Designers have to do what is best for the user
First, why are you designing something? Users will not use products they don't like. If they are saddled with using them (e.g. their employer determines what they have to use at work), they will hate using products they don't like.
If you don't care if anybody ever uses your design, then you are not bound by the wishes or taste of anybody other than yourself. This is a situation more likely to occur if you are an artist trying to express yourself in design. If you are working in industry as a designer for some kind of product which is supposed to be sold to customers (or given away in return for advertising eyeballs, fame, or whatever else), then you are hurting your own goals by making a product which alienates users.
Assuming that almost none of us is in the first situation, as a user experience designer it is your job to make a product which fits the users' needs in functionality, aesthetics, and some other aspects.
This doesn't mean that the user makes the design decisions and you implement them. It means that you have to create a design yourself, and it has to be aligned with the user needs and taste. There are some points to consider here:
Sometimes designers get terrible requirements
Sadly, this happens all too frequently, for several reasons.
Some users can't predict what they will like
This is true of practically all users. No human can predict their feelings well (read up on affective forecasting for details). And while a designer has learned to evaluate the goodness of a design analytically, non-designers in general cannot do it and just react to designs on an emotional level. Add to this that a designer has also learned to imagine a design before it has been implemented, and the mental image a user is judging to predict his future reaction is fuzzy at best.
This is why you end up with users who come to you and say "The background is boring, can you put an underwater picture there? And to stay with the maritime theme, please make the text blue". This will result in poor, unreadable design, and it is likely that the user won't be happy with it once it is implemented. But he cannot see the disadvantages of his idea.
The way you should react in this case is to use objective arguments about what is wrong with it. Something like "First, it will be very hard to read blue text on blueish background. Second, if you put a picture in the background, the text gets very hard to recognize. I am afraid that it will be very hard on the eyes, no matter how pretty the sea looks." If this doesn't convince the user, throw together a very rough prototype and if they really made a big mistake, they will frequently be able to see it and let you choose a different design.
Even in this case, it is your job to try to find out what the user was trying to achieve with his demands, and offer a better solution for his goals. In the example above, you could try to include maritime elements in the visual theme, or even use a faded underwater picture overlaid by textblocks with a solid-color background.
This problem is harder when it concerns functionality and not appearance. The users may insist that their idea is great even when shown a prototype, and only recognize its drawbacks after they have had a few sessions with using the finished version. If you find yourself in the middle of such a problem, try to force user testing with functional prototypes, however low-fi, as soon as possible.
Some users have a truly terrible judgment of existing designs
This almost never happens with actual users. If you make a design which is very hard to read, almost everybody who reads it will notice this. On the other hand, other stakeholders like the product owner may insist on keeping the bad design. (See details about this case further down.) This heading assumes that it is an actual user who wants something you consider bad.
The point here is that, if you have a truly unusable design, only a tiny percentage of all users will decide that they like it. So, just gather data from multiple users. If 99% of them hate the design, change it to what you know will be more usable, even if there is a vocal 1% who insists that they absolutely love the rainbow animation coming out of nowhere and hiding all content until you have chased it away with at least three clicks. And if the majority of them like the design, then probably there is nothing wrong with it and you should leave it the way the majority likes it, even if it is, in your opinion, bad. Designers can judge designs wrong too.
Be aware that your judgments as a professional are superior where matters of usability or efficiency are concerned, but not matters of taste. If you make a clear, understated design, all harmonious lines with a limited palette muted colors, and the majority of your customers tell you that it is boring and they want to see some red and some yellow there, than nothing makes your taste better than theirs. They happen to like primary colors, and if you are making the product for them, not for yourself (see caveat at the top), you should change it.
This doesn't mean that you should just slap a few colored patches at incongruous places, even if that's what the users ask. They ask for it because this is the best solution they come up with for their problem. You should try to understand their aesthetics, even if they are foreign to you, and work with them. If they like garish primary colors, make a design based around primary colors with bold contrasts instead of just adding a few lackluster splotches to a design which was never meant to contain them.
Be aware that users don't see designs the same way as a designer, and this is fine. An object is engaging for us humans if it is sufficiently stimulating intellectually, aesthetically or emotionally. A designer is accustomed to pay attention to each detail, to see everything about a design in depth, to think about structure, color, texture, typography, affordances and whatnot. A user notices much less details. So it is common that designers prefer minimal designs, which nevertheless tell them whole stories, while users need a few "in your face" details to overcome the threshold of boredom. Comic sans is not popular in user-designed texts because all users gathered and decided to use it; it is popular because each user was independently attracted to its unusual shape which is overwhelming to the point of kitsch in a designer's eyes.
Users frequently don't have much clout in the design process
It would be great if we could gather our requirements from users. But it happens so frequently that we are paid by a corporation which is going to sell the product to the users. We get our requirements from marketeers, a product owner, or even from HiPPOs. There you have the site owner who is in love with the title which blinks between red and yellow, or the one who insists that the customer's information should only be available after four clicks through pages with promotional content.
The reason is that these people are not users. The first tactic doesn't work for them. They never use the prototypes with the eye of a user, and when the blinking title catches their attention, they don't resent the distraction like a user would, but are filled with pride at the sign of their own branding.
In matters of pure aesthetics, the users' taste is not inherently better or worse than the owner's. If the owner insists on something you consider ugly, just do it, you can't prove that it is ugly, and the users will have to live with it. But if you suspect that it is counterproductive to the owners' goals (e.g. the owner is making a site aimed at a Polish audience and wants a black-red-gold color theme), you should voice your concern and try to gather data to find out if the users will really abandon the product more than with an alternative design.
The problem comes with functionality. Arduous navigation for common tasks, unintuitive information architecture, and data gluttonous forms are not to the taste of either the users nor the user experience designer, but product owners love them, because they are so well aligned with their view of the product. Here it is your professional duty to stand on the side of the users. Do lots of user research 1) to ensure that you have judged the situation properly and the design really is bad and 2) to have some hard data with which to convince the suits that their favorite design is costing them money. This is about the only argument which can dissuade them. Of course, the research might show that it isn't such a bad design after all. Then you should leave it alone, no matter what your personal opinion is. If you can't get them to see the point even after it has been empirically proven, then in the short term you still have to make the design their way and not yours (or the users'), because this is what you are getting paid for. Long-term, I would suggest searching for another job – not only is it not fun to work for people too full of themselves, the company might also have not so rosy a future if they ignore the users' opinions.
As a hired designer, you have to design for others' needs and not for your own. If you are in a situation where your theoretical knowledge, professional experience and user testing data say that your proposal is superior for objective reasons, you should
- never automatically dismiss what the user wants – experience is on your side, but he has an intuitive detailed understanding of his workflow you can't have
- inform your discussion opponent, project management and possibly other team members why you think that the proposed design is bad, using an objective evaluation (not "pink with orange is hideous, every fool knows it")
- find out what needs or fears the user (or other stakeholder) was trying to address with their proposition and try to find a way to incorporate a professional solution for them into your design
- in the end, follow the decision your boss takes, even if it is against your position.
- in extreme cases, you are always free to leave the project (and possibly your job) if it really matters to you on a personal level. If you can't dissuade your product owner from gathering users' data ostensibly for use within the product but then selling it to advertisers, then you can make the moral decision to leave the project and the company rather than design this unethical product.