How we review.

Research reviews.

Each launch review starts with the exact product identity and the available primary material. We look at product pages, published specifications and relevant manuals or support documents. The article links its sources so readers can inspect the configuration and claims directly. Photographs supplied by a manufacturer are labelled as manufacturer imagery.

We examine the intended job, the mechanism used to do it, navigation or control, installation, physical requirements, connected features, routine maintenance, consumables and support questions. We then describe the trade-offs those features create. Inferences are expressed as expectations, possibilities or questions, not as measured performance.

A research review cannot establish pickup efficiency, noise, battery degradation, obstacle-detection reliability, service quality or long-term durability without appropriate evidence. We leave those issues open when the source material does not resolve them. A long feature list does not earn an invented numerical score.

Owner accounts and community discussions.

An owner account can expose a useful question: a difficult transition, an awkward service step or a feature that behaved differently after an update. When a review uses such an account, it should identify and link the discussion, describe it as an individual report, and separate it from independently established facts. We do not turn a handful of posts into a failure rate or a claim about what most owners experience.

The launch library primarily uses product documentation. It does not claim a survey of Reddit or a statistically representative customer panel. We do not reproduce other publications' reviews. Community evidence can be added as identifiable, relevant accounts are evaluated.

Planned hands-on reporting.

For a physical evaluation, we will first record the model and region, software or firmware version, supplied accessories, unit source and relevant setup conditions. We will describe the task and operating settings so a reader can understand what the result does and does not represent. Repeated tasks should remain comparable, and interventions should be recorded rather than edited out of the result.

• Floor care: floor surfaces, debris, routes, preparation, pickup observations, wet-cleaning conditions and dock maintenance.
• Outdoor machines: mapped area, boundary arrangement, slope conditions, weather, routes, edge work and interruptions.
• Pool and window cleaners: compatible surfaces, cleaning modes, debris, missed areas, placement, retrieval and service.
• Companion and workshop products: setup, supported functions, control behaviour, task limits, learning requirements and ongoing costs.

Testing must follow the manufacturer's instructions and remain within a safe, suitable setting. A product loan does not justify bypassing operating precautions to obtain a dramatic result.

Corrections and revisions.

Material updates should state the date and explain the change. A firmware improvement, revised accessory or different regional model may require a new assessment rather than a silent specification swap. The original basis of a verdict should remain understandable.