If you have spent time as a solo web jockey or your job has you interacting directly with clients, you've probably been faced with the client who asks for something you feel is wrong. If you're new to this, it may seem like a dangerous situation to be in, when in reality it's a great opportunity to establish yourself as an expert and demonstrate you considerable knowledge with well-formed arguments and supporting data/examples. Sometimes the client just isn't quite getting it and it gets a bit adversarial.
I have spent a good deal of time coaching friends, employees, partners, and so on (designers, developers, architects, etc.) on the best ways to deal with clients who have trundled down the wrong path and need some correction. Usually these people feel a great deal of trepidation in realigning the client for fear of losing the business or, worse, being overridden and forced to create something with which they don't agree (but will bear their names).
Conveniently, I don't have to reiterate all the options and steps you can take. Somebody has done a pretty good job of outlining them for me. You can read the full article, How To Explain To Clients That They Are Wrong, to get all the details. You can also read a little behind-the-scenes at the author's blog. It may be worth keeping this client perspective in mind, as mentioned in the article:
...[M]any clients still regard creative digital agencies and freelancers as either kids living in their parents' basement or shady professionals out to take them for every last penny.
For those of you who just want the distilled version, here you go, with some of my own tips peppered within...
First, determine if the client is even wrong or this is just a knee-jerk reaction on your part. As freelancers, it's easy to let your ego get the best of you. Remember, the client knows his/her business, you know the technology or design rules.
Next, speak the client's language. Ask the client what the business benefit is of the request. Don't try to snow the client with techno-babble or designer-speak. If you can get the client to verbalize the business goal, you are off to a good start. If the client cannot verbalize it, perhaps the client will realize that the request is couched in vanity instead of a tangible reason.
As part of all this make sure you come off as the expert. Dress well. Speak well. Spell well. Brand yourself well. Grammarify well. Make up new words well. Be confident, support your opinions with examples and facts, and be prepared to offer alternatives (hybrid solutions even). And don't be late (to meetings, on deadlines, to bed).
Don't hide from the client or the issue, address it quickly, in person (or perhaps on the phone, but not via email) and with supporting documentation (sign-off letters, email verification). I am a fan of the direct approach. Like a Band-Aid, just tear it off, it will be over more quickly. The article I am referencing is a little less aggressive about being direct, but if you are honest and humble (add some humor) then you should be fine.
If the client is insistent, you may need to back down. The client is the one paying, after all, and if you can document that you have attempted to prevent the client from shooting his/herself in the foot, then you will be fine. Consider making the client sign off that he/she is going against your recommendation. If that's too aggressive, just send an email verification. If you are familiar with A/B testing, now is a great time to propose it. If you aren't, you should go buy a book. In the end, spend some time looking at the results of the change to see if it ended up being more effective.
If you've gotten this far, then you should also go read the comments at the original article. There are some good ones in there, sprinkled among the self-aggrandizing ones.
As of today I am an author over at Web Standards Sherpa. I wrote an article discussing the process of juggling a no-budget, tight-timeframe web site for 
