numéro 25 février 2023 #RetourSur… 4 ans de collection numérique la collection numérique ↓ 56 How The University of Colorado Boulder (UCB) is Delivering Data as a Product Back from Educause #EDU22 , a data-centric story Préambule : En tant que membre de la délégation française au dernier congrès Educause , nous avons gardé contact avec un de nos hôtes visités , Brad Weiner , Chief Data Officer in The University of Colorado Boulder (UCB) , et lui avons demandé de présenter une vision de cet établissement
US
sur une approche centrée sur la donnée .
En accord avec les auteurs , cet article vous est présenté dans sa langue d’origine .
Institutions of higher education have substantial data assets yet little capacity for converting those assets into actionable intelligence .
(Borgman & Brand , 2022) .
To solve this challenge ,
CU
Boulder (UCB) is working to deliver curated data sets , as a product , to end users with appropriate training and governance safeguards .
These data products span institutional domains including student success , admissions , retention , enrollment , teaching and lear- ning , among others .
This is a fundamental change from “analyses as a product” which limits analytic capacity by hiring a small core of data experts who act as information conduits to campus .
These individuals are difficult to hire , expensive to train , and cannot scale to meet exponential campus data demands .
In this paper , we will outline the philosophical and tech- nical underpinnings of our process , so other campuses may learn from , and build upon , our efforts .
↘
BUILD FOR AVERAGE USERS
,
NOT EXPERTS
.
Imagine a typical data user .
Do you imagine people writing Python code or creating pivots in Excel ? Although we often build for the former , at
UCB
, we are focusing efforts on the latter .
We are doing this by : ● Augmenting relational databases with a secure , user-friendly , object storage layer .
● Focusing on “boring rectangles” that can be opened in Excel .
● Reducing knowledge barriers by standardizing analytic choices .
Rather than including four variants of the same feature , we will include the most common version , document the ratio- nale , and provide choice paths if different options are required .
auteurs Benjamin Croft , Director of Analytics Engineering , Todd Schaefer ,
IT
Service Engineering , & Brad Weiner , Chief Data Officer , The University of Colorado Boulder données