How it works

A recommendation you cannot check is just an advert. So here is the whole method.

1. We only use numbers you can verify

For each of the 21 bikes in the database we record published specifications only: RRP, claimed weight, frame material, groupset, stack and reach for a size 56/M, official tyre clearance, stock tyre width, smallest chainring and largest sprocket, and whether it takes mudguards or a rack. Every one of these is on the manufacturer's own site. We record no subjective ratings, because you cannot audit an opinion.

2. Specifications become five scores

Those raw figures are converted into five 0–100 traits by a fixed formula that runs identically for every bike:

• Comfort — 45% riding position (stack divided by reach; a higher ratio means more upright), 35% stock tyre width, 20% frame material compliance.
• Speed — 50% weight, 30% how low and long the position is, 20% how fast the stock tyres are.
• Climbing — 55% weight, 45% lowest available gear in gear inches. A light bike with a 34×30 bottom gear is worse on a 20% ramp than a slightly heavier one with 34×36, and the score reflects that.
• Versatility — 65% official tyre clearance, 35% whether it takes mudguards and a rack.
• Value — spec quality per pound spent: frame material, electronic vs mechanical shifting, brake type and weight, divided by price.

3. And one sanity check

Those five traits alone will happily recommend a gravel bike to somebody who only rides tarmac, because fat tyres and an upright position score well in isolation — while being slow and simply the wrong tool. So there is a sixth dimension that scores whether the type of bike matches the surfaces you described. It is weighted heavily enough that a gravel bike will not win a tarmac-only search, and a race bike will not win a gravel one.

4. Your answers set the weightings

The five traits never change. What changes is how much each one counts. Say you ride mountainous terrain: climbing is weighted 1.35 instead of the 0.25 someone in the Fens gets. Say you ride 100km+: comfort is weighted 1.3 instead of 0.4. Tell us you are new to road riding and we add a further 1.25 multiplier to comfort, because beginners are consistently sold race geometry they cannot hold and end up riding on the hoods anyway.

Budget is a hard filter, not a weighting. We drop anything below 75% of your minimum as under-specified for what you told us, and anything more than 12% over your maximum. Bikes between your maximum and that 12% ceiling are shown, but flagged as a stretch — never quietly slipped in.

5. The match percentage

Each bike's weighted average is mapped onto a 50–97% scale. Nothing scores 100%, because no bike is perfect for anyone. The "why" bullets are simply the highest-contributing dimensions for that bike, and the trade-off line is the lowest-scoring dimension you told us you cared about. We show it because a recommendation that hides the downside is not advice.

6. How we make money

Affiliate commission. When you click through to a retailer and buy, we may receive a percentage of the sale. It costs you nothing extra.

The important part: commission rates play no part in the ranking. The engine described above never sees a retailer, a commission rate, or a stock level — it only sees the bike's published specifications and your answers. We choose which retailer to link to purely by price and availability. If the cheapest reputable seller of the winning bike pays us nothing, that is still the link you get.

What we are still missing

This is an early version and we would rather say so. Prices in the buy panel are not yet live — we are integrating retailer feeds. The database covers road, endurance, all-road and gravel bikes only. We do not yet account for frame sizing beyond a single reference size, wheel quality, or how a bike actually rides, which no specification sheet can tell you. Treat the shortlist as a way to get from a hundred options down to three worth test-riding.