InsightArise

Reference sites built from public records

Governments publish enormous amounts of useful data. Much of it is hard for the people it is meant to serve to read and compare. We do the reading.

Start here: how to read public complaint data A complaint count reflects how many vehicles are on the road, and for how long, not only how well they are built. This short guide explains what our per-year figures and index correct for and what they cannot, works through one dated example from the Cars site, and shows how to read Taiwan procurement records. insightarise.com/reading-the-numbers/

A federal safety database is open in the sense that anyone may download it. In practice it is a search box that only works if you already know the exact model name the agency filed the record under, or a multi-megabyte archive of fixed-width text. The answer to a plain question — is this car worse than the others, and is it getting worse? — is in there, but nobody is going to sit down and extract it.

InsightArise builds the sites that answer those questions. Each one takes a public dataset, normalizes the records, applies documented comparison rules, publishes the limits of each comparison, and writes the result out as ordinary web pages, rebuilt on a schedule so the numbers stay current.

Sites

InsightArise Cars Owner complaints, safety recalls and federal defect investigations for 92 vehicle models across 21 brands from the 2018 model year onwards, normalized by time on the road, with trends. Built from U.S. NHTSA records. Nine languages. cars.insightarise.com 標案智庫 · InsightArise Tenders Taiwan government procurement, read from the official e-procurement notices: an index of tender, award and failed-award notices since 1999, with full award details mainly for recent years (older history is index-only), agency and supplier histories, evaluation-committee index, price ranges from awarded cases and a re-tender radar. Traditional Chinese. Updated twice a day. tenders.insightarise.com

How we work

Sources are named, not implied

Every figure on every page traces back to a named public dataset, with the date it was pulled printed on the page. Where the source has a permanent record — a recall number, an investigation ID — we link to it.

The method is documented

Cars publishes a full methodology page that states exactly how each number is computed, which comparison group it uses, and — just as important — what it does not account for. Tenders sets out its sources, update schedule and coverage limits on its About page. A comparison that hides its own limits is worse than no comparison.

Statistics come from code, not from a language model

No statistic on any of our sites is calculated or estimated by a language model. Counts, indices, rankings and trends come from fixed, tested code, and the sentences that describe them on data pages are built from templates we designed and maintain. The guides and long-form articles on Cars, and the explainer on this site, were drafted using an AI model (Anthropic's Claude); the About page says exactly where AI is and is not used.

Rebuilt on a schedule

Data is refreshed from the source on a fixed cycle — Cars is rebuilt monthly; Tenders takes in new notices twice a day and recomputes its statistics daily — so a page that says "updated" really was recomputed from fresh records. Each site publishes the date its data runs through.

Corrections

If a number looks wrong, tell us. We check the report against the source record and correct mistakes in our processing.

Independence

InsightArise is independent. It is not affiliated with, endorsed by, or speaking for any government agency, regulator or manufacturer, and it does not sell reports, leads or introductions. Our sites may carry third-party advertising through Google AdSense; each site's privacy page describes the cookies involved. Advertising has no effect on any figure or ranking: the figures are computed by code from public records before any page — or any ad slot — exists.

Contactcontact@insightarise.com
Sites live2
LanguagesCars: English, 繁體中文, 简体中文, 日本語, 한국어, Deutsch, Français, Español, Português · Tenders: 繁體中文