We selected a series of models and produced half a dozen to fundamentally unlock the capabilities and efficiencies of core county tools and processes. Step 3: We deliver high performance models to unlock county potential in machine learning. Having interviewed stakeholders and cleaned/combined the data, we presented a fleet of machine learning applications in a practical prioritization framework that could impact the entire county. "Any other vendor would have dropped off long ago, but you guys kept plowing." Six months into this process, the CIO marveled. Through interviewing stakeholders and navigating data gatekeepers, more than 450 tables and millions of data points were cleaned and stitched together. Step 2: We complete a data audit, build and present applications, and earn trust across multiple organizations. Over six months, we traversed approvals but also hundreds of data tables across multiple sources - column by column and field by field, building a rigorous map with their analysts of county data to understand the available inputs for machine learning and their readiness. Together with the central technology team, we wrestled through months of security protocols, data transports, and data quality qualifications. Together, we codified and captured dozens of needs from key organizations - from the Department of Public Works to the police department and onto the Department of Health. Storytellers partnered with the county CIO and a key program manager to educate more than 50 stakeholders. Step 1: We begin a journey of trust building with county executives, operational leaders, and data gatekeepers. UNLOCKING A PATH TO MULTI-MILLION DOLLAR COUNTY-WIDE MAINTENANCE EFFICIENCY If this journey inspires you, please reach out for a free, no-cost consultation. "I've never had a partner return this amount of value." - VP of Technology We then walked with the institution through how to implement A/B testing with ML outputs, supported their internal engineering teams with ETL work to smooth ingestion and action, then administered the tests - driving a more than 50% lift on key metrics. We did everything we said we'd do - clean and map data, interview stakeholders, build a prioritized list of opportunities, and build multiple prototypes. Step 2: Building prototypes, implementing them in production, and measuring lift via A/B tests. The lack of hand waving around the grit required to combine and clean data gave our client the confidence to step into a model prototype. We would map and prepare their data, table by table and column by column, build a representative data hygiene artifact and address findings, interview stakeholders and define helpful predictions, consolidate our data hygiene work with recommended science applications, then soberly talk through what is possible (and what isn't) given the data. We walked them through an unflashy, data hygiene obsessed path to high performing data science. We presented to the Head of Education, the VP of Admissions, and the VP of Technology. Step 1: Radical candor leads to unexpected opportunity. LIFTING KEY PERFORMANCE METRICS OVER 50% THROUGH DATA SCIENCE
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |