Web

EGO's Insights From Building Our First No-Code App

In December 2021, a team consisting of a tech lead, a BA, a PM, a UI/UX designer, and a QA engineer built a web app in six weeks without a single line of code.

January 26, 2022
$
HK$

For our needs, we decided to build a simple web app that helps you order water online. That means it should allow users to sign up and log in, make an order, and send a payment.

ENKI: The About page of the desktop version of the web app
The About Us page of ENKI. Source: EGO

However, on the water supplier side, they should have not only a dashboard of orders or rights to edit and remove users' accounts but also have order data predicted. In real life, such data help reduce logistics costs for suppliers.

And our application developers wanted to see if using no-code tools brings advantages convincing enough to use them.

Typical Expectations From A Project Involving No-Code Tools

One of the biggest advantages of no-code tools is that they help people with no prior experience with hardcode software development to create apps. However, other advantages work for any web application company as well:

  1. Flexibility. Just imagine changing the app's business logic with a single drag-and-drop movement or upgrading a feature in a couple of clicks.
  2. Development speed. Testing is automated, slight functionality changes in the existing app are fast. And most startups today are in a rush to get their slice of the market before they run out of money.
  3. Reduced cost. Obviously, if you can make apps faster than before, it will cost you less because most teams use the T&M model, i.e., you pay for the amount of time the team spends on building your app.

For startup owners, no-code tools convey more transparent possibilities and limitations, allowing them to better communicate with the development team and plan the product’s further growth. And the development speed of no-code apps makes it possible to attract investments with fully functional MVPs rather than concepts, presentations, wireframes, or prototypes. 

And while the pros are all rainbows and unicorns, the cons are much more severe because no-code tools have limitations you must know beforehand to evaluate if these tools fit your project. 

And in the specific case of Bubble, the proprietary tool that we used, there is no access to its source code, so you can’t evaluate its security. In addition, you can’t export the source code of the app you expect to own.

So no-code tools appear to occupy a specific niche on the market of development tools we wanted to explore.

A Few Words on ENKI

It took the EGO web development services company three two-week sprints to develop a solution, a progressive web app, or PWA. 

A PWA is a responsive app that can be opened with a web browser on a desktop or mobile and be downloaded and added to the home screen of a smartphone for quick launch. 

ENKI: My Orders mobile interface
The mobile view of the User’s Orders page. Source: EGO

To deliver such a solution, we used the no-code tool Bubble.io, and a machine learning plugin called Peltarion.

The project involved our tech lead, a UX/UI designer and business analyst, and a manual QA engineer.

Working with Bubble.io

Bubble interface screeshot: The making of ENKI
Working with Bubble.io: the interface. Source: EGO

Our research showed Bubble is a perfect choice for web applications not requiring great dynamics and fast information exchange. Those usually appear to be news sites, landing pages, online stores, marketplaces, and online course platforms.

In addition to that, you can connect your Bubble-based app with any third-party service, provided it has a corresponding API. We're talking here about almost everything from social networks and payment services to media players and cloud storage solutions.

Bubble.io scheenshot showing the plugins we used to build ENKI
List of plugins we used in Bubble for ENKI. Source: EGO

Overall, Bubble is helpful in startup app development as a tool to dramatically cut the amount of time and resources necessary to build an MVP. It means that even if your app is too complicated for the Bubble possibilities, in most cases, you still can build an MVP with Bubble to start communicating and collecting data from users while the main app is developed via a more classical code-based approach.

Yet once the database the Bubble app operates with becomes too large – it seems like 100K rows of data feel like the threshold – its processing becomes intolerably slow. And there's no way to extract the app from Bubble to continue development with other technologies.

Hence Bubble is a poor choice for games and apps that allows users to perform quick actions.

ENKI: The My Orders screenshot with the Cart popup activated
The user accessing their cart in the ENKI web app. Source: EGO

While Bubble does not allow building native mobile apps, third-party services can help with that. Yet we were unhappy with the output: such mobile apps didn’t work and look well enough. Alternatively, a user can install a Bubble's app as a progressive web application that can be downloaded to the device and added to the home screen.

Working with Peltarion

Peltarion is a no-code plugin compatible with Bubble. Its ease of model creation and deployment makes it a great option for small-scale projects that need a few hundred prediction requests per month.

What we liked in Peltarion at EGO is its approach to documentation. There’s an elaborate knowledge center, free online courses, webinars available for learners, and should you have any questions, the support team is quick and helpful in their responses. 

However, all of that comes at the cost of data preprocessing limitations and restricted model analysis. There’s a lack of embedded functionality for model retraining and monitoring, too.

However, in the case of ENKI, our research showed it would be sufficient for our water consumption prediction feature.

ENKI: Dashboards and prediction models screenshot
Admin dashboard in ENKI. Source: EGO

Once we inserted the data into Peltarion, it ran preprocessing first. This is a process that includes:

However, all these stages are quite limited in their flexibility. As of December 2021, you’d prefer to preprocess your data with other machine learning platforms for more advanced data analysis and transformation before inserting it into Peltarion. We figured we’d need to use Colab and Python to transform the data according to our project's needs.

Once the dataset is preprocessed, you can use it for your modeling experiment and deployment in the end.

Peltarion provides a handy wizard to create default models that you can later modify. Or you can use different preloaded model templates. Once training is set, you can run experiments and see their outcome.

Model validation requires its deployment, i.e. connecting with your app. To properly monitor and finetune your model at this stage, you'll want to integrate third-party services and plugins, which was our case as well.

All in all, Peltarion satisfied our needs and was easy to master, though our prior knowledge of machine learning concepts was extremely helpful.

Bottom line: EGO's insights

Enki web app screenshot
Enki web app screenshot. Source: EGO


First of all, the ENKI project helped us expand our expertise. We're now ready to suggest Bubble for new clients if they're not sure about their ideas and need quick ways to explore them and get user feedback.

We ascertained that Bubble is great for quick prototyping, small-scale businesses, and content projects.

However, even in these cases, Bubble has important limitations we’ll be communicating every time we’ll suggest using it:

  1. The general approach to making smooth apps is basically about making sure there’s not much data to process, and the way it is processed is as simple as possible. 
  2. There will be limitations everywhere: in the ways the app can be designed, ported to the mobile platforms, supported by browsers, and, eventually, scaled.
  3. Once you have a working app made with Bubble, there’s no way to quickly migrate it if Bubble has no more tools to help you expand your app’s functionality. You’ll be left with the only option of rewriting it from scratch with the language/framework/tool of your choice.

We will follow Bubble's development and keep track of its changing limitations before we consider using it again. But overall, it's a worthy option.

Peltarion is a handy plugin to do basic machine learning stuff for small-scale projects if you don't want to write code. It has nice documentation and learning materials and is quite flexible when it comes to creating your own models, especially considering their paid plans. We’ll definitely consider using it again.  

If you want to know a bit more of the technical details behind the ENKI project itself, get in touch with our app development team, and we'll be glad to share our internal docs on it.

devops ninja animation
devops ninja animation
devops ninja animation
devops ninja animation
LIKE THIS ARTICLE? Help us SPREAD THE WORD.

More Articles

Back to blog