Your decoupling improved.
It might just be block length.
Here is a result that looks like progress. The same interval, a week apart, and aerobic decoupling falls from 3.82% to 2.12% — a 44% improvement. Power identical at 156 W. Average heart rate down from 104 to 100 bpm.
It is a convincing set of numbers, and most of it is an artefact of one detail: the first block ran 119 seconds and the second ran 180.
What decoupling actually measures
Aerobic decoupling compares the second half of an effort with the first, as a ratio of output to heart rate. If you hold the same power while your heart rate climbs, the two have decoupled and the number goes up. Hold power with a flat heart rate and it stays near zero. As a durability signal over a long steady effort it is genuinely useful.
The assumption underneath it is that your heart rate was already settled when the measurement started. Over an hour at endurance pace, it was. Over two minutes, it was not.
Heart rate is slow, and the maths does not know
When power steps up, heart rate does not follow immediately. It rises with a time constant somewhere around 30 to 60 seconds and takes two to three minutes to approach a steady value. That lag is normal physiology and nothing to correct for — but it interacts badly with a metric built on halves.
On a 119-second block, heart rate is still climbing through nearly the whole thing. The second half is therefore substantially higher than the first, and decoupling reads high. On a 180-second block, heart rate has time to level off. The second half is flatter, and decoupling reads low.
The longer block produces the better number, whatever your aerobic system did. Extend an effort by a minute and you will “improve” without training.
The other half of that result
The heart rate figure has a separate problem. 156 W at around 100 bpm is very low intensity — a cadence drill or a recovery block rather than a threshold effort. At that end of the range, heart rate is dominated by things that have nothing to do with fitness: ambient temperature, sleep, caffeine, hydration, time of day, how recently you ate.
104 to 100 bpm is four beats. For the same rider on the same block a week apart, that is comfortably inside normal day-to-day variation. It might be adaptation. Nothing in those two sessions can tell you it is.
What a like-for-like comparison needs
None of this makes cross-session comparison useless. It makes it conditional. Before two efforts can be compared on decoupling:
- Duration within about 10%. This is the one that broke the example. A minute of difference on a two-minute block is not a rounding error, it is most of the result.
- Long enough to have settled at all. Below roughly five minutes, decoupling is mostly measuring the lag. Over that, it starts measuring what you wanted.
- Similar intensity. Decoupling at endurance pace and decoupling at threshold are different measurements wearing the same name.
- Comparable conditions. A hot day inflates heart rate at any power. So does a bad night.
When the blocks do not match, the honest answer is not a smaller number with a caveat attached. It is that this pair cannot be compared on this metric, and something else should be looked at instead.
Better questions for the same data
If the sessions are structurally different, two comparisons survive that decoupling does not:
Power at a fixed heart rate. Take the last 60 seconds of each block, where heart rate has settled in both regardless of total length, and compare the power held there. Fewer assumptions, and the lag has washed out.
Heart rate at a fixed power. The same trick the other way round, and usually the more sensitive of the two at low intensity.
Neither is exciting. Both are harder to fake.
Why this matters more now
Any assistant with access to your training data will compute decoupling on two mismatched blocks without hesitating, because arithmetic does not know that heart rate is slow. It will return 44% and describe it as a meaningful aerobic adaptation signal, and it will sound certain.
The fix is not a better model. It is asking for a refusal:
Compare [label] across these sessions. Before reporting anything, check the blocks are comparable: same label, duration within 10%, similar intensity. If they are not, say so and do not report decoupling — the shorter block will read higher on heart-rate lag alone. List every block you excluded and why.
That instruction is doing real work. Without it you get an answer. With it you get an answer or a reason there isn’t one, and over a training block the second is worth considerably more.
Not medical or coaching advice. This is about what a metric can and cannot support, not what you should do on Tuesday. Fatigue, illness, and injury are not visible in training data.
More questions the data can answer · All articles · Connect your training