October 2020 Summaries

The authors share their recipe for enriching customer data at Census using the Clearbit API, Snowflake External Functions, and dbt. They discuss the challenges of integrating a data enrichment process into an existing ELT pipeline, including maintaining freshness, ensuring consistency, limiting resource usage, and handling different types of enrichment pipelines (using datasets vs APIs). The authors then dive into their implementation using Snowflake's external functions, which allow them to call external handlers, such as Clearbit's API, from within SQL queries. They also discuss caching for performance and execution time savings. Finally, they outline the model definitions used in dbt to compute email values for users that aren't represented in user_clearbit_data, including views to filter freshest data and join users with enriched records. The authors note that this process provides a worthy replacement for external workers running enrichment jobs against the data warehouse and are exploring ways to make it more turnkey for customers.

This is a summary of how SaaS companies can leverage their product usage data to generate more revenue by turning free trials and freemium services into a valuable sales qualification strategy called "product-led" or "PQLs". PQLs are potential customers who have already had a positive interaction with the software, often through freemium versions or free trials, and have achieved specific qualifying actions. These leads convert at a higher rate than traditional marketing-qualified leads (MQLs) and sales qualified leads (SQLs), making them the most valuable type of sales lead for SaaS companies. To create a PQL scoring model, businesses need to identify their most important metrics and use logistic regression to score those metrics and identify qualifying actions. By leveraging PQLs in 4 steps - setting up data architecture, creating a score on unified accounts, routing accounts to the right place, and tailoring sales methods for each type of PQL - companies can generate more revenue by focusing their marketing efforts on generating high-quality leads that are likely to convert into paying customers.

The growth of self-service adoption has led to a need for structured account hierarchies to identify companies and track user behavior, as traditional sales often requires more effort to step in after initial adoption. To build an actionable hierarchy, one must identify real-world companies using the product, which can be done by analyzing domain names in users' email addresses. The data model should include a workspace concept, such as organizations or teams, where the owner is typically the user who created it. Aggregating users into companies involves grouping by domain and counting users or workspaces, with options to ignore personal email domains or roll them up to a single account. With an established hierarchy, businesses can use it to enrich existing accounts, identify net new logos, orchestrate sales and marketing efforts, and more.

Airbnb's Data University program was launched in 2016 to democratize data-driven decision-making across the company. The program was modeled after Google's internal training programs and aimed to give all employees access to information, regardless of their role. By prioritizing foundational classes over advanced ones and making the curriculum approachable for every employee, Airbnb was able to successfully scale its data team beyond just hiring more people. The program's success can be attributed to three key factors: clear expectations from leadership, measuring progress through metrics such as weekly active users and Net Promoter Score, and empowering data scientists to educate their colleagues. This approach has allowed Airbnb to inform every decision with data and increase transparency across the organization. By adopting a similar approach, companies can learn from Airbnb's history and take steps towards transparency and education in their own organizations.