A much better analogy would be a graph of a cake with ingredients along one axis and quantity on another. Just like you can't really tell what an IEM is going to sound like by looking at a graph, you can't really tell what the cake will taste like either. What you can tell, though, is some of the characteristics it might have.
For example, knowing what your favourite cake tastes like, and knowing which ingredients (and how much of each ingredient) is used, you can more or less tell if the cake on the graph is going to taste similar or different to the one you like, and where it might differ. If you don't have a sweet tooth, too much sugar would be a giveaway that this particular cake may not be to your taste, or that too much flower will likely make the cake too dense, or too little water too dry.
Likewise, if your favourite IEM has a certain amount of treble and you know how it's represented on a graph, and a second IEM has clearly more or less, you can start pointing at aspects of the sound of the second IEM that you think may or may not be to your liking, even without hearing that IEM for yourself. Likewise a graph can show anomalies in FR areas that you know you're sensitive (or insensitive to), like a bass hump or midrange dip that may or may to be to your liking, when compared to a known sample, your preferred IEM.
So, going back to reviews and user impressions, if someone calls an IEM bright or lively, and you look at its graph compared to a known quantity, you can tell pretty quickly where that brightness or liveliness is coming from without having heard the IEM. You may not know exactly what it sounds like, or how the different frequencies play on each other in combination with different tips, cables and sources, but it gives you a very good starting point to identify possible strengths and weakness, based on your preferences of course.