March 2018 Summaries

Heap has introduced a groundbreaking behavioral attribution product that allows marketers to precisely track the influence of user actions on conversions across various channels without the need for engineering resources. Unlike traditional attribution models from companies like Adobe or Google, which focus solely on channels, Heap's solution captures every user action, such as email openings, blog post readings, or video views, and consolidates them into a single user identity. This codeless approach makes real attribution accessible to all marketers, enabling them to understand the entire customer journey and attribute marketing efforts to user actions seamlessly. Marketers can easily tag and measure the impact of specific actions on conversions, customize attribution models on the fly, and gain insights into which marketing investments yield the highest returns, as demonstrated by a scenario where different channels are analyzed to determine their contribution to revenue. Heap’s behavioral attribution thus provides a comprehensive view of user engagement, allowing marketers to optimize their strategies effectively.

Heap Engineering embarked on a rigorous process of intentionally breaking their data collection system to identify and address potential failure modes, inspired by practices like Jepsen and Chaos Monkey. This involved simulating availability zone (AZ) outages, isolating server components, and using artificial traffic to test system resilience without risking production data. Initial tests revealed issues with Redis dependency, leading to timeouts and Collector overloads, prompting the implementation of circuit-breakers and refactoring of Redis commands to improve responsiveness. Throughout this iterative testing, they uncovered and resolved various issues, such as slow Redis commands and inadequate logging during Collector restarts. The outcome was a more robust and reliable data collection system, reducing lost events and PagerDuty alerts significantly. The process not only strengthened the system's ability to handle AZ outages but also improved overall system performance, demonstrating the value of an agile and exploratory approach to system hardening.

Heap conducted an analysis of over five million anonymized website visits to determine which advertising channels were most effective for eCommerce businesses in terms of traffic and conversion rates. The study found that search engines like Google and Bing had the highest conversion rates, with Google at 8.2% and Bing at 7.6%, significantly higher than social media platforms like Facebook, which had a conversion rate one-third less than Google's. Interestingly, smaller shopping carts (under $200) boasted higher conversion rates compared to larger ones, with Google leading in conversions for smaller purchases and Bing for larger ones. Despite having lower conversion rates, social media platforms like Facebook and Snapchat drove the greatest amount of traffic to websites, accounting for approximately 13-14% of visits. The analysis highlights the substantial variation in conversion rates within each source and underscores the impact of shopping cart size on conversion effectiveness.