Many (if not most) companies reach a point when data becomes a priority. This implies building out an internal practice to integrate into existing systems and processes to deliver the sought insights. In a field so wide, relatively recent and infamous for its buzzwords-per-second count, formalising problems and making explainable decisions is the only route that won’t see you run out of resources and people’s patience.
This talk explores how we approached this challenge at mx51 armed with lessons from engineering and statistics. We start by defining the optimisation problem formally(-ish) and then applying it to actual decisions faced along the way, including technology selection, warehousing and data lake (both Snowflake), ETL, visualisation and ML-driven insights.