You would usually group in an array things that are related for example if you have LEDs connected to pins, you could do
const byte ledPins[] = {2,3,4,5,6,7}; // pins driving my leds
const byte ledCount = sizeof ledPins / sizeof ledPins[0] ; // number of pins calculated by the compiler
which would be better than
const byte ledPin1 = 2;
const byte ledPin2 = 3;
const byte ledPin3 = 4;
const byte ledPin4 = 5;
const byte ledPin5 = 6;
const byte ledPin6 = 7;
const byte ledCount = 6;
as you don't need to modify the count manually if you add a led later on or modify the code as you can use a for loop to efficiently "talk" to all LEDs
for (byte i = 0; i < ledCount; i++) pinMode(ledPins[i] , OUTPUT);
or the fancier Range-based for loop
for (auto && aPin : ledPins) pinMode(aPin, OUTPUT);
but if you have a pin that does not "conceptually" fit in the same group, say you have also a button or a sensor connected to an analogPin, then you would declare a special constant for those.
const byte buttonPin = 8;
const byte sensorPin = A0;
and not add those into the array.
➜ keep things organised in a way that makes sense and that will help read the code. (and that would take the same amount of SRAM memory)