WovokaStudio
← All work
ROUTE 04 · E-commerce operations · Amazon · live

An Amazon brand that reports on itself and asks before it acts

Amazon SP-APIGitHub ActionsPythonSupabaseClaudeStatic site + Mapbox
bujivanilla.com ↗

Sales and traffic pulled twice a day, analysed every Monday · scheduled jobs, since Sep 2026

Try itPick any of the 21 farms in the grower network and step into its 360° tour. Open the farm map on bujivanilla.com ↗

PROPOSED CHANGEAPPROVESALES + TRAFFIC · 2× DAILYMON · AI REPORT
2×daily automatic sales and traffic pulls
1weekly AI analysis, every Monday
21farms on the interactive map, each with a 360° tour
0store changes without a partner's yes

Figures from the project's own repository and configuration as of October 2026.

The problem

Buji Vanilla sells single-origin vanilla from smallholder farms on Amazon. Running an Amazon brand properly means daily reports, weekly ad tuning, review monitoring and restock timing, which is a full-time job. The partners have other full-time jobs.

What we built

  • Data that pulls itself. A scheduled job calls Amazon's Selling Partner API twice a day for sales and traffic, and weekly for the full report set, then commits the files to a versioned record.
  • Weekly AI analysis. Every Monday an agent reads inventory, ads, coupons, ratings and reviews, writes the weekly report, and drafts specific proposals such as bid changes, listing edits and restock warnings.
  • Private dashboard with passwordless sign-in. It shows live numbers, a "what customers say" summary drawn from review analysis, and an approvals queue where either partner can approve a proposal.
  • An expert council. Advice is grounded in fact-checked practitioner material on Amazon, food brands and vanilla, kept in the repository, so the AI's recommendations cite something real.
  • Brand site with tagged Amazon links that earn the brand referral bonus, and an interactive map of the 21 farms in its grower network.

Where the AI is

AI does the analyst's job and the people keep the decisions. Every change is a proposal with a reason, every approval is logged, and nothing touches the store without a yes. That's how a two-partner brand gets the attention of a full-time e-commerce manager.

How fast

The dashboard, automatic pulls and approval flow were built within days of moving the site off a no-code builder in mid-September 2026.

How this applies to your business

Any marketplace or e-commerce business where the data is available but nobody has time to read it:

  • Amazon, Shopify and multi-marketplace sellers
  • Franchise and multi-site retail reporting
  • Any weekly "pull the numbers, write it up, decide" routine

Have a version of this problem?

Start with Validate: two weeks, a working prototype, and a fixed price for the build.