Google Maps knows where every jam in a city forms, has photographed most of its road signs and lamp posts, and has watched it grow from orbit for two decades. Nearly all of that has been used to route individuals to dinner. Google is now packaging it for the people who plan the city instead, and the significant part of the announcement is not the three new products. It is where they live.
The change is the delivery, not the data
Google Maps Platform has always been an API: your software asks a question about one address, one route, one place, and gets one answer back, billed per request. That shape is right for an app and wrong for analysis. You cannot ask an API what happened to average delay at four hundred junctions over five years, because you would be making four hundred thousand requests and reassembling the answer yourself.
The three new products land as datasets inside BigQuery, Google’s cloud data warehouse. That makes them joinable — against each other, against Earth Engine satellite data, and crucially against the organisation’s own tables. A planner can put ward boundaries, a budget line or a complaints log on one side of a query and twenty years of Google’s traffic history on the other. That is the capability that did not exist before, and it is a larger change than any individual dataset.
What the three actually contain
- Imagery Insights derives the condition and presence of physical infrastructure from Street View and aerial photography. Google’s own example is a utility auditing poles and signs across a region without sending a crew to each one.
- Roads Management Insights exposes historical and current traffic by road segment — which junctions congest, at what hours, and how that changed after a flyover opened. This is the one with the clearest public-sector use, because it can answer the question that road projects are almost never evaluated against: did it work?
- Places Insights aggregates businesses and points of interest in an area — density, categories, opening hours, how busy places get — for retail siting and service planning.
Alongside them, Earth Engine data is reachable from the same warehouse, so deforestation, surface water, fire risk and heat can be measured in the same queries as the traffic.
Note the word aggregates in Places Insights. It is doing load-bearing work. The underlying signals come from individual phones, and Google has to blur them into counts before publishing or it would be selling a surveillance product. That aggregation is also why the data answers questions about patterns well and questions about specific places badly.
What it would change for a city like Dhaka
Planning agencies in most of South Asia work from surveys that are expensive, occasional and out of date by the time they are published. A traffic count is a few people at a junction with clickers for three days in a dry month. Roads Management Insights replaces that with continuous measurement that was being collected anyway.
The practical consequences are specific. A transport authority could see from years of real data whether a U-turn scheme reduced delay or merely moved it a kilometre down the road. A power distribution company could audit poles across a whole thana from a laptop. Earth Engine already shows, season by season, how much of a city’s wetland has been filled — a question that is politically contested precisely because nobody has been able to settle it with a number.
The part worth arguing about
None of this is free, and the cost is not only the invoice. BigQuery bills per query and the Maps datasets are licensed, so a planning department’s evidence base becomes an operating expense that has to be renewed.
The deeper problem is that the evidence cannot be audited. A survey, however crude, has a method you can inspect and repeat. A vendor dataset has a method you are told about. When a planning decision is challenged in public — and road and land decisions always are — "Google’s data says so" is an argument the losing side cannot test, and the agency itself cannot reproduce after the contract lapses.
That is not hypothetical. Google published the COVID-19 Community Mobility Reports during the pandemic, governments worldwide built analysis on them, and Google discontinued them in 2022. The data was useful, free and gone, and nothing obliged it to continue. A dataset you do not own is a dataset that can be repriced, changed in definition, or retired between one planning cycle and the next.
The honest conclusion is still in favour of using it. The realistic alternative in most cities is not better, independent data; it is no data, and the same decisions taken anyway on the basis of whoever argued most confidently in the meeting. But an agency adopting this should treat it as it would any other single-supplier dependency: export and keep what it uses, write down the questions it is answering, and not let twenty years of someone else's traffic history become the only record it has.
