Mattermost’s Data Stack Explained: How They Leverage Unlimited Data For Customer Analytics

Overview

Who is Mattermost?

As you’d expect from an open-source tool, they offer hundreds of third-party integrations and connect to popular DevOps and developer workflow tools.

How Mattermost Use Real-time Customer Data

Real-time events, Alex says, allow teams at Mattermost to understand better how the customers navigate through the product and use it in general. These insights are then used to segment their audiences and build user cohorts. They also run A/B tests on various product-related features and measure their overall impact on the conversion and customer retention rates.

For their internal use-cases, the Mattermost teams make extensive use of data models, visual dashboards, and reports to track various aspects of their performance and overall business health. This includes financial forecasting, tracking their KPIs, and key product usage metrics.

Mattermost’s Data Stack: An Overview

  • Data Collection and Synchronization — RudderStack SDKs, Stitch Data, Heroku Connect, Custom Scripts
  • Warehouse — Snowflake
  • Data Transformation and Enrichment — RudderStack Transformations, DBT
  • BI and Data Querying Toolset — Looker
  • Job Orchestration — Apache Airflow

How Data Flows Through Mattermost’s Data Stack

  • Mattermost leverages RudderStack’s web, mobile, and server SDKs to collect user events in real-time and route them to their data storage infrastructure. For routing, they utilize RudderStack’s integration for Snowflake, their data warehouse.
  • Once all the data is dumped into the warehouse, they use Apache Airflow for job orchestration and scheduling. For data enrichment and transformation, they leverage DBT to convert all raw data across various sources into an aggregated data stream.
  • For business analytics and BI use-cases, Mattermost uses Looker to build visual dashboards and reports on top of the DBT data models.

Here’s a visualization of Mattermost’s data stack:

Mattermost’s Customer Data Stack (Click to open in a new tab)

Data Collection and Storage

Alex noted that with RudderStack, Mattermost overcame event volume limitations — a problem they faced with their previous vendor. Because of the vendor’s pricing model, they could capture only 2% of their user event data, which meant missing out on valuable customer insights. With RudderStack, Mattermost collects 100% of their event data, giving them rich insights into their customer journeys.

Like their high-trust customers, Mattermost is very data privacy and security-focused. Hence, all the events that are tracked by RudderStack are stripped off any PII (Personally Identifiable Information) before being sent to the data warehouse. They’ve simplified this cleansing process through RudderStack’s Transformation feature, which allows them to strip sensitive information (name, email, etc.) on the event stream in-transit.

Learn how RudderStack Transformations can be used to protect the PII in your event data.

Other Data Sources

For storing and processing their Salesforce data, Mattermost uses Heroku Connect — popular data integration and synchronization service. Mattermost sends all the sales data to their data warehouse through Heroku Connect and syncs the processed and enriched data back to Salesforce through custom scripts. This way, their sales teams always have the most up to date insights on customers and users.

Mattermost uses Stitch, a popular ETL solution, to collect the data from all these cloud sources, then dump it into their data warehouse.

Data Warehousing, Orchestration and Enrichment

For job scheduling and orchestrating, Mattermost uses Apache Airflow’s capabilities for dynamically instantiating pipelines. Alex also appreciates Airflow’s integrations, timely alerting, logging, and monitoring mechanisms.

Finally, Mattermost uses DBT (Data Build Tool) to transform and enrich the data in Snowflake. Alex and the team leverage DBT to define data sources and test the results of their transformations. They also use DBT’s modular SQL queries, which they can update and execute quickly and easily.

Activating the Data for Cutting-Edge Analytics

Also, since the DBT data models contain aggregated data, the Looker views are comprehensive, enriched, and accurate. This allows their team to make the little connections between the data, which usually go unnoticed, which is often where they find the most powerful insights.

Along with the ability to import or export visualizations outside Looker’s dashboard, teams within Mattermost can also use custom webhooks to build third-party visual workflows, eliminating the need to use any other external applications.

In Conclusion

Today, Alex has helped Mattermost set up a robust customer data stack that allows them to get an end-to-end view of their customer journeys. They can also translate this view into useful insights to build a product that improves the user experience and, in turn, boosts their business.

RudderStack is the CDP for developers. - https://rudderstack.com