How AeroEdge works

Drop in a ride file. The AeroEdge advanced model does the rest.

Your number, in one line

Your number is how well you cut through the air. Smaller is faster.

Engineers call it CdA. A rider sat up on a road bike is around 0.30. A good time-trial position gets near 0.20 or below. At racing speed the air is usually the biggest thing slowing you down, so this is the number most worth knowing.

All we need is a ride file

Any file off your bike computer works: a plain FIT file from a Garmin or Wahoo, or a Notio, GiBLI or AeroSensor file if you ride with a wind sensor. No special kit is needed to start.

The advanced model takes it from there. From that one file it:

• Finds the parts of the ride it can trust: steady stretches, loops, and out-and-backs. You do not have to mark anything.
• Works out the wind. Measured every second if you have a sensor; otherwise worked out from the weather for the exact time and place you rode.
• Holds your tyres and chain to real data, corrected for the day's temperature, so their cost does not leak into your aero number.
• Checks itself. The same ride is worked out more than one way, and against your GPS altitude. When the answers agree, the number is good.
• Reads your corners from the GPS: how hard you cornered, and how much speed you left on the table.

Each run also keeps the weather from the day you rode it, so the number will not drift if you open the ride again next year.

A test day, start to finish

• Make a session for the day, and add a bike if you track your kit.
• Drop each ride into it. The model reads the file straight away.
• Tell us your tyre and your chain so the model can price them.
• Riding indoors? Tick the indoor box and give us the temperature.
• Want a specific stretch? Drag across the ride and it gets its own number. Otherwise the model's auto-detected runs are already there.

The advanced workbench

The analysis page is the full workbench, built for coaches, athletes, fitters and engineers from Level I. Drop in a ride from any supported device and it runs all three models on it. It shows you what it found and how sure it is; it does not tell you what the result means for your position. That reading is yours to make.

Road surface, and why it starts at no correction

Tyre rolling resistance is measured on a smooth lab drum, and real roads are rougher, so the workbench lets you scale it. Only one level is measured: typical asphalt at ×1.50, from comparing drum and on-road rolling resistance for the same tyres. Velodrome ×1.00 is the drum baseline itself. The three levels between them are our estimates — nobody has published figures for them.

There is one more catch, and it is why the default is no correction at all: those factors were calibrated against a different lab's drum from the tyre table now in use, so they read high today. A surface-corrected Crr comes out too big, and the CdA worked out from it comes out low by roughly 0.01 m². Use it to compare two of your own rides on the same road, not to quote an absolute number.

Where to test

The model needs a ride that comes back to where it started: a loop, or out and back on the same road. It does not have to be flat; roughly 100 m of climbing over 6 km or less is fine, with each direction holding for at least 90 seconds without braking.

Wind hurts through its gusting, not its strength: a gusty 2 m/s wrecks what a steady 4 m/s would sail through. Set a test venue in your profile and AeroEdge watches the forecast and flags the next calm window for you.

Testing two setups? Ride A, then B, then A again, in the same session, so a real difference cannot be confused with the day drifting underneath you.

Reading your result

The big number at the top is your CdA for this ride. Under it, a line says where the wind came from: measured by your sensor, looked up from the weather, worked out from your ride, or still air. That line matters, because wind is what makes or breaks the number.

Everything else is the model showing its working: the ride drawn from your GPS, the virtual-elevation trace sitting on your real altitude, the power you would need to hold each speed, and your corners. Want all of it? It is behind Open raw data.

On a paid plan you also get the give or take: how far the number could be out, and which input is making it wobble. It is measured from your own ride, not guessed, and a wide one is the ride telling you to test again before you act on it.

If you are on the free plan

You get your number straight away, and it is the same one paying members get; we do not round it off. You also get the ride drawn from your GPS, your cornering, and the power to hold each speed in the air you actually rode in.

Every free result says "Confirm this in advanced analysis.", because a single ride read quickly is a good answer, not a checked one. Paying keeps your rides, unlocks the full workbench, and shows you the give or take.

Comparing two setups

Two numbers that look different can still be nothing but noise. Each setup gets a CV%: how much your own repeat laps disagreed with each other. Under 3% is reliable. And a gap between setups smaller than the day's own scatter is within noise: real on paper, meaningless in practice. Ignore it and test the next thing.

Team analysis: one road, every ride

Everything above is about one ride. Team analysis is the other question: the same road, ridden more than once, by you on different days or by different riders. It is where you find out whether a change actually made you faster on real tarmac.

It comes with a coach or team plan, not with the athlete plan. On an athlete plan the page still lists your roads and shows you what is behind the lock.

How a road gets there

You do not draw a road. When a ride is added, we read its GPS and look for a road we already know. If it matches one, the ride joins it. If it does not, that ride starts a new road. So a road builds itself out of the rides you upload, and the number beside it is how many of your rides are on it.

Two things follow from that. A ride with no GPS — a turbo session, a velodrome — cannot land on a road, and it says so instead of failing quietly. And a road is shared: if two riders ride the same tarmac, they are on the same road. You see only your own rides on it, plus the rides of any athlete you coach.

Pick the road, then ask it four questions

Click a road and the newest few rides on it are compared for you straight away. Then four tabs ask four different things about the same road, the same rides and the same stretch, so switching a tab never re-picks anything.

• Cornering — where the bends are, and what each rider did in them. Every bend is a pin on the road. Click one and you get every pass through it: entry speed, slowest point, exit speed, how long it took, and how far behind the best pass it was. The list leads with the bend where the passes differ most, so the first row is usually the one worth reading. Corner radius is deliberately not shown. The same bend measures anywhere from 22 to 60 m depending on which pass you ask it, so two such numbers side by side are not a difference between riders. There is a button that re-finds the bends with the GPS shaken about, which tells you which pins are real and which are noise.
• Energy efficiency — how many watts each ride saved against what that same ride would have cost alone, in still air. A positive number means the day helped you: a tailwind, a wheel to sit in, or a better position. Each rider is measured against their own tested CdA, so this is not a ranking between riders. Click one pass and every other curve is redrawn as a difference from it. That is the A-against-B test: when the two passes share a rider and a day, it cancels the wind out.
• Power overlay — every ride drawn on the same distance axis, one channel at a time: speed, power, height, wind or CdA. The same kilometre is the same kilometre for all of them. Click a ride in the list to hide it and the chart zooms in on what is left; the table underneath still shows every ride.
• Time differences — where the seconds went, metre by metre. No model and no physics, just clocks. Pick a reference ride and which of its passes to measure against. Every ride starts at zero, so this is time lost on this stretch and not a race result. A ride that went round more than once gives one curve per pass.

Two controls that drive all four tabs

The range. You can type a from and to in kilometres, but the quicker way is to drag across any chart. Everything follows: the map, every table, every summary. Double-click to go back to the whole road.

The baseline. Leave the box blank and every rider is measured against their own tested CdA, which is what you usually want. Type a number and it becomes a what-if. The rows will say the number is yours and not a measurement, because a typed CdA must never borrow the authority of a tested one.

Saving a stretch of road

Click the road itself on the map and you can save that spot as a named stretch: "the drag up to the roundabout", "turn 3". Pick how long it is in the same box. A saved stretch is timed on its own two ends, so two passes are compared over the same road and not merely a similar amount of it. Picking one lights it up on every chart, so you can see where on the road you are looking.

Under 1 km, a stretch is timed but its energy is never compared. Over that short a distance the difference between two passes is wind, draft and the line you took, not your shape in the wind.

What it refuses to answer, and why

Refusals are the point of this page, not a fault in it. A number it will not give you is a number you cannot be misled by.

• An outbound leg against a return leg. The same road the other way has the opposite wind, so a gap between them is not a difference in riding. The two directions are offered separately and never share a chart. (A LAP is no longer refused: each pass gets its own curve.)
• Any stretch where the rides did not overlap. A gap in a curve means "no measurement here", never zero.
• A saving bigger than air resistance can explain. Those stretches are left blank rather than drawn, and the page counts them for you.
• A ride in a bunch. The energy tab will still answer, but the answer includes the draft, because one file cannot tell draft and air resistance apart. Rides in a group are labelled so you know.

Every tab also carries a short "How this tab works" box on the page itself, and a "Show me round" button that walks you through the controls.

If you have a coach

Link a coach or fitter to your account and they can open and analyse your rides while the link is live. You stay the owner of everything, and you can cut the link whenever you want.

The words we use

Plain meanings, so nothing on a screen has to be guessed at.

• CdA — how well you cut through the air. Smaller is faster. Around 0.30 sat up on a road bike, near 0.20 or below in a good time-trial position.
• Ride — one file off your bike computer.
• Road — a piece of tarmac we recognised from GPS. Rides join it.
• Pass — one trip through a bend or along a stretch. One ride can make several passes of the same bend.
• Stretch — a named piece of a road that you saved to come back to.
• Baseline — the CdA a ride is measured against. Normally that rider's own tested number.
• Still air — no wind at all. It is the fair thing to measure a ride against, because it is the same on every day.
• CV% — how much your own repeat laps disagreed with each other. Under 3% is reliable.
• Give or take — how far a number could be out. Measured from your own ride, not guessed. A wide one means test again before you act on it.
• Within noise — a difference smaller than the day's own scatter. Real on paper, meaningless in practice.

The full method

How the analysis works covers the whole method, with a worked example from a real ride: the virtual-elevation trace, the cross-checks, the wind sensors we read, and where the give or take comes from. For the test protocol in depth, see the AeroEdge Academy course.