MSS valuation and GIS coverage modelling
Valuation tooling for mobile satellite spectrum, built on population and terrain data rather than national averages. Because a satellite band is worth what the ground under it is worth.
Terrestrial spectrum benchmarking usually runs on national aggregates: price per MHz per head of population, adjusted for licence term. It is crude but it works, because a national mobile network really does aim to cover most of a country.
Mobile satellite spectrum breaks that assumption completely, and the standard method quietly gives the wrong answer.
Why population per head fails here
A satellite band’s value is not distributed like a country’s population. Its advantage is precisely in the places terrestrial networks do not reach: at sea, over mountains, across the areas where putting up a mast is uneconomic or impossible.
Valuing it on national population credits it for urban users who will never use it, and gives it no credit for the maritime and remote coverage that is the entire commercial case. The directional error is large and consistent.
Doing it geospatially instead
The tooling works on a grid rather than a country.
Gridded population data is intersected with terrain and with modelled terrestrial coverage to classify every cell: well served, marginally served, or genuinely uncovered. Maritime zones are brought in separately, since the addressable demand there is vessels and routes rather than residents, and needs its own density model.
The result is an addressable population and an addressable area that mean something for this band specifically, instead of a national figure that describes a different product.
Feeding it into value
From there the valuation is conventional and deliberately so: addressable demand drives a subscriber and ARPU model, that drives cash flows, those discount at WACC, and the whole thing runs as a Monte Carlo because the uptake assumptions in remote and maritime segments are wide.
The geospatial work is upstream of the finance, not a substitute for it. What changes is the input quality, and that is where the error was.
The comparison that made the case
The clearest output was running both methods on the same band and showing the gap. Where a naive per-head benchmark and a coverage-weighted model disagree by a large multiple, the argument stops being about methodology preference and becomes about which number to put in front of a board.
Practical notes
Most of the effort was not modelling. It was projections, resolution mismatches between population rasters and coverage polygons, and coastline handling, where a small definitional choice about how far offshore to count moves the maritime figure substantially. That definition is an input to the model, stated explicitly, rather than an accident of the data.
Method only. The datasets, bands and commercial inputs are not published.