I needed a restaurant recommendation, so I did what every normal person would do: I scraped every single restaurant in Greater London and built a machine-learning model.
It started as a very reasonable problem. I was tired of doom-scrolling Google Maps, trying to disentangle genuinely good food from whatever the algorithm had decided to push at me that day. Somewhere along the way, the project stopped being about dinner and became about something slightly more unhinged: how digital platforms quietly redistribute economic survival across cities.
Because once you start looking at London’s restaurant scene through data, you stop seeing all those cute independents and hot new openings. You start seeing an algorithmic market - one where visibility compounds, demand snowballs, and who gets to survive is increasingly decided by code.
Google Maps Is Not a Directory. It’s a Market Maker.
The public story of Google Maps is that it passively reflects “what people like.” More stars, more reviews, better food. But that framing obscures how the platform actually operates. Google Maps is not just indexing demand - it is actively organising it through a ranking system built on a small number of core signals that Google itself has publicly acknowledged: relevance, distance, and prominence.
“Relevance” is inferred from text matching between your search query and business metadata. “Distance” is purely spatial. But “prominence” is where the political economy begins. Google defines prominence using signals such as review volume, review velocity, average rating, brand recognition, and broader web visibility. In other words, it is not just what people think of a place - it is how often people interact with it, talk about it, and already recognise it.
Visibility on these ranked lists determines foot traffic. Foot traffic determines how quickly reviews accumulate. Review accumulation then feeds directly back into the prominence signal. The system compounds. Early discovery generates demand. Demand generates data. Data generates future discovery. This creates a cumulative-advantage dynamic that looks remarkably similar to the way capital compounds in financial markets. This is essentially Robert Merton’s Matthew Effect applied to kebab shops - ‘unto every one that hath shall be given.’
This disproportionately rewards chains and already-central venues. Chains benefit from cross-location brand recognition. High-footfall areas generate reviews faster, meaning venues in those zones climb the prominence ranking more quickly even at identical underlying quality. By contrast, new independents face a classic cold-start problem: without reviews they are hard to find, and without being found they struggle to accumulate reviews at all. What looks like neutral consumer choice is therefore better understood as algorithmically mediated market design.
In economics, this dynamic closely resembles the logic of a market maker: an intermediary that does not merely reflect underlying supply and demand, but actively shapes liquidity, matching, and price discovery. Platforms like Google Maps perform an analogous function for local services by controlling visibility rather than prices directly. In the language of digital economics, ranking algorithms act as attention allocators, steering demand toward some firms and away from others.
Building a Counterfactual City with Machine Learning
... continue reading