Three surfaces
People ask their network for a recommendation, search for options, then buy somewhere else entirely. Each step loses the context of the last, and the person whose advice started it gets nothing from the outcome.
Recommendations From People You Actually Trust
CURD puts search, social and a marketplace in one app, and scores what it shows you against your real network rather than against an anonymous crowd. A recommendation carries the weight of who it came from, and a referral that converts pays the person who made it.
Build Your Social Commerce PlatformAbout the Project
CURD is a consumer platform that merges three things usually kept apart: a social network, a search experience and a marketplace. Its stated premise is social validation — the platform analyses interactions within a user's own community and scores that data to produce personalised recommendations and search results drawn from people the user trusts, rather than from strangers or paid placement. On top of that sits a refer-to-earn model: users recommend brands to their network, the platform tracks the referral, and a commission follows a successful deal. Payments made through CURD return loyalty points, cashback and brand vouchers.
Industry
Social Commerce and Consumer Marketplaces
The overlap between social networking, discovery and transactions, where the recommendation and the purchase belong to the same product rather than to three separate apps.
The Product
Search, social and marketplace in one application
The platform positions itself as the place where search, social media and marketplaces become one surface, with AI selecting what is relevant based on the opinions of people the user trusts most.
Social Validation Engine
Interactions inside a community, scored into recommendations
CURD describes its AI as analysing interactions within the user's community and scoring that data to drive personalised, socially validated search results and recommendations.
Refer to Earn
Referrals tracked to the transaction, commissions paid out
Users refer brands they trust and earn commission on successful deals, with the platform tracking attribution and handling payout. Transactions through CURD also return loyalty points, cashback and brand vouchers.
Engineering Stack
React and Node services with PostgreSQL on AWS
The stack recorded for this project: a React client, Node services, PostgreSQL for relational and graph-shaped data, and AWS infrastructure.
Build your idea
Talk to our experts
Scope a social commerce, marketplace or recommendation platform with rewards and attribution built in.
Discovery Broke Into Three Apps, and None of Them Knows Who You Trust
A recommendation from a friend and a sponsored result look identical once they reach a feed.
Anonymous ranking
Conventional ranking optimises for engagement or for whoever paid. Neither answers the question a buyer is actually asking, which is whether someone they trust has done this before and what happened.
Unrewarded influence
Word of mouth drives real transactions and is almost never attributed. Closing that loop means tracking a recommendation through to a completed deal and paying the person who made it.
From a recommendation
to a transaction,
with credit paid back to whoever made it
Build Your Consumer Platform
How the Platform Works
-
Build the Trust Graph
Build the Trust Graph
- The social layer captures who a user is connected to and how they actually interact, not just who they follow
- That relationship data is the substrate for everything else: search, recommendations and referral attribution all read from it
- PostgreSQL carries relational and graph-shaped data together, which keeps connection queries close to the transactional record
-
Score and Rank by Social Validation
Score and Rank by Social Validation
- Interactions within a community are scored to decide what a user sees, so relevance is a function of trust rather than of engagement alone
- Search and recommendations draw from the same scoring, which is what makes results consistent across the app
- A ranking model tied to a user's own network has to work from the first session, before much interaction history exists
-
Close the Loop in the Marketplace
Close the Loop in the Marketplace
- Discovery and transaction sit in one product, so a recommendation can be followed through to a completed purchase
- Referral attribution links the person who recommended, the brand and the resulting deal
- Rewards — commission, loyalty points, cashback and vouchers — settle against that attribution record
-
Run It as a Consumer Mobile Product
Run It as a Consumer Mobile Product
- A React client over Node services, deployed on AWS, carries a feed, a search surface and a marketplace in one app
- Feed, search and checkout have different performance profiles but share one identity and one trust graph
- Consumer expectations set the bar: the social half has to feel as immediate as the apps it is competing with for attention
What This Build Demonstrates
Consumer product engineering where the social graph is the core data asset.
Graph Data as the Product Foundation
Connections and interactions are not a feature here, they are the input to search, ranking and payouts. Modelling them well in PostgreSQL keeps that data next to the transactions it has to explain.
AI Applied to Relevance, Not Decoration
Scoring community interactions to rank results is the product mechanic itself, which puts the quality of that model directly in front of every user on every screen.
Money Inside a Social Product
Commissions, cashback and vouchers mean consumer-app expectations on one side and payout correctness on the other, in a single codebase.
React and Node on AWS
A stack chosen for iteration speed on the consumer surface, with the relational core kept strict enough to support attribution and rewards.
More Case Studies
Global presence
