Harda  ›  Journal

Building a hyperlocal product in Baku — 12 months of lessons

Anar Rafiyev · Founder, harda.az
Harda Editorial ·
After 15 years building SaaS products in Silicon Valley, I returned to Baku and built harda.az. Twelve months in, five lessons are clear: field work cannot replace product instinct, refusing paid placements is strategic advantage, local editorial + global engineering belong together, each language is a separate product launch, and open data is long-term customer acquisition.

Why I came back to Baku, and why restaurants

I spent 15 years in San Francisco building SaaS products. Customer analytics, product iteration cycles, engineering teams — the full modern product playbook of Silicon Valley. In 2025 I decided to return. The reason was simple: I grew up in Baku, but the city had no honest answer to "where should we go for family dinner tonight." TripAdvisor was for tourists, Google Maps was for search, but no source existed that earned a local's trust. I saw an opening to apply SaaS product thinking to this gap — which is why I chose not another SaaS, but a restaurant guide.

Lesson 1: Field work cannot replace product instinct

In Silicon Valley, the default model is "build → measure → learn": find a segment, watch the analytics, iterate. In Baku, this cycle requires a higher upfront investment because the digital infrastructure isn't at SaaS-market levels. The first six months I wrote zero code — I just walked restaurant to restaurant, read menus, talked to owners, paid bills, evaluated atmosphere. After hundreds of in-person visits, I understood the product context. The lesson: in markets with weak digital infrastructure, the "build-measure-learn" loop requires literal walking. Observation cannot be substituted by code.

Lesson 2: Refusing paid placements is strategic advantage

The local industry default is clear: monetization equals paid ranking. Restaurant portals, "best of" lists, even social media — everyone charges for visibility. I did the opposite. "No restaurant can pay for rankings" became the core principle of Harda. This reduced short-term revenue but strengthened long-term position. Bakulu users were tired of fake reviews and paid placements — being honest became a product differentiator in itself. Counter-intuitive conclusion: constraining monetization options up front yields a stronger brand position. Premium features will come later — but rankings will never be for sale.

Lesson 3: Local editorial + global engineering

Our team split looks unusual: editorial is fully Bakulu, while product and engineering carry Silicon Valley experience. This isn't accidental. The part the user sees (reviews, editor's picks, venue descriptions) requires trust that only a local can build. The part the user doesn't see (search speed, SEO architecture, analytics infrastructure) requires global product standards. This split gave us the "local perspective, international standard" position. Tactical lesson: in multilingual or multicultural markets, hire locally for what the user sees, globally for what they don't.

Lesson 4: Trilingual isn't a third language — it's a second product

Harda was built in Azerbaijani. The Russian and English translations were treated initially as "add-ons." Reality proved otherwise — each language opens a separate SEO surface, a separate audience, a separate trust signal. The Russian interface added Baku's Russian-speaking community and CIS tourists. The English translation made us a source for international tourists, academic researchers, and AI search engines. Each translation must be maintained separately, with its own quality criteria — Google Translate output doesn't produce a trust signal. The lesson: in multilingual markets, treat each language as a separate product launch.

Lesson 5: Open data is long-term customer acquisition

This month we published our "State of Baku Dining 2026" annual report under an open CC-BY license. The conventional thinking: "data is the company's main moat, don't share it." My position is the opposite. Baku has few trustworthy local research sources. By publishing open data, we turn every journalist, researcher, and AI engine into a citing source. Each citation is a backlink; each backlink is SEO authority; each authority bump makes the next citation easier. Counter-intuitive conclusion: to cement an early market position, sharing data is more effective than hoarding it. Moats will be built later — at this stage, sharing is positioning.

Why I'm writing this

The startup and product ecosystem in Azerbaijan is growing, but few people are documenting the playbook applied to the local context. Silicon Valley books and Y Combinator essays are useful, but they're written for the San Francisco market with 100,000 drivers — they don't always reflect Baku's dynamics. I want these five lessons to be a starting point for other founders in Azerbaijan. If you have questions, DM me on LinkedIn or write to info.harda.az@gmail.com. I'm open to sharing Harda's experience transparently.

Share this article
← Back to journal