How Fast Must You Ride to Measure an Aero Gain?

Your skinsuit is supposed to save 4 watts. You do a velodrome session, and the number moves — but is it the suit, or is it the noise? The honest answer depends on one thing most riders never think about: how fast you rode the test. The faster you go, the smaller the aerodynamic change you can actually trust. Here is the data behind that — from thousands of real velodrome runs on our platform.

What "resolution" means in aero testing

Every aero test gives you a CdA — your drag area, in square metres. But no measurement is a single perfect number; it comes with scatter. Ride the same position ten times and you get ten slightly different CdA values. The width of that scatter is your measurement resolution: the smallest genuine change you can tell apart from random noise.

If your test resolves to ±0.003 m², then a kit change that saves 0.002 m² is invisible — it drowns in the noise. A change that saves 0.010 m² stands out clearly. Knowing your resolution is what separates "I think it's faster" from "I can prove it's faster." The good news: you don't have to estimate it — every test we fit reports its own measurement error, so your resolution is a number you can read, not guess.

The physics: why speed sharpens the picture

Aerodynamic drag rises with the cube of speed. The power to overcome it is:

P_aero = ½ · ρ · CdA · v³

The power to overcome rolling resistance, by contrast, grows only linearly with speed. So as you ride faster, aerodynamics takes over a bigger and bigger share of the total power — and CdA, the thing you're trying to measure, leaves a stronger fingerprint in the data. At 52 km/h the aero signal towers over everything else; at 38 km/h it's a smaller part of a noisier whole.

That's the theory. Here's what it looks like across real sessions.

Our data: resolution by test speed

We took 931 velodrome setups across 167 sessions (indoor track, rolling-lap detection) and measured the run-to-run scatter of CdA in each one — the pure repeatability of the method. Grouped by the speed the rider actually held:

Two things jump out.

First: ride faster, resolve finer. At 51 km/h the method resolves a CdA change of about 0.0022 m² — roughly a good aero helmet or a well-fitted skinsuit. Drop to 40 km/h and the floor nearly doubles to 0.005 m² — now those same small gains hide in the noise. Controlling for how many laps were ridden and how steadily the rider held power, scatter falls about 3.6 % for every extra 1 km/h of test speed. (That trend is the regression across the well-sampled 40–53 km/h band; the sparse ≥53 km/h row — just 27 setups — ticks back up, which is small-sample noise, not a real reversal.)

Second — and this is the honest twist: in watts, the floor barely moves with speed. It hovers around 3.3–3.8 W across the well-sampled band (nudging up only in the thin ≥53 km/h cohort). Why? Because as you speed up, your CdA resolution gets finer, but each unit of CdA is also worth more watts (that v³ again). The two effects nearly cancel. So the real lever isn't "detect fewer watts" — it's that at speed you resolve a smaller aerodynamic change, which is exactly what matters when you're chasing marginal kit gains.

Methodology: how we built this table

We don't aggregate other people's numbers — we measure our own. For each setup we take the individual valid laps, compute CdA per lap from the power-and-speed physics model, and take the standard deviation across laps as the measurement scatter (CoV = scatter ÷ mean CdA). "Smallest change you can trust" is the 95 %-confidence minimum detectable difference between two same-length setups (1.96 · σ · √(2/n)); the watt column is that difference converted at the row's own speed with P = ½·ρ·ΔCdA·v³, ρ = 1.2 kg/m³. Every figure is reproducible from the stored test data. Stand: August 2026.

It's not just the average — how well you test decides your number

The table above is the median. But at the same speed, two riders can land far apart — because how steadily you hold power, how carefully you build each setup, and even the power meter you use all feed into the scatter. Across our velodrome sessions:

That's roughly a 3× gap between a clean, disciplined test and a rushed one — at the same speed. The sharpest sessions resolve a change of about 2 watts; the loosest can't reliably see 8–10. The best ones reach 0.3 % — genuine wind-tunnel territory. The lever is in your hands: ride like a metronome, nail your setup, and use a power meter you trust, and you land at the sharp end.

The part we won't oversell

• Faster riders have lower CdA in our data (0.25 at 39 km/h → 0.17 at 54 km/h) — but that's selection, not cause. Time-trial specialists ride fast and sit aero. Riding faster sharpens your measurement; it does not, by itself, make you more aerodynamic.
• How steadily you ride matters as much as how fast. Power steadiness was as strong a predictor of scatter as speed itself. A smooth, locked-in effort out-measures a ragged one at the same speed.
• There's a physical floor you can't out-test. Independent lab work puts CdA repeatability around 1–2 % even in a research wind tunnel (García-López et al., 2008), and a 2025 reliability study of indoor-velodrome testing reports a 1.67 % coefficient of variation and a smallest detectable change of 0.002 m² (Hopker et al., 2025) — essentially the numbers we see across the platform. An athlete getting off and back on the bike moves CdA by roughly that much. Our best cohorts already sit at that floor; below it, you're limited by the rider, not the method — and no amount of processing changes that.

How to run a test that resolves small gains

1. Ride the test at your real race speed — or faster. Every km/h buys resolution. Don't measure a 50 km/h position at 42 km/h.
2. Hold power like a metronome. Steady watts beat high watts. Surges widen the scatter.
3. Give it enough laps. More clean laps tighten the setup average; aim for at least 8–10 per configuration.
4. Re-measure your baseline. Bracket kit changes with a repeated baseline so you can see the session's own noise floor.
5. Trust the error bar on your result, not a rule of thumb. Every test reports its own measurement error — it already reflects your speed, your steadiness and your session. A change bigger than that error is real; a smaller one isn't yet. Riding faster and steadier shrinks it, so smaller gains become provable — but you never have to guess your own floor, we compute it for each test.

Test it yourself

Every velodrome session you fit on afasteryou.com reports its own measurement error and tells you whether a change actually cleared the noise. Start a free trial →

FAQ

Does riding faster lower my CdA? No. Faster riders tend to have lower CdA because they're aero specialists — but for you, riding the test faster only sharpens the measurement, it doesn't change your actual drag.

What's the smallest aero gain a velodrome test can detect? It depends on your test — and we tell you: every result carries its own measurement error. Typically that's around 0.002 m² (≈4 W) at 50 km/h and looser near 0.005 m² at 40 km/h, but trust the exact figure shown on your session, not a generic number. Big kit changes (helmets, skinsuits, wheels) are usually well above it and show up clearly.

How does a velodrome test compare to a wind tunnel? For repeatability, they're in the same class. Independent studies put wind-tunnel CdA repeatability at 1–2 % (García-López et al., 2008), and a 2025 study measured 1.67 % for velodrome testing (Hopker et al., 2025). Our velodrome sessions sit right there — median 1.6 %, the very best down to 0.3 % — so a well-run track test is wind-tunnel-class for telling changes apart, without the wind-tunnel cost.

Why do the best tests beat the average by so much? Because how you test matters as much as where. The sharpest 10 % of sessions resolve to 1.0 % or better; the loosest 10 % to 2.7 % or worse — a roughly 3× gap at the same speed, driven by power steadiness, setup care and equipment (your power meter included).

Why do you report a measurement error on each test? Because a number without an error bar can't be trusted. We fit each lap, check how consistently the laps agree, and report the larger of the two — a clean, steady run gets a small error; noisy data gets an honest, bigger one.

Is indoor better than outdoor for this? Yes. Outdoors, changing wind and conditions add noise no method can remove; the velodrome is where aero testing is at its sharpest. Testing on the road? See why your test road matters most.

How much data is this based on? The platform holds 2,000+ velodrome sessions and 13,000+ setups since 2018. The indoor rolling-lap dataset covers roughly 670 sessions and 4,565 setups; this resolution analysis draws on 931 setups across 167 of them. Stand: August 2026.

Aerodynamic power model: P = ½·ρ·CdA·v³. Measurement scatter, minimum detectable difference and watt conversions computed directly from stored per-lap test data, August 2026. Independent repeatability benchmarks: García-López et al. (2008), J Sports Sci 26(3):277–286 (wind tunnel, r > 0.96, CoV < 2 %); Hopker et al. (2025), Eur J Sport Sci (indoor velodrome, CoV 1.67 %, smallest detectable change 0.002 m²). No rider or operation is identified; all figures are platform-anonymous aggregates.