Xert’s Training Advisor: Strong Model, Demanding Inputs

Every xert training advisor review I’ve read makes the same mistake: it treats Xert as a coaching app with an unusual dashboard. It isn’t. Xert is a physiological model with a coaching app bolted on top, and the model is genuinely the best thing available to a self-coached rider. It is also the only one that will quietly rot if you feed it the wrong diet. Six months of well-executed sweet spot will make TrainerRoad’s Adaptive Training smarter and make Xert stupider. That is not a bug in the software. It’s the direct consequence of how the maths works, and almost nobody explains it before you’ve paid for a year.

MPA is a real model, not a fitness score

Most platforms describe your ability with a single number and a set of percentage bands. Xert describes it with three: Threshold Power (TP), High Intensity Energy (HIE, in kilojoules), and Peak Power (PP). From those it computes Maximal Power Available, a running figure for the hardest single-second effort you could produce right now, given everything you’ve already done in the ride.

Take a rider I’ve been tracking through a winter and a TT season. Signature at the end of March:

Fitness Signature (Xert Online, 29 Mar)
  Threshold Power      292 W
  High Intensity Energy 21.4 kJ
  Peak Power          1,043 W
  Last breakthrough    22 Mar (5:14 @ 361 W, Focus 4:38)
  Status               Fresh

At 400 W, she is spending 108 W above threshold. Divide 21,400 J by 108 W and you get roughly 198 seconds before MPA has fallen all the way to 400 W and she cannot hold it any longer. Ride at 340 W instead and the same 21.4 kJ stretches to about seven and a half minutes. Recovery runs the other way and much more slowly: sit at 200 W, 92 W under TP, and HIE refills at a fraction of the rate it drained, which is why Xert’s MPA line crawls back up after a hard climb rather than snapping back.

This is the part worth paying for. On a Garmin head unit with the Xert app, MPA is drawn live next to your power, so on the fourth kicker of a Peak District gravel loop you can see that you have 6 kJ left rather than 21, and that going over the top at 480 W will empty you completely. No other consumer tool tells you that in real time. Not intervals.icu, not TrainerRoad, not Zwift. When I put the same eighteen months of files through the platforms covered in AI Coaching Platforms, Tested, Xert was the only one that could answer “can I do this effort, right now, after what I’ve already done today” rather than “how hard was yesterday”.

The decay is the whole story

Xert recalibrates your signature when you produce a breakthrough: an effort that exceeds what MPA said was possible. No breakthrough, no new information, so the model does the only honest thing available and decays the signature toward what your recent training load supports. Low intensity work sustains TP reasonably well. It does almost nothing for HIE or PP.

Here is what that looked like across an eleven-week block of base and sweet spot, three or four rides a week, nothing above about 108% of TP:

TP fell 16 W, about 5%. HIE fell 5.5 kJ, about 26%. That asymmetry matters more than the headline number, because HIE is what sets the shape of every prescribed interval. The Advisor kept issuing workouts, kept marking them complete, and every one of them was scaled to a signature that had drifted a quarter of the way into fiction on the anaerobic axis.

Then came the compounding problem. With TP at 276, a workout calling for 4x8 minutes at 105% prescribes 290 W. In March the same session would have asked for 307 W. She passed it comfortably, produced nothing near MPA, triggered no breakthrough, and the model decayed another notch. Three weeks later the same session asked for 285 W. The Advisor was training her downhill and reporting green ticks the entire way.

The reality check came from a club 25 in mid-June: 55:41 for 40 km at 296 average, normalised 301. Xert’s TP said 276. intervals.icu, running its eFTP off the power curve, had her at 291 the whole time because it will take a best-effort 20-minute club climb as evidence and Xert will not, unless that effort clears MPA. Two models, the same files, a 15 W gap, and the more sophisticated one was the one that was wrong.

Feeding it properly costs about 40 minutes a month

None of the above is an argument against Xert. It’s an argument that Xert has an input requirement the marketing underplays, and that requirement is specific enough to schedule.

What actually resets the signature is an effort that clears MPA, which in practice means going properly deep at a duration that isn’t already well described in your recent data. If all your hard riding sits between 5 and 20 minutes, Xert knows that region and your 8-minute test will land under the MPA line without moving anything. You have to attack the parameter that has drifted. For HIE and PP, that means short and brutal: 30 seconds flat out from a rolling start, then 2 minutes recovery, then a maximal 2 to 3 minutes. For TP, a genuine 20 minute maximal on a consistent gradient.

Xert will build this for you. Set Focus duration on the Workout Designer or pick one of the prescribed Breakthrough workouts, and it constructs an interval sequence designed to walk MPA down to your current power at a chosen point, so that a maximal effort there will clear it. The one that has worked best for me is a 3x1 minute ramp into a 90-second all-out, targeted at Focus around 2:00. Total session, 42 minutes including warm-up. Do that every third week and the decay never gets a grip.

Two other things that stop the model lying to you. First, check your power sources agree: a wheel-on Tacx reading 4% low indoors versus Assioma Duos outdoors will produce a signature that is a compromise between two incompatible realities, and Xert has no idea it’s happening. Second, audit your breakthroughs rather than celebrating them. A dropout on a cold morning that produces a phantom 1,340 W for two seconds will lift PP and distort every MPA calculation for weeks. Xert lets you exclude an activity from signature calculations; use it. I’ve thrown out four false breakthroughs in the last year, all of them from the same pedal battery going flat.

Who this is wrong for

If you ride three structured indoor sessions a week all winter and never go maximal, Xert will describe you worse than a free intervals.icu account will, and you should either change how you ride or use something with lower input demands. TrainerRoad’s progression levels ask nothing of you beyond completing sessions. That’s a real advantage for a rider whose life doesn’t include regular full-gas efforts.

But if you race, or do club 10s, or ride gravel events where the surges are the event, you are already producing near-maximal efforts most weeks, and Xert will convert them into something none of the alternatives can offer. The Advisor’s daily XSS targets are only as good as the signature underneath them, and the signature is only as good as your last breakthrough.

Open your Fitness Signature panel and look at the breakthrough date. If it’s more than 28 days ago, the numbers the Advisor is working from are extrapolated, not measured, and the workout it hands you tomorrow morning is guesswork wearing a lab coat. Go and do the 42-minute session first.