I've had the pleasure of running dozens upon dozens of conversion tests in my career for some very well established companies. These tests have covered e-commerce, lead generation, and ad revenue generation, along with every type of creative approach you can imagine.

Through this time I've learned a lot of very interesting things about testing a site to increase performance, along with some other very interesting things to consider when you perform your own A/n or Multi-Variate testing:

1. Park any emotional attachments at the door

Clients becoming emotionally attached to a creative piece is the number one problem that I've seen occur with any type of test. Whether its attachment to a writing style, image, trade dress, new creative, or old creative, it always hurts the test. Before you start any test, any stakeholders need to be aware that increasing conversion is the goal. This applies universally to any test, be it ecommerce driven, lead driven, or ad driven. As the business owner, you must have clarity in your mind about how far you're willing to go. If you ran a test and found that your trade dress and branding elements converted lower than a new version would you be willing to change it? If the verbiage that you yourself wrote - which you feel is the best thing ever written - bombs, would you separate yourself from it enough to allow positive change?

If you don't ask these questions, you can run into some very rocky situations. I've run multiple tests where the client had made up their mind before about the outcome, with highly detrimental results. These clients chose what they wanted to win, and when it didn't, insisted that their "top pick" be used instead. They would explain away hard results with empirical data like someone truly in denial. End result? Their company suffered lower conversions than they could have enjoyed.

2. Don't get too cozy with Rich Media

This goes right along with number 1 above. A client has a simple, straightforward website that immediately offers the cost, details of the product, call-to-action, and other supporting materials. They then decide that they want sexy, they want flashy, they want cool. They end up with a flash driven, AJAX bejeweled site with the production value of a Michael Bay company. Guess what? Unless you're in a very specific niche where Rich Media does work, your site will bomb.

If you are running a direct response website you need to remember what your goal is: sales. This means giving the consumer a positive experience with your site. Old school usability rules. Do they have to wait for a pre-loader, an effect? Do they have to play detective just to get a price or and "add to cart" button? If so, the odds are your shiny new site will bomb.

3. Start simple, then get complicated

You've heard this before in a variety of ways. When it comes to site testing for direct response it means making your website into an online salesman. In all the tests I've run, the two easiest things to try out that add to conversion are bullet points and emphasizing your response mechanism. The first one is straightforward, now for your response mechanism. This is the thing you are trying to get your user to do - it can be filling out a lead form, clicking an "add to cart" button, or watching a video. Adding emphasis to this mechanism to "break it out" of the rest of the page is what you're looking for.

4. Don't stop testing

You can never say that you have the "best" version of a website. You may have a great conversion for a time, but things change, and you can always make things better. Additionally, if you're like most businesses, the seasons of the year will effect you. A time generic page may do great, but why do you think retailers always put up seasonal ads?

You must get into the habit of planning tests for different seasons and events. Just like a retailer you need to find ways to elicit a consumer to purchase using anything you have, which brings us to the final point.

5. Beware anyone who tells you they "know" what's best

We've all met a version of this particular marketer. They will always claim that testing is irrelevant because they've built X amount of websites that drove Y amount of traffic and they "know what people want".

Bogus! If they knew what people wanted, they would be mega billionaires hanging out with Bill Gates and Warren Buffet. This person will regularly be disguised as a Marketing Manager, all the way up to a C-level executive. Whatever their title is, they're full of it if they make such a silly claim.

Testing is one of the few ways a business can find out what people want. The others are market testing, focus groups, and a few other highly valuable methods. Making a website based on an over-sized ego is not one them.