How It Works
A data experience journey: from data sources to data nodes to analysis apps to collaboration to useful data artifacts (UDATs).
Data Experience Overview
Data experience encompasses all interactions with the data. For the data scientist, this means mapping data into data nodes and building analysis apps. For the domain expert, this means using analysis apps to create and share useful data artifacts (UDATs).
1) Data sources
Leverage new and existing data sources, such as data warehouses, data lakes, and emerging data types. Pick and choose any analysis ready datasets and map them into individual data nodes.
2) Data nodes
The data node is the fundamental building block of the data mesh architecture.
Each data node ties a data source, algorithms and a smart API together into an independently deployable, containerized analysis server:
- Data map: Choose your analysis ready datasets (e.g. data warehouses, data lakes, and emerging data types) and map them into individual nodes.
- Algorithms: Use classic statistics or integrate scripts (e.g. R, Python, ML/AI).
- Smart API: Embed data-specific apps and connect nodes.
3) Data mesh
4) Data portal
The data nodes you selected are available on the centralized analysis site. Each node comes with its specific set of analysis apps. Every analysis performed is tracked as history, and the history is a basis for collaboration with your team.
5) Analysis apps
Analysis apps are embedded in the data node. They are built by the data scientists to guide the domain experts through their analyses.
The apps are designed to ask specific questions of the dataset in the node and provide a result format for the answer. They enable domain experts to run complex statistical or machine learning models through a point-and-click user interface.
6) Point-and-click to set parameters
Each analysis app asks a parameterizable question. These parameters enable domain experts to create cohorts, choose analysis variables, pick analysis algorithms, and sort the results. Each app allows a controlled range of parameter combinations to guide the domain experts.
7) Analysis results
Results are presented in a prioritized format, typically as lists ranked by probability. Visualizations such as clusters, graphs, and scatter plots are used to show the relationship between individual variables.
If the results indicate that your question needs to be refined, you can go back and re-parameterize your inputs. If they show something of interest, you can share the results with your team.
8) Share discoveries
Annotate your discovery and share with a single click.
9) Reproduce analyses
All actions, such as analyses, UDAT creation, and sharing, are automatically saved into your history, enabling them to be instantly reproducible.
10) Useful data artifacts (UDATs)
UDAT is a structured data object, created when you or an algorithm extract something useful from the data source. UDATs are frequently used as signals or starting points for further investigation, for example a gene signature or a defined patient cohort.