KASPR Datahaus · use case

Where no one runs a speed test

Crowdsourced measurement follows people: tests happen where users choose to run them, which is thinnest exactly where universal-service money goes. KASPR probes address space on a schedule, so a district with no engaged users is measured the same way, at the same hours, as a capital. Indonesia, district by district, is the worked example.

The sampling problem

A behavioural sample answers a different question

Consumer-initiated testing is invaluable for what it measures: the experience of users who test, where they test. But it is a behavioural sample — its density tracks engagement, handsets and habit, not need. For market benchmarking that is the right sample. For infrastructure policy it is the wrong one, because the districts policy most needs to see are the districts least likely to generate tests.

A designed sample inverts this. KASPR measures every district on the same schedule regardless of what anyone does, so coverage does not thin out where attention does. The trade is explicit: nothing here is a user’s experienced speed. What it is instead is presence and stability — whether infrastructure answers, and how steadily — for every boundary on the map.

The panel

434 Indonesian districts, measured identically

Each district carries a monthly connectivity index against its own baseline of 100, a latency and a jitter level, with history to 2019 and data through Jun 2026. The ten districts below fell furthest year on year — the shortlist a universal-service review would start from.

The spread

Most districts hold; a tail slides

The distribution of year-on-year change across all measured districts. The centre of the distribution sits near zero — most districts are steady — and policy attention belongs in the left tail, which a national average would erase entirely.

What this is for

Reading it as a regulator

Three uses follow directly. Targeting: the left tail is a candidate list for intervention, at the administrative level funding is actually disbursed. Verification: a subsidised district should move right in this distribution over the funding period — measured by a third party, not reported by the operator being funded. Response: the same districts carry hourly anomaly records, so a disaster’s connectivity impact is visible within hours at the boundaries relief planning uses.

The measurement approximates infrastructure remotely — IPv4 only, ICMP-reachable addresses only, third-party geolocation refreshed monthly — and every KASPR surface says so. Those caveats are the terms on which a consistent world-scale panel is possible at all.