Building an Nginx log analytics system from scratch involves collecting logs, parsing them, and storing the data in a database for analysis. The system uses Fluentd to collect logs from Nginx, Kinesis Data Firehose to process and store the data in S3, Athena to query the data, and Cube.js to build a dashboard for visualization. This setup allows for efficient querying of the data, even with large datasets, by partitioning the data and using pre-aggregations. The system also enables easy addition of other data sources, such as Snowplow events, into the same S3 bucket and merge results in Athena.