Spectrum benchmark pipeline
A reproducible pipeline that turns messy regulator data into comparable international price benchmarks, and then into finished reports without anyone touching a spreadsheet.
Spectrum price benchmarking is a genre of analysis that is done badly almost everywhere, because the hard part is invisible.
Why the numbers are usually wrong
To compare what two countries paid for spectrum you have to reconcile at least six things: currency and the date you converted it on, licence duration, bandwidth, population, whether the price includes annual fees or only the upfront sum, and whether the coverage obligations attached were trivial or enormous.
Miss any one and you produce a number that looks authoritative and is not. The most common failure I see in published benchmarks is population: using a current figure against a price paid eight years ago, which quietly rewrites the answer.
The pipeline
The work was moved out of spreadsheets entirely.
Ingestion pulls from published regulator sources. A normalisation layer applies the conversions in a fixed order with the basis recorded for every value, so any figure in the output can be traced back to a source row and a set of assumptions. Benchmark construction then builds peer groups and computes the comparison set.
Output generation produces the finished chart set, the workbook and the document in one run. Nobody formats a table by hand, and rerunning after a data correction takes minutes rather than a day.
What changed
The important outcome was not speed, although it was much faster. It was that the assumptions became explicit and reviewable. When someone disagreed with a benchmark, the conversation moved from “your number looks wrong” to “you annualised over fifteen years and I would use twenty”, which is a conversation you can actually resolve.
Versioning the outputs meant a benchmark from six months ago could be reproduced exactly, which matters when a figure ends up in a regulatory submission.
Note on scope
This describes method and tooling only. The pipeline was built in a commercial context and the underlying datasets, peer groups and outputs are not published here. Nothing in this page reflects any specific client, country position or valuation.