diff --git a/docs/decisions/0003_superset.rst b/docs/decisions/0003_superset.rst
new file mode 100644
index 0000000..a7a269a
--- /dev/null
+++ b/docs/decisions/0003_superset.rst
@@ -0,0 +1,92 @@
+3. Superset as Data Visualizer
+##############################
+
+Status
+******
+
+**Draft** (-> Accepted)
+
+Context
+*******
+
+As we found in development of Insights, creating web-based visualization of data is a complicated
+problem requiring a fair amount of resources and technical knowledge of both data and UI design. The
+complexity and expense of adding new visualizations contributed to the stagnation of the Insights
+project, in spite of it having a tremendous quantity of data to draw from.
+
+Creating our new analytics system gives us an opportunity to bring in tools that are more flexible
+and able to support our needs without introducing a lot of new technology for small site operators
+to manage.
+
+Decision
+********
+
+The Open Analytics Reference System (OARS) will use `Apache Superset`_ as its primary data visualization
+and exploration tool. When developing a plan to support analytics in Open edX installs we have looked
+for a solution that offered at least the level of functionality we were able to get from the Insights
+user interface, as well as the following:
+
+- Open source
+- Deployable and configurable to work with OARS using Tutor
+- Able to use the LMS as an authentication and authorization provider
+- Capable of handling large quantities of data
+- Minimal new technology / expertise needed beyond the usual Open edX stack
+- Well supported
+- Comes with a large set of visualizations to choose from
+- Presents an API that allows authorized users to further access their data (for example via
+  download or Javascript extensions to the LMS)
+- Allows extension and customization for novel use cases
+- Allows operators to create and share queries / visualizations with each other
+- Allows site operators to choose from a wide variety of analytic database backends if necessary
+
+Superset simple meets all of our needs without ballooning our tech stack or introducing vendor lock-in
+or expensive paid solutions.
+
+.. _Apache Superset: https://superset.apache.org/
+
+
+Consequences
+************
+
+Superset will be integrated into OARS via a Tutor plugin that:
+
+- Allows it to share the existing Tutor redis and MySQL services
+- Integrates it with our chosen analytic database (detailed in a future ADR)
+- Creates our default suite of visualizations
+- Creates OAUTH integration with the LMS, with default permissions allowing different permissions for
+  site operators, course instructors, and analytics users
+- Any custom UI components that are needed to support our use cases (ex: a template showing all of an
+  instructor's courses)
+
+Rejected Alternatives
+*********************
+
+Rewrite the Insights Dashboard / New Custom UI
+----------------------------------------------
+As stated in Context, this is a complicated problem and would be an expensive distraction from our
+core mission of education. Superset provides all that we could hope to create and much more with
+a tool that is available today. Several Javascript libraries exist that simplify the visualization
+aspect, but still require a great deal of work to create and maintain a site framework to drive
+them.
+
+Grafana
+-------
+While `Grafana`_ can be used for a variety of data visualization purposes, it is primarily for supporting
+live services and lacks a lot of the visualizations that Superset has.
+
+.. _Grafana: https://grafana.com/
+
+
+Redash
+------
+`Redash`_ is similar in many ways to Superset, but seems poorly supported and possibly defunct. There
+have been few commits and no responses to Github Issues in the last 3 months. It would also seem
+to require operators to run Postgres in addition to the service itself, which isn't ideal.
+
+.. _Redash: https://redash.io/
+
+Several other options
+----------------------
+Most other packages in this space are either Javascript front-end components, make assumptions
+about data sharing that are incompatible with our desired permissions, lack features we need, or do
+not provide things we need in the open source versions of their products.