How to Build a No-Code Data Stack

In the past few years, we have witnessed a tremendous evolution of the tech ecosystem, as no-code tools have rapidly reshuffled how we produce and manage software. Today, anyone with a laptop, an internet connection, and a brain can build a website or an app using no-code tools. The data ecosystem has largely kept up with this evolution, and a floppy of no-code data/ machine learning tools has recently emerged. The goal: lowering technical barriers at each layer of the modern data stack, to empower as many people as possible to leverage data science. No-code data tools generate a lot of excitement in the data community. The reason is, these tools are pioneering the movement towards data democratization, one of the five pillars of the next wave of innovation in the modern data stack. Today is about taking a deep dive into the no-code ecosystem, understanding the power/limits of these tools, and finding out about the available solutions.

‍

What's no-code?

A no-code platform uses a programming method that doesn't involve writing code. Instead, users work with a Graphical User Interface (GUI) where they can use templates, logical sequences, and drag-and-drop to achieve the desired output. The latter can be data ingestion, transformation, analysis, machine learning, etc. A no-code data platform basically enables people to perform all sorts of data manipulations without using code, which was traditionally impossible. These solutions offer a shortcut - leveraging the power of code but abstracting away the complexity so that users can focus on design and logic

No-code strategies focus on four key areas:

1. Providing drag and drop capabilities or other visual ways of configuring a process.
2. Creating filters and data queries to allow for instant customization of a process.
3. Using APIs to consolidate data from various sources.
4. Appealing to non-technical users, empowering them to conduct a given process.

A no-code platform differs from a low-code one. Low-code simply stands for a reduced amount of coding. You can build your workflows/ data science models without writing a single line of code. However, it is also possible to customize them by writing your own code, which gives increased flexibility.

Let's look at how no-code works in the realm of data science/machine learning:

Machine learning is a pain to learn: you need to prepare data, train your model, tune your parameters, etc. Machine learning models are traditionally built using scripting languages such as python, which adds complexity. A no code/ low code machine learning tool such as Google autoML allows you to create machine learning models fast. How? Mostly by automating some steps in the machine learning procedure. Some tools, like autoVIML can automatically put your data through different machine learning models in order to uncover which one fits your data the best. Automating parts of this procedure saves data teams tremendous amounts of time.

‍

What's at stake?

No-code data platforms play a key role in democratizing access to data. The widespread adoption of these tools won't happen overnight, but once it does, it's gonna impact non-technical users as much as data engineers/developers, while allowing tremendous productivity gains for everyone.

‍

For business users

A no-code data platform removes technical barriers, empowering non-technical users to manipulate data. This is key, as there has been a shift in the way businesses conduct analytics work. In fact, companies today engage in operational analytics - an approach consisting in making data accessible to "operational" teams, for operational use cases (sales, marketing, ..).

We distinguish it from the more classical approach of using data solely for producing dashboards and BI reports, an approach in which only data scientists/data analysts wrangle with data. Instead of using data to influence long-term strategy, operational analytics informs strategy for the day-to-day operations of the business. To put it simply, it's putting the company's data to work so everyone in your organization can make smarter decisions.

In practice, it means sales, marketing, and other business teams grapple with data and analytics every day, conducting various analyses that enlighten their short-term strategy. This trend will only accentuate, which means that more people, and especially non-technical profiles, are going to with data as part of their daily routine. Easy to use data/machine learning platforms allow users with no coding skills to optimize day-to-day operations and to solve business issues more efficiently. The learning curve is much smoother for these platforms, meaning a user can conduct value-generating analyses quickly, rather than taking months to master complicated software.

More importantly, no-code is about empowering non-technical people to analyze and route data between different cloud services on their own, without being forced to take a ticket and wait for data engineers and data scientists to do the work for them (which can take days, or weeks).

‍

For data engineers and data scientists

First, no-code doesn't mean no coders, and data engineers/scientists won't disappear anytime soon. Let's not forget a no-code tool is built with code, and these platforms exist because there are brilliant developers who are always working to find better solutions.

No-code data solutions empower every role of the new modern data stack, and especially the technical wizards. In fact, these platforms automate what can be automated in data manipulation processes.  Data engineers, data scientists and other highly-trained specialists can stop engaging in repetitive tasks that can be automated and can instead focus on solving larger, more interesting problems, like bias in system design.

A no-code data tool shouldn't only be used by non-coders. In fact, data people who are completely fluent in python can also gain by using a no-code approach to data analytics. Using a no-code approach allows you to process data much faster than you would with a more traditional, code-heavy approach. Coders also have a better grasp of the logic of how no-code  works than a person with no knowledge of programming at all.

‍

For businesses

The adoption of no-code tools means that organizations won't need to rely on a team of engineers and technical data people to leverage their data. Technical talent is scarce and expensive. Small and medium companies, representing 99% of the US market, usually can't afford to hire big data teams. No-code platforms represent a new way to deploy custom data processing solutions that would otherwise be unattainable to most organizations. Small companies can start leveraging machine learning and data analytics without having to make substantial investments.

‍

The downside of no-code tools

On the flip side of the coin, no-code tools also have a share of limitations that you should take into account before you decide to build a no-code data stack.

‍

Rigid templates

By design, no-code tools offer less refined options than what is available with coding. And that's completely normal, in the same way that you can express more things with words and writing than with hand gestures. Now, of course, hand gestures come across quite handy, but it doesn't mean you should only be using this as your only means of communication. No code tools provide various templates that can be configured to meet a set of use cases. However, as soon as you encounter that inevitable edge use-case, or need to introduce a layer of customization, no-code tools don't do the trick anymore. This limitation of no-code is especially penalizing in the realm of data processing since there is an endless number of different use cases at each layer of the data stack.

For example, you will quickly get stuck if you decide to solely rely on no-code platforms for data science and machine learning purposes. When designing a machine learning model, you want your model to be as fine-tuned as possible. With a visual interface and drag-and-drop functions, you're quickly going to encounter the boundaries of these platforms. The reason is, these platforms are based on models. Developers who designed a data science/machine learning tool usually decide to propose simple models that are easy to understand and use, with the aim of making machine learning and data science accessible to business users. The issue is, a simple model lacks flexibility because one can only develop within the framework of the model. Now, it's possible for platform designers to choose more elaborate models allowing more flexibility. But the learning curve to master these tools will obviously be much steeper than with simpler solutions. This kind of goes against the no-code platforms' purpose, which is to make data science and machine learning accessible to non-technical users. There is therefore a key tradeoff between ease of use and fine-tuning of machine learning models. That's why you should be aware of who you're purchasing the software for: data scientists and machine learning engineers, or business users? The ultimate end-user will impact the kind of software you go for.

‍

Platform dependency

The issue with no-code tools is that it's a matter of time until you run into the issue of vendor lock-in. Imagine you decide to use no-code tool to orchestrate your reporting workflows. The problem is, once these workflows are built around this tool, you're completely dependent on the tool. If the price rises, you have to keep paying or rebuild the entire system by yourself. In both cases, it's not an ideal position to be in.

‍

Which no-code tools for your organization?

Below, you will find a no-code tools landscape, which can hopefully help you choose a tool adapted to the needs of your company.

For a full no-code tools benchmark and more classification criteria, click here. For a list of no-code ETL tools, click here. 

‍

Subscribe to the Castor Blog

‍

More modern data stack benchmark?

Find more benchmarks and analysis on the modern data stack here. We write about all the processes involved when leveraging data assets: from the modern data stack to data teams composition, to data governance. Our blog covers the technical and the less technical aspects of creating tangible value from data. If you're a data leader and would like to discuss these topics in more depth, join the community we've created for that!

‍

At Castor, we are building a data documentation tool for the Notion, Figma, Slack generation. Or data-wise for the Fivetran, Looker, Snowflake, DBT aficionados. We designed our catalog to be easy to use, delightful and friendly.

‍

Want to check it out? Give the CastorDoc software a test drive with our free 14-day demo.