What We Learned Building for Filipino Micro-Businesses

Building for a Market Nobody Studied

When we started Elizabeth.ai, we could not find market research on Filipino Facebook micro-sellers. There were reports on Philippine e-commerce (Lazada, Shopee), on social commerce in general, and on SME digitization. But nobody had studied the specific workflow of a Filipino seller taking orders via Facebook comments in Taglish.

We had to learn everything firsthand. Here is what surprised us, what worked, and what we got wrong.

Lesson 1: Filipino Sellers Are Incredibly Resourceful

Filipino sellers have already invented their own systems. Through our research into how sellers currently operate, we see them using:

• Color-coded notebooks — Different colors for different order statuses
• Voice notes — Recording orders on their phone while scrolling through comments
• Family assembly lines — One person reads comments, another writes orders, another sends DMs
• Spreadsheet templates shared in seller Facebook groups, passed around like recipes

These were not unsophisticated people waiting for technology. They were entrepreneurs who had already solved their problems with the tools available. Our job was not to teach them how to manage orders — it was to eliminate the repetitive parts so they could focus on what they were already doing well.

Lesson 2: Price Sensitivity Is About Value, Not Poverty

When people hear "micro-business" and "Philippines," they sometimes assume the sellers cannot afford any software. That is wrong.

Filipino micro-sellers spend money on tools that clearly deliver value. They pay for GCash merchant accounts, Canva Pro subscriptions, ring lights, and packaging materials. They invest in their businesses.

What they will not pay for is software with unclear value. They need to see exactly how a tool saves them time or money, ideally within the first day. This is why our Free tier (100 orders/month) is so important — it lets sellers experience the value before committing any money.

Our Pro tier at PHP 2,499/month works because sellers can do the math themselves: "This saves me 3 hours per day. My time is worth more than PHP 83/day." The value proposition has to be obvious and immediate.

Lesson 3: Taglish Is Not Optional

Our first parser prototype was English-only. We quickly realized this would cover less than 20% of actual orders. Filipino sellers and buyers communicate in Taglish — and not the formal, textbook version. Real Taglish includes:

• Abbreviated particles ("po" becomes "p," "ko" becomes "q")
• Regional variations (Kapampangan sellers use different abbreviations than Bisaya sellers)
• Emoji as communication (not just decoration)
• Creative misspellings that are consistent within a seller's customer base

We built a Taglish-aware parser because we had to. There was no shortcut. And making it work well — understanding "pabili po ng 2 adbo" as an order for 2 adobo — required real message data, not synthetic test cases.

Lesson 4: Mobile-First Is Not Enough; Mobile-Only Is Closer

We designed our dashboard as "mobile-first" — responsive, works on phones. But observation showed us that many sellers never open a laptop at all. They manage their entire business from a smartphone. Some use tablets, but the phone is the primary device.

This meant our dashboard could not just "work" on mobile — it had to be the best experience on mobile. Large tap targets, minimal scrolling to reach key actions, no hover-dependent UI. The desktop experience is nice to have; the mobile experience is the product.

Lesson 5: Trust Is Built Through Consistency, Not Features

Filipino sellers are cautious about new tools. They have been burned by products that promise everything and deliver nothing. The way to earn trust is not a long feature list — it is consistent, reliable performance on the basics.

Our approach: nail order parsing accuracy first, then add features. If the core product (read comment → parse order → send confirmation) works reliably 99% of the time, sellers trust us. If we shipped 20 features but parsing failed 10% of the time, nobody would use us.

This lesson shaped our entire engineering culture. We invest heavily in reliability — the 3-provider AI cascade exists because we refuse to let a single API outage affect sellers who depend on us.

Lesson 6: "Simple" Is Not "Easy"

Making Elizabeth.ai simple to use was the hardest engineering challenge. Behind a "Connect your Facebook Page" button is OAuth flow handling, webhook registration, page token management, and permission verification. Behind "Your orders appear automatically" is a real-time pipeline involving webhook processing, hybrid parsing, intent classification, inventory checks, and Messenger API calls.

The principle: every hour of engineering complexity should remove a minute of seller complexity. We absorb the difficulty so they do not have to.

Lesson 7: Community Feedback Loops Are Gold

Filipino sellers talk to each other. They share tips in Facebook groups, recommend tools during live sessions, and warn each other about bad products. This word-of-mouth network is more powerful than any ad campaign.

We are setting up direct feedback channels — not just support tickets, but conversations with sellers about what works and what does not. When a seller in a Pampanga food sellers group mentions Elizabeth.ai, that recommendation carries more weight than anything we could write on our website.

What We Got Wrong

In the spirit of honesty, things we initially got wrong:

• Underestimating Taglish complexity — Our first NLP approach was too English-centric. We had to rebuild the parser from near-scratch.
• Overbuilding the dashboard early — We shipped features nobody asked for while the core parsing still had issues. We should have focused even more narrowly.
• Assuming sellers want chatbots — Sellers do not want to chat with an AI. They want the AI to handle orders quietly and efficiently. The persona feature adds warmth, not conversation.

Learn more about our team, or read the founder story: Why I Built Elizabeth.ai.

---

Want to see what we have built? Try Elizabeth.ai for free — 100 orders, full features, no credit card. We would love your feedback.