What it does
Africa's universal service funds hold money but lack data on where to spend it. USF Compass maps where people live and where networks reach, works out why each community is unconnected, proposes a remedy with a cost range, and ranks the projects so a fund can see what its budget would buy.
It separates two problems. A coverage gap means there is no usable 4G signal, and a tower or upgrade fixes it. A usage gap means the signal is there but people are not online, which needs cheaper devices, power or skills rather than masts. The tool recommends; the fund decides, with the weights it chose shown next to the ranking.
Zimbabwe is shown in full depth. Other countries use the public open-data layer only.
Two data layers
Public layer (all countries)
Built only from open data: population, phone-reported cell sites (OpenCelliD), open speed tests (Ookla) and terrain. Anyone can see it.National layer (Zimbabwe)
Adds the national spatial data (NSDI): operator sites, homesteads, schools and clinics. The sites stop at 2018, so the app presents them as a 2018 sample register, standing in for the regulator's current one. Only totals per area are published, never site locations.Missing data is unknown, never “no coverage”. An area with no phone reports may have no towers, or may just have no data. Such areas get their own hatched style on the map and never count as a gap.
Where people live
Population comes from WorldPop (2020, 1 km grid). It is added up into hexagons of about 5 km² (H3 resolution 7), which are the building block for every count in the app: people covered, people without 4G, and people a project would reach. Districts and wards are the sum of their hexagons. In Zimbabwe, NSDI homesteads, schools and clinics (with OpenStreetMap schools) add detail inside each area.
Who gets a usable signal
Each hexagon gets a status per technology (2G, 3G, 4G): covered, not covered or unknown. Two kinds of evidence feed it.
Terrain model
For every site the model traces 360 radials over the Copernicus GLO-30 terrain model and works out the signal along each: a standard propagation model (Okumura-Hata at 900 MHz, COST-231 Hata from 1500 MHz) plus knife-edge diffraction over hills. A place counts as covered when the signal at a phone held 1.5 m above ground beats the threshold: −100 dBm for 2G, −105 dBm for 3G and −115 dBm (RSRP) for 4G.Measurements
Open speed tests show service where people already use it. A hexagon counts as having usable 4G when its tests average at least 10 Mbit/s download. Phone-reported cell sites (OpenCelliD) confirm which sites are still on air.Why people are unconnected
Every district and ward is labelled by the group holding the most people:
- Coverage gap: people with no usable 4G.
- Usage gap: people covered by 4G but not using mobile internet.
- Mostly connected: people covered and online.
- Unknown: not enough data, or a tie between groups.
Who is online is estimated from the national mobile-internet rate, spread across hexagons by how many devices ran speed tests there for each person. Where there are no tests, the national rate stands. Each area panel lists the evidence behind its label.
What fixes it, and the cost
Upgrade to 4G
An existing 2G/3G site gets 4G. Upgrades are tagged by whether phone reports confirm the site still has no 4G, or need checking with the operator first.New site
Candidate sites on a 3 km grid, moved to the highest ground within 1.5 km, then chosen greedily for the most people. Their coverage is traced over terrain like any other site.Satellite Wi-Fi
For pockets too small for a mast, such as a school or a small village.Backhaul is microwave or satellite, because no clean open fibre-route data exists for Zimbabwe yet. Power is grid when the GridFinder grid lies within 2 km, otherwise solar. Every cost is a range(low, central, high), and the subsidy is what is left after the operator's own revenue over ten years: subsidy = lifetime cost − lifetime revenue, never below zero. Projects needing no subsidy are shown as commercially viable: a case for licence obligations rather than the fund.
Loading…
Which projects come first
Projects are picked one at a time. Each pick takes the project with the most benefit per subsidy dollar, then removes the people it covers from every remaining project, so no one is counted twice.
benefit = people × (1 + w_poverty × poverty + w_remote × remoteness) + w_facility × (schools + clinics) × 1000
Loading…
To show how sure the ranking is, it is rerun with costs and coverage drawn at random from their ranges. Each project shows its rank range and people range across those runs.
Does it work?
Loading…
Known limits
- The national sites stop at 2018 and have not been validated. They are a sample until the regulator's current register replaces them.
- Phone-reported cell sites are thinnest in rural areas, which is where most gaps are. That is why missing data shows as unknown.
- Costs are indicative placeholders with sourced ranges, and still need confirmation by the fund.
- There is no open fibre-route data, so fibre backhaul is not considered yet.
- School lists are incomplete: national data and OpenStreetMap together still miss some of the ministry's 11,371 schools.
- The delivery monitor (checking that funded projects connected people) is planned for 2027. The back-test stands in as proof of method.
Data sources
Loading…