ggplot2 has two bar geoms, and choosing between them is most of the work. geom_bar() counts rows: ggplot(orders, aes(x = category)) + geom_bar() draws one bar per category, as tall as the number of orders. geom_col() draws the values you give it, so you summarise first and map the total to y.
Mapping a column to fill splits each bar by group. By default the pieces stack; position = "dodge" puts them side by side, and position = "fill" stretches every bar to 100 percent to compare shares. colour on a bar only changes its outline.
A common slip is geom_col() on raw rows: it stacks one thin segment per row, which looks right but isn't one bar per group.
Count orders per region, split by channel, side by side.