Paystub

When we took up the project, it had an existing design. But the challenge was it wasn't completely prepared for the rapid project development and support. The problem was in a different grid on each page, a problem with colors, and no unified design system.

We've outstaffed additional designers, and after assessing the design system, they proposed UI/UX changes to the user flows. All implemented features helped us save time for the pay stub generator development and support in the long run.

The calculation of taxes has its nuances. The federal and medical security taxes are calculated in the USA according to clear formulas, the same for all the states. When it comes to taxes for a specific US state, they're calculated according to different formulas depending on the state, and tax rates can change yearly. Furthermore, there is no open electronic source indicating the formulas for tax calculation for each state.

Finding the appropriate API was the single right solution. Although it took effort due to a specific industry, it helped us in two critical ways. Firstly, API simplified how we created tax calculation algorithms, considering the relevant info from sales tax rates, exemptions, and other factors influencing the final calculation. Secondly, API helped minimize errors and improve accuracy when delivering final pay stub reports for businesses and individuals.

The API for the tax calculation was expensive. We agreed with the client on optimizing the processing of each user step when entering payroll data. At the client's request, we had to perform the following optimization without major changes to the design of the pay stub generator.

We used various client-side optimizations, tracking each user step to assess the importance of each user request. Also, we used server caching to store the results of tax calculations for each state. Following the user journey in the tax calculation process, the analysis helped us set the right amount of requests. In turn, we reduced query performance to the tax calculation API by about 10 times.

The PDF generation on the project required many resources. Since we used Next.js development technology with deployment on Vercel, we had to consider separating the PDF generation. Vercel is a cloud-based platform allowing developers to build, deploy and easily manage serverless web apps and APIs. But the reason for separating PDF generation was because Vercel didn't allow the use of more than 50Mb for Serverless functions.

Separating PDF generation enhanced better organization and management of the files, optimized the process, and reduced the workload of the pay stub generator. We used firebase functions for PDF generation as a solution, allowing us to use many resources. It solved the resource-related problem and made it possible to scale the project.