The pilot pin witness mark method pressroom_tom described is exactly right ? that's the gold standard for dialing in feed timing on progressive dies.
What I'd add is the acceleration profile matters as much as the feed window angle. On a mechanical press the feed accelerates and decelerates on a fixed cam profile. On a servo, the feeder controller usually lets you set accel/decel ramps independently. If you're getting inconsistent feed lengths (say ?0.1mm variation), it's almost always the decel ramp being too aggressive ? the strip overshoots the target position and the feeder brake can't catch it in time.
We run Komatsu H2F presses with Sankyo feeders and the magic number for us is 70% of max feeder speed with a decel ramp of about 15 degrees before the feed window closes. That gives the strip enough time to settle before the pilots engage. Going faster than that and we start seeing pilot pin wear on one side ? exactly the witness mark symptom.
One more thing: if you're running thick stock (2mm+), add a pilot release delay. The strip needs a few milliseconds to relax after the pilots release before the feeder pulls. Without that delay, the strip is still slightly compressed from the pilots and your feed length comes up short by 0.02-0.05mm. Doesn't sound like much but over 15 stations it compounds.
Also check your loop sensor position if you're using a loop control feeder. I've seen shops where the loop sensor was set for their old mechanical press speed and the servo's variable speed confused the sensor ? it would call for more strip during the dwell portion of the stroke when the press was barely moving, creating a huge loop that then whipped during the fast portion.