Building the Restaurant Data Layer South Africa Has Been Missing
The unglamorous technical project behind dish-first discovery, price intelligence and every dining insight that follows.
The quiet infrastructure problem
There is a category of technology that never gets celebrated but on which entire industries run: infrastructure. Payment rails. Number-plate systems. Address databases. The unglamorous layer that makes everything else possible.
South Africa's restaurant industry has, for years, operated without that layer. There is no national, structured, continuously-updated dataset of restaurant menus. There are directories, profiles, delivery-app listings, and social pages — each holding a fragment, none holding the whole picture, none able to talk to the others.
That gap is not a niche technical complaint. It is the reason the country cannot answer basic questions about its own food economy. And closing it is the central technical project behind Dine South Africa.
The premise
A restaurant industry without structured menu data is an industry operating blind — to its own prices, its own coverage, its own gaps. The dataset is the foundation for everything else.
What "structured" actually means here
Structured menu data is not a PDF of a menu made searchable. It is the menu decomposed into its component parts: individual menu items, each with a name, a description, a price, a category, a brand, a location, dietary tags, and a relationship to the restaurant and cuisine it belongs to.
This decomposition is what makes the dataset useful. A dish becomes a queryable entity — findable by name, by price range, by cuisine, by dietary tag, by geography, by brand. A location becomes a queryable entity. A cuisine becomes a queryable entity. Every dimension that a diner, a restaurant, or an industry analyst might want to filter on becomes a first-class piece of data.
1 entity
Per menu item
Name, price, category, tags, brand, location
Relational
Connected to menu, brand, location, cuisine
The whole becomes searchable
Why this is harder than it looks
The hard part is not the schema. The hard part is keeping it current.
Menus change. Prices change. Restaurants open and close. A structured dataset is only as valuable as it is accurate, and accuracy at national scale is a continuous engineering problem, not a one-off data entry exercise. The dataset has to be readable from live menus, matched against canonical dish definitions (so that "beef burger" and "classic burger" aggregate correctly), and recalculated whenever the underlying menus update.
This is why the Price Index recalculates on demand, not on a schedule. A stale dataset is worse than no dataset — it gives false confidence. The engineering imperative is freshness, and freshness is a continuous process.
Average burger price
Live Price Index data — renders automatically as platform data grows
What becomes possible once the layer exists
Once the structured layer exists, a surprising number of things that seemed like separate problems collapse into the same solution.
For diners, dish-first search — answering "where can I get this dish, near here, at this price?" — becomes a single query rather than a forty-minute research project. The diner doesn't experience the dataset; they experience the speed and confidence of the answer it produces.
For restaurants, the dataset makes pricing competitive rather than guesswork. The Price Index tells an owner where their dishes sit in the market — against the average, the range, the province — without a single manual competitor-pricing exercise.
For the industry, the dataset is the foundation for any serious analysis of affordability, cuisine coverage, and dining trends — the numbers that currently don't exist because the data underneath them doesn't either.
For tourism, the dataset is what makes a city's dining scene packageable and marketable, the way a wine route already is. Without the data layer, this is impossible.
The principle underneath all of it
There is a simple principle behind the technical work: a fact is more useful than an opinion. A menu's actual price is more useful than a reviewer's impression of value. A dish's actual location is more useful than a guessed neighbourhood. A cuisine's actual coverage is more useful than an industry's assumed maturity.
The entire dataset is built to surface facts. Every menu item is a fact. Every price is a fact. Every location is a fact. The opinions — the ratings, the reviews, the rankings — are welcome on top, as trailing confirmation. But they are no longer the foundation.
That inversion — facts first, sentiment second — is the technical project. And it is the reason the industry it serves will, finally, be able to see itself.
The engineering thesis
Build the unglamorous layer. Surface the facts. Let every other product, feature, and decision sit on top of a foundation that is structured, current, and national. Everything else gets easier when the data is honest.
— The Dine South Africa Insights Team
Related: The State of Restaurant Discovery in South Africa · How Much Does Eating Out Really Cost in South Africa?
Dine South Africa Insights
Technology
