Skip to main content
Honeyguide Monitor
EN · English NL · Nederlands

Honeyguide Monitor · 3.4.0

Methodology

Honeyguide Monitor is a continuously maintained measurement engine with two explicit methods: a public Benchmark and a private Company Scan. They share infrastructure, not scores.

Continuously maintained engine

The engine keeps running: provider adapters, content checks, cost telemetry, queues and correction paths. A public result is still a completed, immutable snapshot with a measurement window and publication date. "Up to date" means the latest completed compatible snapshot is displayed, not real-time model behavior.

Public Benchmark — HONEYGUIDE-BENCHMARK-0.4

Each selected company is scored with one search-enabled Understood call and two fresh search-enabled Recommended calls (1U+2R) under HONEYGUIDE-BENCHMARK-0.4, public schema 0.4.0 and depth HGB-PER-COMPANY-1U2R-0.4-PILOT. Recommended uses one category-fixed Dutch customer need; only the pre-assigned target scores. Coverage, freshness and the measurement window are published with the data.

Compare within a category first. The whole-population view is exploratory.

Company Scan — CIX-FUNCTIONS-2.1 + STANDARD-2.0

One company is measured with CIX-FUNCTIONS-2.1 and STANDARD-2.0: 16 distinct opportunities producing 20 responses across profiles, fit, factual checks, target-blind recommendations and comparisons. Score-bearing signals produce two axes, Understood and Recommended, locked before any explanation is written. U6 and R5 are diagnostic-only.

Public and private Understood/Recommended share brand names but are different methods. They are never presented as the same score, rank or delta.

Dutch measurement, two presentation languages

Questions are Dutch and the market is the Netherlands regardless of the website or report language. English and Dutch report views show the same locked result.

Environment and window

Each artifact discloses its tested model environment, evaluator environment, content edition and measurement window. A model, prompt or reducer change creates a new method or snapshot version.

Source roles

Official, independent and user-generated sources have different evidentiary roles. Single-source dependency and user-generated fragility are reported, never hidden.

Designed-sample limits

Scenarios are designed, not sampled from real buyers. We publish counts and variation, not confidence intervals or causal claims. Indices are not percentages of buyers.

Independent 250 selection

The Netherlands 250 is selected before scoring. Payment never changes membership, scenario selection, scores or public position.

Freshness, statuses and corrections

Every member has a status. Missing values are "Not available", never zero. Snapshots carry a freshness state and are immutable; corrections create a new version with lineage. Factual errors can be reported via the contact page.