this analysis

• Prioritise fixes at the funnel stage with the largest revenue-weighted drop-off.
• A/B test proposed changes before a full rollout.
• Monitor the funnel monthly from a single dashboard instead of separate platform reports.

09Related service

This project demonstrates methods used in our marketing analytics services.

10Apply this to your data

If you are facing a similar question, we can look at what your data can support and what a useful answer would look like.