What the price-app autopsy taught us
When the national agency launched a voluntary price-reporting app, the data arrived fast and wrong: the same RM20 petrol station reported as a budget grocery store, submissions from places that could not be geolocated to a district, prices conflated across subsidized and unsubsidized tiers. That is not a vandalism story — it is an architecture story. A price index built on anonymous voluntarism has no way to weight a shelf, no provenance for a number and no defence against signal decay once novelty fades.
The doctrine we shipped elsewhere transfers whole. From AnKampung: events only move along legal states, and a submission without a verifiable source is quarantined, not averaged in. From AnTIE: an aggregate is published only above the floor — item, place, week, band — so the index is dense where crowds are real and honest where they are not. From AnTeraju: grey until proven. AnHarga band we cannot substantiate stays visibly unsubstantiated, and the visual language says so.
Consent, not crowds
The feed has two honest sources. First, the opted-in aggregates from AnKedai ledgers — actual transacted prices at actual counters, in banded form, with the merchant revoking in one tap. Second, a enumerator-maintained sampling frame over the staples basket, anchored to BPN's published structure, so the index does not silently drift toward whatever people felt like reporting this week. e-Harga already proves the government wants machine-readable shelf prices; AnHarga is the private-built, ministry-signatory-owned way to finally get them without a surveillance drag along with them.