Start from the question
Chart choice follows from what you're asking.
Comparison across categories → bar chart. Almost always the right answer, and the easiest to read accurately because position along a common axis is the most precise visual encoding.
Change over time → line chart. Relationship between two measures → scatter plot. Part of a whole → stacked bar or treemap. Geographic pattern → filled or symbol map. Distribution → histogram or box plot.
Pie charts
Pies are hard to read because comparing angles is much less accurate than comparing lengths. Beyond three or four slices they become guesswork.
They're defensible for a simple two-or-three-way split where the point is "roughly half". For anything else, a bar chart answers the same question better.
Dual axis, carefully
A dual axis layers two measures on independent axes. It's genuinely useful for a measure and its target, or for two different units over the same time period.
The risk is that independent axis scales let you imply any correlation you like. If both measures share a unit, synchronise the axes (right-click the secondary axis) so the comparison is honest.
When a table is right
Sometimes people need to look up an exact number, not see a pattern. A crosstab (text table) is the correct answer there, and dressing it up as a chart makes it worse.
A useful middle ground is a table with light colour encoding — a highlight table — which keeps precise values readable while making outliers visible.