Price Radar: daily competitor repricing that knows when to ask a human

Every morning at 7:00, Price Radar checks your competitors, reprices your catalog within your margin rules, and asks you only about the changes it is not sure about.

The problem. Small online shops lose on both sides: too slow to react to competitors, or too fast and they undercut themselves below cost. The "fix" is usually a person with a spreadsheet at 8 a.m., or an AI tool nobody trusts with real prices.

What it does

- Catalog with cost and minimum margin per product; competitor offers per product

- AI reads a competitor page and extracts the price as strict JSON with a confidence score

- Deterministic repricing engine: undercut the cheapest valid competitor slightly, clamp to cost floor and ceiling

- Anomaly quarantine: competitor jumps over 30% or prices below your cost are never used

- Approval queue: changes over 10% wait for a human, small ones auto-apply, bulk approve on mobile

- Every decision has a plain-English reason ("Kestrel Supply dropped to €12.62. Undercut it by 1%. Stays €0.31 above the cost floor. Change is 15.7%, over your 10% limit, so it waits for you.")

- Dashboard: auto-applied today, waiting approval, anomalies, margin protected, price history

The story. I run this exact logic in production for a photo and video gear shop in Budapest: about 700 products, 4 competitors, every morning. The guardrails are not a feature list, they are scars. Base44 let me package it as something any shop owner can try in an afternoon. The demo shop "Northwind Tools" uses fictional data; press "Run repricing now" to watch a morning happen, "Reset demo" to start over.

Built on Base44. Entities for products, competitors, offers and price changes; a scheduled backend functi

Price Radar: daily competitor repricing that knows when to ask a human

Every morning at 7:00, Price Radar checks your competitors, reprices your catalog within your margin rules, and asks you only about the changes it is not sure about.

The problem. Small online shops lose on both sides: too slow to react to competitors, or too fast and they undercut themselves below cost. The "fix" is usually a person with a spreadsheet at 8 a.m., or an AI tool nobody trusts with real prices.

What it does:

- Catalog with cost and minimum margin per product; competitor offers per product

- AI reads a competitor page and extracts the price as strict JSON with a confidence score

- Deterministic repricing engine: undercut the cheapest valid competitor slightly, clamp to cost floor and ceiling

- Anomaly quarantine: competitor jumps over 30% or prices below your cost are never used

- Approval queue: changes over 10% wait for a human, small ones auto-apply, bulk approve on mobile

- Every decision has a plain-English reason ("Kestrel Supply dropped to €12.62. Undercut it by 1%. Stays €0.31 above the cost floor. Change is 15.7%, over your 10% limit, so it waits for you.")

- Dashboard: auto-applied today, waiting approval, anomalies, margin protected, price history

The story. I run this exact logic in production for a photo and video gear shop in Budapest: about 700 products, 4 competitors, every morning. The guardrails are not a feature list, they are scars. Base44 let me package it as something any shop owner can try in an afternoon. The demo shop "Northwind Tools" uses fictional data; press "Run repricing now" to watch a morning happen, "Reset demo" to start over.

Built on Base44. Entities for products, competitors, offers and price changes; a scheduled backend function for the morning run; one LLM integration call with a strict JSON schema; the repricing engine is plain code so the same input always gives the same price. Mobile-first layout, dark UI, Inter.

Try it: https://price-radar-abb5e005.base44.app

LinkedIn post: https://www.linkedin.com/posts/ihor-kolosov_base44challenge-ugcPost-7518628647407202306-XOqX/

Ihor Kolosov, AI systems engineer, Budapest. I build AI systems that run real businesses. Happy to answer anything about the engine or how it was built.on for the morning run; one LLM integration call with a strict JSON schema; the repricing engine is plain code so the same input always gives the same price. Mobile-first layout, dark UI, Inter.

Try it: https://price-radar-abb5e005.base44.app

LinkedIn post: https://www.linkedin.com/posts/ihor-kolosov_base44challenge-ugcPost-7518628647407202306-XOqX/

Ihor Kolosov, AI systems engineer, Budapest. I build AI systems that run real businesses. Happy to answer anything about the engine or how it was built.