Flat 2.0

Flat 2.0 is a retreat, and it is worth naming as one. Flat design removed every cue that suggested a surface had thickness, and within a couple of years the people building flat interfaces were quietly putting a few of them back: a soft shadow under a card, one gentle gradient across a header, a slightly lighter plane for the thing that floats. The name comes from NN/g, whose usability work on the original style is the reason the correction happened at all. What makes it a style rather than a patch is that the restoration is rationed. Depth is spent only where it says something.

The parts list is short. One elevation level, or two at most, held for surfaces that genuinely sit above other surfaces. A gradient shallow enough that you have to look for it. Layered planes, each a shade apart, instead of drawn borders everywhere. And feedback on interaction: a control that lifts, brightens, or presses when the pointer meets it. What stays is everything flat design was good at, the geometry, the confident colour, the absence of fake leather and stitching, so a Flat 2.0 screen still reads as flat at a glance and only reveals the depth on inspection.

The argument for it was never aesthetic. Testing found that stripping depth stripped signifiers: people could not tell which rectangles were buttons, missed controls entirely, and slowed down on pages where colour and position were the only clues that something was clickable. Ghost buttons and unlined text fields were the worst offenders. Restoring a shadow does not restore the pretence that a button is a physical object, which is what skeuomorphism was for. It restores the much smaller claim that this rectangle is a thing you can act on, and that claim is what an affordance has to communicate before anything else on the page matters.

Material Design got to a comparable answer by the opposite route, inventing a physics of paper and elevation up front rather than walking a purge back, which is why its shadows follow a rule and Flat 2.0’s tend to be a house choice. Either way the discipline is the same in practice: pick one shadow, decide which components are allowed to use it, and treat every additional level as something that has to be argued for. A flat interface with shadows on everything is not Flat 2.0, it is just an interface with shadows on everything.