Majority of indicators will be "timeseries" (Gini, GDP, temperatures, salary, population, light pollution). Need to store (at least on the raw storage part) the date even if the queryable data will be the "last year". Fortunatelly, that's easier to display / compute as a layer when doing some geo but ....
(Let's say that the application makes us start from a world map and we can 'add layer')
Some scale makes the data hard to valuate. E.g having the exact position of hospitals at country level (even France) might be difficult to interpret. Synthethising the data as a heat map for certain zoom level might help. https://www.utc.fr/ic05/resources/EDA-cours2-3-p2010.pdf (Page 17)
And regrouping data within zoom level might be helpful (as you know at which zoom level you are, you can add / remove geo layer)
https://nomadlist.com/ looks to have simplified by having their atomic data to "a city" and everything is link to a city.
The ultimate simplification would be to generate a map with a layer that shows the likely hood factor with intensity based on ponderated layers as described before : http://oi46.tinypic.com/5uo5lk.jpg