Note: Examples in this answer are drawn from reviewing D&D 5e homebrew content. This is the most common use case and one that most users will understand, however most of this advice equally applies to other systems.
1. Break it down into features
Reviewing an entire class, subclass or race and be difficult to do all at once. The variety of features and the way they are distributed make it impossible to provide a single correct answer.
A class may be overpowered because of a feature it receives at 11th level, but be otherwise under-powered up to this level. Attempting to simply assess this class as a whole would leave an inaccurate impression of the overall power of the class and how it will play out.
2. Compare to the closest existing official content
I find the best way to review homebrew is to find the nearest equivalents from official content and compare them feature by feature. It is usually a good idea to draw from at least two examples in your analysis. Compare the relative strengths and situational usefulness of the features.
Some examples of how to choose comparisons:
- A variant subclass; this is particularly easy as you can compare it to the other subclasses for that class.
- A spontaneous full-caster; look at the bard and sorcerer for comparisons.
- A martial class with some spells; you would compare it to the ranger and paladin.
- A subclass adding spells to an otherwise mundane class you should compare it to the arcane trickster rogue and eldritch knight fighter subclasses.
- A spell; find other spells with similar effect as well as spell of the same level.
- An item; try to find items with equivalent properties or rarity. This is simpler in systems with more robust item creation rules like Pathfinder.
3. Look for consistency with existing content
One thing to watch out for is homebrew that breaks away from the existing design patterns for a game. Examples of things to look for in D&D 5e reviews:
- Subclasses giving benefits at the wrong levels
- Features granting static bonuses instead of advantage or proficiency
- Anything that can break bounded accuracy
- Healing that doesn't require expending a limited resource
You also want to look for the type of feature that is granted. Do all other subclasses grant an out-of-combat situational buff at 6th level? Then this homebrew should too. Watch for homebrews that are too focused on a single use case (usually combat), particularly ones that focus more than the existing content.
4. Provide a summary
With your analysis broken up by feature, it can be hard to determine what your overall verdict is. So like the essays you wrote in high-school, you want to add a conclusion with a summary of your review. Here is an example from one of my answers:
This class overall is likely underpowered. Particularly in the early levels where is lacks offensive spells and slot recovery abilities. Once the 14th level ability kicks in, it is probably ok in balance terms but could use some work on thematics.
5. Check if their homebrew achieves their goal
The typical reason for someone to homebrew is that the user has some particular concept or theme in mind that the current content doesn't quite hit. They should give this information in the question, but if they haven't, feel free to ask for it in the comments.
If someone is going for a necromancer theme but their spell choices don't align with that, it is a good idea to let them know that their homebrew might not quite achieve what they were hoping. An example from one of my answers:
You say your intent is to add to the offensive capabilities of the class, but only the spells available at 9th level actually deal damage. The rest are control/healing/buff spells. I don't think this quite hits the theme you were going for.
6. Provide suggestions but don't homebrew for them
"Homebrew this for me" questions are off-topic here, for good reasons. Reviewing bad homebrew with an answer that is just a better version is fundamentally the same thing. So you should avoid rewriting their whole homebrew to a better version.
Instead, provide suggestions for which features need modifications and what those potential modifications should be. Examples of the type of advice that is useful:
- "Replace this 1st level spell with a 2nd level one, some good options are..."
- "Increase the damage of this ability in line with X"
- "Consider replacing this with Y feature from Z"
7. Encourage playtesting
At the end of the day no amount of theoretical analysis will be as good as proper playtesting. If their homebrew isn't just fundamentally broken and doesn't have major problems with balance or style, you should advise them to try playtesting it.