No Code: Almost Enough Is Still Not Enough

At Einblick, we’ve consistently heard that the availability and scale of data has grown faster than the supply of data scientists available to tackle that data. But is it enough to turn to “no code” tools, which allows non-technical users to participate in data science as well? 

To review some strengths first, no code environments share several appealing features – namely:

  1. No Technical Requirements: Lower the barrier to entry to enable more users to access insight. 
  2. Time Savings: Encapsulate repeated tasks into easy-to-use modules to simplify workflows.
  3. Action Oriented: Focus on getting to the right deliverable, using established workflows and pathways whenever possible. 

However, practical data science questions stray from the platonic ideal of no code in a few key ways.  

  1. Operator Limitations: No code tools are dependent on the vendor developers, as users are limited to just the operators already included. But analysis is dynamic, and users will almost always stray outside of a predefined path. 
  2. Frustrating to Technical Users: Projects are collaborative efforts, which might see participation from across skill levels. Users with more technical skills might be faster in a data science notebook, but have to change their workflow to be the no code version. 
  3. No Skillset Learnings: Citizen data scientists are being asked to learn complex tasks specific to a single software rather than general skills development. 

These challenges make it impossible for organizations to rely on no code tooling. The best platform for data science needs to go beyond “no” code and marry the features that make life easier with the adaptability and extensibility of coding. 

Einblick makes it possible to merge code with no code, allowing for both to exist as first class concepts in a single adaptable platform: 

  • Existing base of no code functions: You don’t always need a notebook. We agree with no code tools that mundane tasks in data exploration and model building are better pre-packaged.
  • Full notebook integration: When analysis needs more flexibility, write your own code and bring your own functions. Even custom code runs on top of Einblick’s progressive computation engine, meaning the super-fast computation you expect from Einblick is still true even for user defined operators. 
  • But Code and No Code are interchangeable: Einblick is built on top of open source packages, so you can easily export visual-based AutoML outputs into raw Python. Conversely, publish user written code into a visual operator, allowing it to be shared in a library and reused throughout the organization. 
  • Real-time collaboration: When no code user becomes blocked, a multiplayer environment creates a frictionless way for data scientists to drop into a workspace and assist. No time is wasted on missing context or mismatched environments. Through this interaction, less technical users should become cross-trained as well by seeing what is possible. 

When put together, the resulting tool is one that allows users to select the easiest way to move through analysis. Highly efficient no code capabilities can knock out most analyses quickly. Additional modules of code and expert data scientist support fills in any gaps. The resulting hybrid balances both “no code” and “full code,” to enable analysts to tackle any problem.