One price can be misleading
Restaurants can price the same item differently by branch, delivery platform, promotion, size, or time. Flattening all of those observations into one value may make a chart cleaner, but it can remove the reason the difference exists.
Compare like with like
A defensible comparison keeps branch, city, item identity, source, and observation date available. Then the team can decide whether it wants a branch-level comparison, a grouped market view, or a summarized range.
Ranges often tell a better story
For many commercial questions, the minimum, maximum, median, and distribution are more informative than a single average. They show whether the market is tightly priced or whether specific branches or brands are creating meaningful variation.
The result should remain inspectable
A chart or AI answer should be a doorway into the supporting observations. When someone asks why a number moved, the team should be able to trace it back instead of rebuilding the analysis from scratch.
Use the summary as the starting point, while keeping the evidence and context available behind the result.