McLaren's Baku upgrade: aero validated, straights still the weak point

The flow-vis paint was still streaked across the rear-left corner of Lando Norris's MCL40 when the third practice session at Baku closed on 20 September 2026. For anyone who works on the technical side, that detail matters more than any number on the timing sheet. Flow-vis is normally applied on Friday and wiped clean before the final optimisation phase. Keeping it on until the last session means McLaren was still experimenting, still cross-checking track data against the wind tunnel right up to the edge of qualifying. The qualifying result was modest: Oscar Piastri third, Lando Norris fifth. The real story of the Azerbaijan Grand Prix sits somewhere else.

McLaren brought its biggest upgrade package since Hungary to Baku: revised sidepods, a new engine cover, floor, diffuser, rear suspension and rear wing. It was a visibly larger shape change than the Hungarian package, but with a smaller performance impact.

Expectations were aimed at the wrong target. McLaren had taken three pole positions in the previous four rounds. Baku is a circuit whose main straight runs longer than two kilometres, where top speed decides most of the lap time — no home track for a car that is strong in medium and high-speed corners. Norris called the car miles off after the opening practice, 1.5 seconds adrift, though most of that gap came from an aborted lap. By the third practice session, the deficit to Piastri was down to eight tenths.

Andrea Stella's explanation deserves to be read as a technical document rather than a defence. He confirmed the upgrades worked as designed and that the team was happy with the measured data and the performance impact. He also set a clear marker: this was not the Hungary package.

The subtle part is how he described Baku. The circuit has no medium or high-speed corners, meaning the car is constantly pushed into a range where it cannot express its best qualities. The upgrade simply adds aerodynamic load in all conditions without changing how performance is distributed across corner types. Extra aerodynamic load does not automatically convert into lap time when most of the deficit sits on the straights.

Norris pointed directly at where the time went: the straights. He also suffered a lock-up in Q3 and a technical issue that cost top speed.

The two statements look contradictory but fit together. An aerodynamic upgrade, however effective, cannot show up on the timing sheet at Baku when top speed has already capped the performance ceiling. A strategy machine does not run on emotion; it runs on information.

McLaren spent extra time checking its own work. The late flow-vis run shows the team was still narrowing options between configurations across the first two sessions, then converged on a single solution for both cars and wanted one final confirmation. Stella said the results came back as expected.

Under the cost cap, every upgrade package is an investment decision: which circuit do you bet on. McLaren chose to bring a large package to a race where it holds no natural advantage, and used that same race as a data laboratory ahead of the coming run of rounds. The cost is points in Baku; the benefit is correlation data for the rest of the season.

This is where a five-layer verification habit earns its keep. A claim that an upgrade works is only meaningful once you know what it rests on: measurement on the car, correlation with the wind tunnel, or the driver's feel. Based on my experience following practice sessions, I always separate those three sources. The driver reports balance feel; the engineers report correlation; the timing sheet reports the final outcome. The three rarely say the same thing.

Norris supplied feel: miles off, painful. Stella supplied correlation: data correct, performance as expected. The timing sheet supplied distance: eight tenths. The three layers contradict each other on the surface, not in substance. My mistake is called Kanté, and I do not want to forget it: a good writer updates the model when reality pushes back, instead of picking whichever layer sounds best.

Do not read the Baku qualifying result as a verdict on the upgrade. Equally, do not read the team's statement as proof it succeeded. Both conclusions outrun the data.

The blind spot sits elsewhere. McLaren's satisfaction with its correlation data is an internal signal, unverifiable from outside. When a team says everything worked as expected, it is talking about its own model — and the model is built from the assumptions the team chose to test. The loop closes on itself.

What is verifiable from outside is a different detail: McLaren kept flow-vis on the car until the end of the third practice session, meaning its final configuration decision came later than usual. That is not technically wrong, but it eats into practice time and reduces the number of qualifying simulation runs. It is a cost that never appears in any aerodynamic data table.

If the upgrade only adds aerodynamic load across all conditions without shifting the performance correlation, the question worth asking is this: is McLaren using this season to optimise each race, or to build a platform for the next one? Those two goals pull in different directions when you have a leading car and a circuit that does not suit it. Do not ask who is driving well; ask which side the system is on.

If the Baku correlation data is right, the real effect of this package will appear at circuits with medium and high-speed corners, where aerodynamic load turns into lap time. That is the marker I will return to when the season closes. A framework only matures after reality pushes back. This time, reality pushed back against both those who rushed to bury the upgrade and those who rushed to declare it a success.