Case Study: Building a Fintech MVP for a Bangalore Startup

The Challenge

A founder reached out to me with a simple problem: small businesses in India struggle to track GST compliance across multiple invoices and vendors. They were either using spreadsheets (error-prone) or expensive software (overkill for small businesses).

He had domain expertise—he'd been in accounting for 12 years. What he needed was someone to turn his knowledge into a working product.

Week 1: Getting Aligned

We spent the first week understanding the problem deeply:

- Who exactly would use this? (Small business owners with 5-20 vendors)

What's the minimum they need? (Upload invoices, extract GST data, generate reports)

What can wait for later? (Multi-user access, bank integrations, automated filing)

By Friday, we had a one-page scope document and a rough wireframe of the core screens.

Week 2: Design

We kept the design simple—this isn't a consumer app that needs to look trendy. It needs to be:

Easy to scan (lots of numbers)

Clear about what action to take

Fast to navigate

We used a simple dashboard layout with three main sections:

1. Upload invoices 2. View extracted data 3. Generate reports

The founder reviewed the wireframes with two of his accountant friends. They gave feedback, we adjusted, and moved on.

Week 3-4: Development

The core technical challenges:

OCR for invoice extraction: We used a combination of AWS Textract and custom parsing logic

GST validation: Cross-checking vendor GSTINs against government database

Report generation: PDF exports that match the format accountants are used to

We shipped features incrementally. By the end of week 3, the founder could upload an invoice and see extracted data. By week 4, basic reporting was working.

Week 5: Testing and Edge Cases

Invoices come in all shapes and formats. Some are PDFs, some are images. Some have GST broken down, others have it combined. We spent this week handling edge cases and improving extraction accuracy.

The founder tested with real invoices from his previous clients (anonymized). This revealed issues we hadn't anticipated—which is exactly why you test with real data.

Week 6: Launch

We deployed to production on a simple subdomain. No fancy launch—just invited 10 potential users to try it out. The founder personally onboarded each one and collected feedback.

Results

Within the first month:

8 of 10 initial users continued using the product

Average time savings: 4 hours per week per user

NPS score: 72

The founder has since raised a small seed round and we're building version 2 with more features.

What Worked

1. Founder with domain expertise

He knew the problem inside-out. When we had questions about edge cases, he had answers immediately.

2. Clear scope

We said no to many "nice to have" features. This let us focus on making the core experience great.

3. Real users from day one

Testing with real invoices revealed issues we never would have found otherwise.

4. Direct communication

WhatsApp for quick questions, video calls twice a week. No miscommunication.

Lessons Learned

- OCR isn't magic: Invoice formats vary wildly. Budget time for edge cases.

Trust but verify: User feedback is valuable, but watch what they do, not just what they say.

Start charging: Even early users. It reveals who actually values your product.

Working on Something Similar?

If you're building in fintech, accounting, or any B2B space and need help turning your idea into a product, reach out. I'd love to hear what you're building.