Low-code/no-code or custom coding?

*For readability, we will use the term ‘low-code’ throughout the article.

With a low-code solution, you need little to no in-depth technical knowledge. Thanks to the user-friendliness of the various tools, you can easily click together your desired solution. But is low-code the logical choice for every data issue? When answering this question, it is important to consider a few things.

Low-code solution or custom coding? 6 considerations for making the right choice

1. Is low-code really that simple?

With low-code, you make complex solutions seemingly simple. You get a nice visual overview, where you can click to zoom in on the details. This is also a danger: because the details are sometimes deeply hidden, you can overlook them. As data maturity increases, a simple setup can become a real spaghetti. This makes it unclear what steps are needed to produce the final tables. This can undermine confidence in the correctness of the data delivered to the end user.

2. Can you maintain focus?

Low-code solutions often have a specific focus. So don't expect a total solution when you start with low-code. Be critical about what you use them for and what not. If you have a complex issue, low-code solutions probably won't offer the flexibility you need. You might choose to do data extraction with low-code solutions and the data transformations using code.

3. Do your colleagues want to work with a low-code solution?

Be aware that the technical people needed to set up, for example, a data warehouse, may prefer not to work with low-code tools. If the tool turns out to be inadequate, they will have to untangle the entire 'spaghetti' and reprogram everything.

4. Are you becoming too dependent on one specialist?

The number of programming languages is limited. Every data programmer, for example, knows SQL, making the work easily transferable. In contrast, there are many different low-code solutions, reducing the chances of finding someone who knows your specific solution. If that person then leaves your organisation, it can lead to problems.

5. What are your performance requirements?

If you start small and stay small, you can easily click something together in a low-code solution. But if you keep adding things, you quickly run into performance issues. You can usually poorly optimize the way your data is transformed in a low-code tool.

6. Will a low-code solution meet your needs in the future?

Keep in mind that low-code solutions may not meet your needs over time if your requirements change, as the details of the setup are difficult to uncover. If you want to easily roll out the solution to multiple teams in the future, such as a data platform with fixed components, a code-based approach is advisable. With Infrastructure as Code (IaC), you can quickly set up the same environment with different parameters.

When to choose a low-code solution?

Low-code solutions are especially suitable for smaller, simpler projects. Think of performing simple data transformations on a single data source, such as reporting marketing results for Facebook and making this data available to your dashboard tool.

If you are sure that everything you want to do falls within the capabilities of a specific low-code solution, it is an efficient approach. However, if you expect to scale up in the future and think the requirements will increase, keep in mind that your tool may no longer be adequate. More complex transformations are often not possible, and combining different data sources becomes difficult.

What are reliable low-code solutions?

When looking for a low-code solution, it quickly becomes overwhelming due to the plethora of tools with large marketing budgets. We are happy to guide you with these reliable low-code solutions.

Extract phase

Tools for extracting data from systems Data ingestion tools like Fivetran and Stitch are very useful. You use built-in connectors already developed by other developers. The advantage is that you don't have to reinvent the wheel but use predefined connections. This is especially useful for extracting data from commonly used systems like SAP, Salesforce, HubSpot, and marketing platforms like Meta and Google.

Cloud providers sometimes also offer low-code solutions for data extraction. For example, Microsoft Azure has many connectors available within Data Factory that allow you to extract data to your data lake or data warehouse. This way, you can extract raw data from your SAP system or Salesforce.

Transform phase

Tools for transforming your data Many low-code solutions are also offered for transforming data. An example is Data Flows within Azure Data Factory, a solution where you define/format all transformations by clicking without using code. We usually do not recommend using this tool because it quickly becomes confusing and there is a lot of layering. You quickly have to click together multiple transformation blocks for relatively simple transformations. In contrast, you could also capture the transformation in 10 concise and readable lines of SQL code.

Data build tool (DBT) also fits the trend where, as a programmer, you need to write less boilerplate code (standard, repetitive code fragments). Common operations are abstracted by the framework, reducing the amount of code you have to write. However, it remains important to master SQL and Jinja to get the most out of the framework and write efficient code.

DBT is not a low-code solution but a framework that helps you capture your data transformations and ensure your data quality. It thus accelerates the work of a programmer or Analytics Engineer. Read here how this can work in practice.

Low-code solution or custom coding? We are happy to think along with you

In practice, we see increasing possibilities for using low-code tools. Your context is very important in this. If there is little technical knowledge within your organisation, low-code solutions can be a godsend.

However, it is not advisable to start small with low-code and then scale up with code as the complexity increases. This requires you to migrate all your data and essentially start over. So think in advance about where you want to go and whether this fits with your (data) strategy.

Don't try to solve every problem with low-code. Choose certain low-code solutions as part of your architecture to speed up work so that your developers have time left for complex issues.

Contact us if you want to discuss the right use and implementation of low-code solutions for your organisation.

Receive data insights, use cases and behind-the-scenes peeks once a month?

Sign up for our email list and stay 'up to data':