Informal public transport is often the primary way of getting around emerging-market cities. However, infrastructure and investment decisions are held back by a lack of reliable data on the relationship between informal public transport, access, and economic vulnerability. Atlas AI overcame this by combining WhereIsMyTransport’s comprehensive Transit Data for Dar es Salaam, Tanzania, with their own spatio-temporal, high-resolution socioeconomic data, gaining a richer understanding of how informal public transport in the city supports economically vulnerable communities.

The challenge

In most African cities, formal public transport—such as government-run or funded bus and rail networks—has limited coverage and fails to meet overall mobility demand. As African cities grow and densify, planners are questioning whether these networks can serve the economically vulnerable communities who benefit most from public transport access to opportunities and services.

In the absence of formal public transport or private vehicles, low-income commuters have long relied on informal public transport—think tro tros in Accra, boda bodas in Kampala, danfos in Lagos—to meet their mobility needs. Yet there is little reliable data on the relationship between informal public transport and economic vulnerability in and around Africa’s cities, making it challenging to understand:

• Which communities are the most vulnerable?
• What opportunities and services do people typically attempt to access?
• What routes do informal public transport operators follow?
• What are the occupation and gender-related impacts?

Addressing these questions benefits from combining data assets. For example, pairing data on informal public transport coverage with data on the socioeconomic characteristics of the communities that rely on this type of transport.

Generating insights by combining data assets

Atlas AI’s data

Atlas AI develops high-resolution, spatio-temporal data for data-sparse environments. They provide insights on economic well being, agricultural productivity, and infrastructure using technology that integrates the latest advances in remote sensing and machine learning with empirical ground data from the field.

Atlas AI’s data sets on socioeconomic conditions provide information at national, sub-national, or local boundary levels. By overlaying Atlas AI’s data about household spending and asset wealth with population density, demographic indicators, and public or informal transport networks, it is possible to make even more precise associations between demand, access, supply, and utilisation.

WhereIsMyTransport’s data

WhereIsMyTransport’s breakthrough data production methodology includes collecting complete Transit Data on entire public transport networks—all formal and informal modes. Data from informal public transport is rarely available from other data providers, despite it being critical for a reliable understanding of emerging markets.

Atlas AI worked with our Transit Data for Dar es Salaam, which follows the standard, structured General Transit Feed Specification (GTFS) format. The priority attributes for Atlas AI were:

1. Calendar
2. Fare attributes
3. Fare rules
4. Feed info
5. Frequencies
6. Routes, Shapes
7. Stop times, Stops
8. Trips,
9. Vehicle attributes

“WhereIsMyTransport’s Transit data is the most granular data on informal transport available in emerging-market cities in Africa, addressing a long-standing and elusive need for planners and infrastructure investors all over the continent.” - Abe Tarapani, CEO, Atlas AI

More about our Transit Data

Combining data assets meant Atlas AI was able to establish whether informal public transport networks in Dar es Salaam provide the most economically vulnerable communities with access to valuable services. They investigated:

1. What patterns emerge from combining population and spending power data with the catchment areas of informal public transport routes and stops?
2. What behavioural differences are evident when adding the data attributes of electrification access, population count and density, spending, and asset wealth?
3. Can demographics be inferred by looking at the drive and walk time to informal public transport stops?

Pre-processing data

Atlas AI first deployed Extract, Transform and Load (ETL) processes to perform descriptive analysis in order to eventually answer their key questions. Atlas AI’s ETL tools automatically extract data from various sources and then clean, customise, reformat, and integrate that data into models. Their tools are specifically designed to work with geospatial data such as GTFS, geojson, and other formats, allowing the analyst to manage, analyse, and visualise data in a single platform environment.

The steps in their data pre-processing methodology:

1) Extract daladala stops from WhereIsMyTransport’s Transit Data for Dar-es-Salaam

Our Transit Data features all formal and informal public transport, meaning Atlas AI needed to extract daladala stops—a single informal mode—from the complete data set. They quickly identified about 3,500 unique daladala stops in the city.

2) Add attributes from Atlas AI data sets

Atlas AI added their attribute values to the extracted daladala stops, including:

• Electrification access
• Population count/density
• Spending (household consumption expenditure)
• Asset wealth (household wellbeing estimate)

The attribution step revealed, for example, that almost 1,500 stops are in neighbourhoods with a population density of 10,000 people per square kilometre. However many of these are quite closely spaced, as shown in the figure below.

3) Create catchment areas around daladala stops

By deploying an algorithm to compute a buffer area around all daladala stops, Atlas AI established a 4km catchment area around the complete informal network—its overall reach encompassing approximately 7.4 million people.

Note: Purple dots indicate daladala stop locations.

4) Understand how socioeconomic attributes vary near daladala stops between different areas of the city

After removing all socioeconomic data that fell outside of the daladala network catchment area, Atlas AI was able to observe how the underlying behavioural layers (electrification access, population count/density, spending, asset wealth) varied in and between areas of the city served by daladala routes. For instance, 411 daladala stops serve individuals living on a household spending budget of less than USD 3.2 per person per day—a commonly accepted threshold of poverty, in 2011 international purchasing power parity.

With their focus on neighbourhoods and locations at high resolution, the findings and insights can then be converted into parameters for models, forecasts, or other decision-making approaches for route planning, service optimisation, and tariff schemes, among other applications.

Note: Colours indicate spending power, with lighter colours indicating lower spending power and darker colours higher spending power.

WhereIsMyTransport data in the Atlas AI Aperture platform

Atlas AI’s data sets measure local market conditions such as demographic and economic information. This market data is turned into visual experiences and strategic insights through their web-based platform Aperture—a decision-support tool that enables B2C enterprises and development organisations to gain visibility into emerging markets and make smarter investment choices. In an urban context, as Atlas AI demonstrated with WhereIsMyTransport’s Transit Data for Dar es Salaam, detailed insights can be generated down to a street or even building level.

Note: Colours indicate spending power, with lighter colours indicating lower spending power and darker colours higher spending power.

Summary of findings

By combining Atlas AI data with WhereIsMyTransport’s Transit Data, Atlas AI found:

• The majority (98%) of daladala stops are located in areas that benefit from consistent electricity access. This allows the exploration of policies such as electrification of the transportation network. .
• Spending power (2011 USD per person per day) as a proxy for socioeconomic vulnerability shows that the informal network serves a varied swath of the urban and peri-urban population, with 2,374 (68%) out of 3,500 stops in neighbourhoods with spending power of less than USD 5.5 per person per day. Over 2 million residents live in the corresponding neighbourhoods with at least one daladala stop.

• The informal transport network structure makes apparent how peri-urban and rural settlements with lower socioeconomic status provide access to areas of higher economic vibrancy, not only affirming the patterns of transport but also helping quantify how communities can be expected to access jobs, services, and other activities.

Note: Colours indicate asset wealth levels, with darker red showing lower asset wealth and yellow showing higher asset wealth.

Atlas AI was further able to identify the population densities with access to the daladala stop—findings that are suitable for informing transport network planning and operations.

Those making investment decisions in the region could also benefit from geographical insights on spending power—the ability to pay. The addition of further data layers in the future could generate other insights, for example, an understanding of how male and female commuters use public transport differently.

“The bottom line is that historical models for transport in Africa have often missed the signal altogether. Models can only be realistic if  they combine both real-world transport behaviour—observed utilisation—with deep socioeconomic understanding—demand. This is where Atlas AI and WhereIsMyTransport come in.” - Vivek Sakhrani, Head of Data Science, Atlas AI

About Atlas AI

Atlas AI is a software engineering company that uses machine learning trained on satellite imagery and ground surveys to produce high-resolution data fabrics on economic well being, infrastructure, and agriculture at 1km x 1km resolution for the entire African continent and South Asia. Atlas AI delivers data and insights to their B2B and B2C customers through their proprietary Aperture(R) platform.