We Were Pricing for Enterprise. Our Market Is Warung Owners.

The Disconnect We Almost Missed

We built Jiwa AI for Indonesian UMKM — the sixty-five million micro, small, and medium enterprises that power sixty-one percent of the nation's GDP. We wrote blog posts about how warung owners deserve world-class marketing. We designed a WhatsApp-first experience because that's where small business owners live.

Then we looked at our own demo form and realized the lowest budget option was "Di bawah Rp 50 juta." Fifty million rupiah. A warung owner pulling in five million a month would see that and close the tab.

Our trust badges said "DTC Brands" and "E-Commerce Leaders." Our pricing started at three hundred seventy-five thousand rupiah with no way to try the product at lower commitment. Every signal on our landing page screamed enterprise — while every blog post we wrote promised accessibility.

We were building for UMKM with enterprise packaging.

Five Changes That Fixed the Mismatch

1. Budget Ranges That Match Reality

We replaced our demo form budget options with ranges an UMKM owner would actually recognize. Instead of starting at fifty million rupiah, the options now start at "Di bawah Rp 1 juta" — under one million. For the English locale, we dropped from "Under S$5,000" to "Under S$500."

This is not just a cosmetic change. When a small business owner sees budget ranges that include their reality, they stay on the page. When the lowest option is ten times their monthly revenue, they leave — and they are right to leave, because the product appears to be not for them.

2. The Coba Pack: Try Before You Commit

We added a "Coba" (Try) pack at ninety-nine thousand rupiah — just under the psychological one hundred thousand barrier. Three posts, two stories. Enough to see the quality of AI-generated content for a specific business without committing to a full calendar.

At roughly two percent of an average micro-UMKM's monthly revenue, this is impulse-purchase territory. The goal is not revenue from this tier — it is conversion. A warung owner who sees their own nasi goreng featured in a professional Instagram post by a virtual influencer is far more likely to upgrade to the Starter pack than one who never tried the product at all.

3. Trust Badges That Say "You Belong Here"

We changed the landing page trust badges from "DTC Brands / E-Commerce Leaders / Growth-Stage Startups" to "UMKM / Brand E-Commerce / Startup Berkembang" in the Indonesian locale, and "Small Businesses / E-Commerce Brands / Growing Startups" in English.

The old badges were aspirational. The new ones are inclusive. When a toko owner sees "UMKM" as the first trust signal, they understand immediately that this product was built for businesses like theirs.

4. Price Anchoring in the CTA

The call-to-action footer now reads "Mulai dari Rp 99rb" — starting from ninety-nine thousand rupiah. Previously it said "Tidak perlu kartu kredit" which is a standard SaaS conversion tactic but means nothing to someone who does not use credit cards in the first place.

Leading with the price removes the biggest anxiety an UMKM owner has when evaluating a new tool: "Can I afford this?" When the answer is visible before they need to click anything, friction drops.

5. Differentiated Feature Tiers

Previously, all three content packs shared identical features — the only difference was quota. This created a strange value proposition: why pay four times more for the Premium pack if it includes the same features as Starter?

Now each tier has a distinct feature set. The Coba pack includes basic Instagram posting, community support, and WhatsApp delivery. As tiers increase, features unlock: revision limits increase, caption translations become available, multiple influencer styles open up, and dedicated support activates at Premium.

This creates a natural upgrade path where each tier genuinely offers more capability, not just more volume.

The Hidden Cost Optimization

While reworking the pricing, we also addressed a cost inefficiency in the AI pipeline. Our multimodal image analysis function was hardcoded to use our most expensive AI model — ten times the cost of the lightweight model — for tasks that did not require that level of reasoning.

Product color extraction, packaging identification, and basic visual analysis are routine pattern recognition tasks. The lightweight model handles them with equivalent accuracy at a fraction of the cost. We now default to the efficient model and only escalate to the premium model for complex brand analysis that genuinely benefits from deeper reasoning.

The result is an estimated thirty percent reduction in per-onboarding AI cost — savings that directly protect the margins that make the Coba pack economically viable.

The Lesson

It is easy to build for an audience on paper and price for a different audience in practice. The market we wanted — UMKM owners who need affordable, accessible marketing tools — was being filtered out by every signal our product sent before they could even try it.

The fix was not a pivot. The product was already built for them. The packaging just needed to catch up.