How we score robots
Every Verdict Score on this site is computed from structured specifications by published code. This page shows the formula, the weights, and a real product's arithmetic worked through line by line, including where human judgement enters and where it does not.
The formula
Verdict Score=Σ ( subscorei × weighti ) ÷ 100
A subscore is a 0–100 rating of one capability, produced by a fixed point table applied to published specifications.
A weight is that capability's share of the final number. The weights of a category always sum to 100.
Nothing else enters. There is no manual adjustment, no placement fee, and no editor who can move a product up a list.
Change the weights and you change the question being asked, not the machine. That is what a ranking page is.
Weights, by category
Each category is judged by its own engine, because the machines answer different questions. A mower is rated on how it cuts and what gradient it holds; a vacuum is rated on suction, mopping and how much of the maintenance its dock takes over. The two scores are never compared: a mower's 84 and a vacuum's 84 mean different things, and no page on this site puts them in one list.
Robot lawn mowers
mower-score-2 · mower-weights-1- Navigation
- 20%
- Cutting
- 20%
- Terrain
- 15%
- Obstacle Avoidance
- 10%
- Ease of Use
- 10%
- Reliability
- 10%
- Features
- 5%
- Value
- 10%
Robot vacuums
vacuum-score-1 · vacuum-weights-1- Cleaning
- 22%
- Navigation
- 20%
- Mopping
- 14%
- Obstacle Avoidance
- 12%
- Automation
- 12%
- Ease of Use
- 5%
- Reliability
- 5%
- Value
- 10%
Robot pool cleaners
pool-score-1 · pool-weights-1- Coverage
- 22%
- Filtration
- 14%
- Navigation
- 18%
- Power
- 14%
- Automation
- 12%
- Ease of Use
- 5%
- Reliability
- 5%
- Value
- 10%
Window cleaning robots
window-score-1 · window-weights-1- Cleaning
- 22%
- Adhesion
- 12%
- Coverage
- 14%
- Safety
- 18%
- Navigation
- 12%
- Ease of Use
- 7%
- Reliability
- 5%
- Value
- 10%
Worked example
This is not an illustration. It is the machine currently at the top of the catalogue, with the arithmetic exactly as the engine computed and stored it. Every line is a specification, the points it earned, and the maximum it could have earned. If the scoring code changes, this page changes with it.
Robot lawn mowers
Segway Navimow X420
Verdict Score
89/100
Confidence
98%
Navigation
9720% of the Verdict Score
- Positioning system83/89
RTK-GNSS with vision
- Reference station6/6
No separate RTK antenna to site
- Zone management5/5
120 mapped zones
- Narrow passages0/5
Not published (not published, estimated from the category)
Cutting
9920% of the Verdict Score
- Adjustable rotary deck15/15
Baseline for cutting to a set height across the category
- Cutting width30/30
430 mm deck
- Height adjustment range11/12
20–95 mm
- Lowest cut8/8
Down to 20 mm
- Edge cutting20/20
Cuts to the lawn edge
- Blade system15/15
12-blade system
Terrain
10015% of the Verdict Score
- Rated slope65/65
84% gradient
- Drivetrain35/35
All-wheel drive
Obstacle Avoidance
8510% of the Verdict Score
- Detection system80/95
3D time-of-flight sensor
- Lift and tilt sensors5/5
Fitted
Ease of Use
10010% of the Verdict Score
- Installation30/30
No boundary wire to lay
- Setup hardware15/15
Self-contained positioning
- Companion app15/15
- Wi-Fi10/10
- Bluetooth5/5
- Multi-zone scheduling10/10
- Cellular connectivity5/5
- Rain sensor10/10
Reliability
8310% of the Verdict Score
- Brand track record42/55
Editorial index, see /how-we-score
- Weather sealing20/20
IP66
- Warranty15/15
36 months
- Time on market0/10
Unknown (not published, estimated from the category)
Features
905% of the Verdict Score
- GPS tracking15/15
- Cellular connectivity15/15
- Theft protection10/10
- Camera vision10/10
- LiDAR sensor0/10
- Multi-zone support10/10
- PIN lock5/5
- Rain sensor5/5
- Wi-Fi5/5
- Bluetooth5/5
- Edge cutting5/5
- Companion app5/5
Value
3110% of the Verdict Score
- Capability per 1,000 euros31/100
38.0 points per 1,000 euros, against 32.8–104.2 across the catalogue
| Subscore | 100 | Weight | Weighted |
|---|---|---|---|
| Navigation | 97 | × 20% | 19.4 |
| Cutting | 99 | × 20% | 19.8 |
| Terrain | 100 | × 15% | 15.0 |
| Obstacle Avoidance | 85 | × 10% | 8.5 |
| Ease of Use | 100 | × 10% | 10.0 |
| Reliability | 83 | × 10% | 8.3 |
| Features | 90 | × 5% | 4.5 |
| Value | 31 | × 10% | 3.1 |
| Verdict Score | 89 | ||
The same subscores, weighted for the question
A ranking never recomputes a product's subscores. It applies a different weight vector to the numbers already on the page above, which is why a machine can top Best for Slopes and sit mid-table overall with no contradiction. Every row below sums to 100. A zero means the question genuinely does not depend on that capability.
| Navigation | Cutting | Terrain | Obstacle | Ease | Reliability | Features | Value | |
|---|---|---|---|---|---|---|---|---|
| best-overall | 20 | 20 | 15 | 10 | 10 | 10 | 5 | 10 |
| best-for-slopes | 15 | 10 | 35 | 10 | 5 | 10 | 5 | 10 |
| best-for-large-lawns | 25 | 20 | 10 | 10 | 10 | 15 | 5 | 5 |
| best-for-small-lawns | 15 | 15 | 10 | 10 | 20 | 10 | 5 | 15 |
| best-without-boundary-wire | 30 | 15 | 10 | 15 | 15 | 10 | 5 | 0 |
| best-value | 15 | 15 | 10 | 5 | 10 | 10 | 5 | 30 |
| best-for-complex-gardens | 25 | 10 | 10 | 30 | 10 | 10 | 5 | 0 |
| Cleaning | Navigation | Mopping | Obstacle | Automation | Ease | Reliability | Value | |
|---|---|---|---|---|---|---|---|---|
| best-overall | 22 | 20 | 14 | 12 | 12 | 5 | 5 | 10 |
| best-for-pet-hair | 34 | 14 | 5 | 17 | 15 | 5 | 5 | 5 |
| best-mopping | 14 | 16 | 34 | 8 | 17 | 3 | 3 | 5 |
| best-for-carpet | 40 | 16 | 10 | 10 | 9 | 5 | 5 | 5 |
| best-obstacle-avoidance | 14 | 22 | 6 | 38 | 10 | 5 | 5 | 0 |
| best-self-emptying | 16 | 14 | 12 | 8 | 40 | 5 | 5 | 0 |
| best-for-large-homes | 18 | 32 | 10 | 12 | 18 | 3 | 5 | 2 |
| best-value | 20 | 16 | 8 | 9 | 8 | 4 | 5 | 30 |
| best-lidar | 16 | 40 | 10 | 16 | 8 | 5 | 5 | 0 |
The Verdict Score is arithmetic, not opinion
Each robot receives eight subscores from 0 to 100. Each subscore is produced by a fixed point table applied to the machine's published specifications: cutting width, rated slope, positioning system, ingress rating and so on. The Verdict Score is those eight subscores multiplied by the weights below.
The same code runs for every product. Nothing is nudged by hand, and no manufacturer can pay for a placement or a point.
Subscores are stable; weights are not
A ranking page such as Best for Slopes does not recompute a product's scores. It applies a different weight vector to the same eight subscores, so a machine's terrain capability is one number everywhere on the site and only its importance changes with the question being asked.
That is why a product can rank first for slopes and mid-table overall without any contradiction.
Where human judgement enters
One input is editorial: each manufacturer carries a reliability index from 0 to 100, reflecting service network, parts availability and track record. It contributes 55 of the 100 points in the Reliability subscore, which itself carries 10% of the Verdict Score, so at most 5.5 points of any final number.
Every value and every change to it is documented alongside the brand. A maker with no track record is given 60, the neutral default, rather than a guess in either direction.
Value is relative, on purpose
The Value subscore is capability per unit of money, positioned against the rest of the catalogue between the 10th and 90th percentile. Value compares prices in euros, and the price of a machine listed only in another currency is converted at fixed indicative rates. Adding a cheaper capable machine lowers everyone else's value score. That is the honest behaviour: value only means anything next to the alternatives a reader could actually buy.
Missing specifications are not zeros
Manufacturers publish different amounts. One prints a full data sheet; another prints six bullet points and a photograph. Scoring the second as though every unpublished figure were zero would rank press offices rather than robots, so we do not.
Where a specification is absent, the points resting on it are filled at the category's own typical rate, slightly below it, so that withholding a figure is never an advantage. The score reports how much of it stands on published data. A machine scored on a complete sheet reads 100% confidence and the estimate never touches it.
What we do not do
We do not claim to have tested a machine we have not tested. Every review states its method, and specification analysis is labelled as specification analysis.
We do not publish invented owner ratings, aggregate star scores we did not collect, or measurements we did not take. Where a specification is still unverified against the manufacturer's documentation, the product page says so.