Since 2013
Continuous production
Live and paid for without interruption, through more than a decade of market change, data-source change and platform extension.
Case study · Market intelligence · Agri-commodities
Daily market intelligence for the global grain trade, in production since 2013.
AgroChart is a subscription platform for grains, oilseeds and feedstuffs: futures, cash prices, USDA supply-and-demand forecasts, tender results and market news, published fresh every morning. We are the engineering side of it, and have been for more than a decade.
Context
Why a day-late price is worse than no price.
Traders, processors, brokers and exporters in the global grain and oilseed markets commit real money against numbers that move every day. A cargo priced against yesterday's basis, or against a USDA forecast someone read wrong, is not a rounding error — it is the margin on the deal.
So the people in this market do not want a data warehouse. They want one place they can open before the day starts, trust without cross-checking, and go back to. That is a harder engineering problem than it sounds, because the moment the data is late, incomplete or quietly wrong, the product has failed at the only job it has.
Challenge
The market's data existed. It just did not exist in one place.
The inputs an analyst needs are spread across exchanges, government agencies and regional cash markets. Futures and currencies come from one kind of feed; USDA supply-and-demand reports come on a publication calendar; cash prices come from regional sources that each have their own format, their own quirks and their own idea of when they are ready.
Pulling that together once is a project. Pulling it together every single morning, for years, while sources change their formats without warning and new market domains keep getting added, is a different problem — and it is the one that actually had to be solved. On top of that, the platform had to serve a market that reads in more than one language, and had to be a commercial product with subscribers, not an internal dashboard.
Solution
A platform built around continuous ingestion, not one-off imports.
We built a market-intelligence platform whose centre of gravity is the data pipeline. Scheduled workers pull futures, currency, cash-price and USDA data on their own timetables, normalise it, and land it in a single model that spans roughly fifteen market domains — futures, cash prices, crop and production, exports, trade, freight, tender results, companies, news and market highlights among them.
On top of that sits the part customers actually touch: interactive charts and a market review for tracking current and historical prices, country-by-country supply and demand analysis built from USDA reports, full-text search across the archive, and PDF and Excel exports for people who take the numbers into their own models. For customers who would rather not leave their own systems at all, a market-data API exposes the same figures directly.
Because it is a commercial product, the unglamorous half matters as much: subscription billing and paywalled access, a multilingual interface, admin tooling so the editorial team can load and correct data without an engineer, and monitoring that surfaces a broken source before a subscriber notices a gap. It runs on Django and PostgreSQL with Celery-driven ingestion — supporting detail, not the point.
Result
Still running, still being extended, more than a decade later.
The platform has been in continuous production since 2013. It is a paid product with subscribers, and it reaches 10,000+ companies worldwide with its market reports — that reach figure is AgroChart's own published claim, and we cite it as theirs rather than as something we measured.
The result we can stand behind on our own account is the longevity. Data sources changed formats, market domains were added, the platform was extended repeatedly, and it kept publishing every morning throughout. Most systems of this age get replaced or quietly abandoned; this one is still the product.
Have data that arrives daily and never lands in one place?
That is the shape of problem this platform solves. Tell us where your data comes from and what it has to produce, and we will tell you what a reliable pipeline for it actually takes.
Outcomes
What the engagement produced.
Four things we can point at, with the source of each made explicit. Where a number is the product's own claim rather than our measurement, it says so.
10,000+
Companies reached — per AgroChart
The platform's market reports reach over 10,000 companies worldwide. This is AgroChart's own published reach claim about its product, not a figure we measured.
~15
Market domains in one model
Futures, cash prices, crop and production, exports, trade, freight, tenders, companies, news and more — normalised into a single product rather than a set of separate tools.
Daily
Automated ingestion
Scheduled pipelines collect and clean external market data on its own publication timetable, so the morning view is current without anyone assembling it by hand.
We do not publish uptime percentages, subscriber counts or traffic figures for this platform. We never measured them, so we do not claim them.
Our role
Developer on a decade-long engagement.
Our role here is engineering, and it is worth being precise about that. AgroChart is not our product: the market expertise, the editorial judgement, the customer relationships and the brand belong to the company that runs it. What we contributed is the system underneath — building it and then continuing to build it, year after year.
It is also the clearest example of how we prefer to work on long-lived products. Not a delivery that ends at handover, but a standing engineering engagement where the same people who wrote the ingestion pipeline are still the ones extending it — which is why a system from 2013 could keep absorbing new sources and new market domains instead of calcifying into something nobody dares touch.
Built with: Django · PostgreSQL · Celery for scheduled ingestion · Redis · full-text search · PDF and Excel export · subscription billing · multilingual interface · production monitoring.
Keep reading
Want a system that is still yours in ten years?
Tell us the data problem, the documents, or the workflow that costs your team the most time. We will come back with what is realistically buildable — and what we would be willing to publish about the result.