Arvest Wine :

See how we designed a platform that aggregates and harmonises scattered data to build a single, reliable source of truth.

React
Python
Scraping
Data Analytics

Client: Arvest Wine
Sector: Data Analytics / Market Intelligence
Deliverables: Web application (data platform)
Delivery date: À compléter — delivery date
Expertise involved: web development, scraping, data engineering
The Arvest Wine platform: À compléter — Arvest Wine platform link

On the fine wine market, information is everywhere but unusable as it stands. It is split across a multitude of sources, in very different formats that are rarely comparable.

The result: the professionals in the sector (merchants, wine sellers, investors) buy and build their stock with no overall view, for lack of a unified reference to track prices and trends.

The stake for Arvest Wine? Turning that huge volume of data into a reliable, unified source of truth that professionals can genuinely rely on to decide.

The real challenge was not collecting the data, but making it trustworthy.

We had to automate the collection of information scattered across many sources in heterogeneous formats, then clean and harmonise it to make it genuinely usable and comparable. Heavy normalisation work, where the slightest inconsistency destroys trust in the whole reference.

To be useful, that machinery had to run continuously and stay reliable over time, despite sources that evolve, change format or disappear. The constraint was therefore to maintain the same quality over the long run, not just on the first few uses.

1.Framing and data architecture
We designed the whole data chain, from collection to transcription. The goal? Build a single source of truth where every price and every trend is reliable and comparable.
2.Automated collection at the source
We built scraping robots (targeted and at scale) that collect data directly at the source, across many platforms. Broad coverage, and collection designed to run continuously.
3.Normalisation and reporting
The raw data is then cleaned and normalised into a unified reference, and reported through clear dashboards on prices and trends. Precise, up-to-date information, ready to use.

Weekly sync meetings

Every week, a checkpoint let us validate progress, adjust priorities and prepare the next sprints.

Figma prototyping

Before coding the interface, the journeys and dashboards were prototyped interactively. Arvest Wine could try the data reporting for real, adjust the UX and validate decisions very early on.

Continuous test environments

As development progressed, we gave the client staging environments. As soon as a building block (collection, normalisation, dashboard) was ready, they could use it in real conditions and send us feedback so we could iterate fast.

The client stays in full control at every step and watches their product take shape in real time: no tunnel, no bad surprise on delivery.

To support a complete data chain (collection, normalisation, reporting) running continuously, we picked several technologies:

Data collection (scraping)Python (curl_cffi, BeautifulSoup, SeleniumBase, proxies)
Back-endFastAPI
Back officeVue 3 + Vite
StorageAWS S3 (Parquet)
InfrastructureDocker, Kubernetes (ArgoCD, Helm), Terraform

À compléter — number of sources aggregated / references tracked

À compléter — volume of normalised data updated continuously

A reliable market reference that turns raw, scattered data into genuine decision support — to optimise buying, pricing and stock management strategies. Data is no longer a headache: it is a competitive advantage.