RealUseBench Scoring rules
← Back to the test bench

Editorial standard · Version 1.0

A score only exists when the workflow leaves a receipt.

RealUseBench separates observed behavior from vendor documentation and unknowns. The same synthetic jobs, timers, fields, exports, and scoring rules apply to every product.

Canonical scenario

Four canonical datasets

  1. Cleaning: three sites, 18 checklist items, two shifts, and one corrective visit
  2. Garage door: two jobs, six doors, and three truck-stock SKUs
  3. Chimney: two flues, eight photos, one repair quote, and one annual return
  4. Air duct: two systems, 16 vents, 12 photo pairs, and one change order

Evidence captured

What earns a claim

Hands-on tested
The task was completed in the product and the resulting artifact was retained.
Vendor documented
An official source states the capability, but our test did not verify it.
Not verified
Evidence was incomplete, contradictory, inaccessible, or out of date.
Zero-weight revenue
Affiliate payout and commercial relationships never affect scoring.

Publication queue

Evidence first. Article second.

Each item remains queued until its synthetic job, source packet, and deterministic checks pass.

  1. 01

    Predeclare each rubric and product plan before testing

    QUEUED
  2. 02

    Publish elapsed time, taps, workarounds, and synthetic exports

    QUEUED
  3. 03

    Date every result and preserve pricing and version sources

    QUEUED
  4. 04

    Correct findings through a visible revision note

    QUEUED