I am drawing up a list of tests I'd like to do on a clients site as they have now been pursuaded of the benefits of proper testing.

My problem is I have a LOT of tests I think need to be done. Everything from changing button colours through to whether providing geographic special offers improves offline footfall in stores.

Obviously I'm going to break them in gently with the basics. Has anyone who's done a lot more of this than me got some suggestions as to where I prioritise aside from that? Do you nearly always find checkout testing has more benefit than homepage tweaking? Image optimisation (style, size, product colour) have more impact on the bottom line than playing with filters (expanded, not expanded, order of, location on page, etc).

I do have my own ideas as to whats most important to test, but I'm basically making assumptions as to which previous assumptions are wrong. If anyone has some tips as to what should be tackled first on an average site that would be very helpful.

In short - it depends on how bad the bad bits are, and where exactly they are.

On a page level, 'Headlines' usually make the most impact, but, of course, if the headlines are already OK-ish this may not be true.
On a site level, I would look at the steps in the conversion funnel, and see if I could easily spot where the biggest leaks are, and plug them first.
Then do a conversion improvement exercise on each step, measuring conversions from one stage to the next, and also from each stage to the final sale.

In the absence of any obvious 'leak' in the chain, and assuming that end sales were high enough to allow testing, I would work backwards from the shopping cart end.


[added]
Forget doing Multivariate to start with - just do straight A/B tests on BIG changes - you will get the results quicker.
Multivariate is best for when you have most of the big changes done already.

Thank you. I know my biggest leak is an insoluble issue, but I might start with testing how to make it have least impact, based on that very good advice - I was putting off even thinking about it!

And big things are always much more fun to do

You have my reply

Gurtie,

You need pretty decent volume to get significant results with MVT.  My experience has been that MVT is frequently overkill for various reasons.  Instead, start with big conceptual tests based on your gut -- just prioritize.  I think of MVT as a tool to "tinker" with pages...  Things like color changes frequently lead to small (but useful) lifts, but more significant changes around messaging, offers, etc often are usually where the bigger wins are.  Also, it's frequently easier to implement A/B than MVT, and therefore easier to "get the ball rolling"...

I generally think of it like this:  first, test major concepts (e.g. completely different page layout strategies) and look for clear winners.  Once you pick a direction, you can tinker with it with MVT -- but figuring out your "bigger strategy" is most important.

Not sure if this makes sense, but hope it helps a bit...

Also, as an aside, which tools are you considering using for MVT, etc?  How do you plan to implement/execute on your client's site?

-seoboy

thanks SEOBoy

Thankfully volume isn't a problem for this one, so I'm hoping to get some decent results pretty quickly. Setting measurables is the bit I'm head scratching about now actually - sometimes its easy to measure a change (more/less clicks) but hard to decide how that needs to feed back in and when to call it significant (eg; a goal set by the client is that more people should be aware of a certain added value service they provide. I can test if more people look at that info, but actually I need to test if it leads to more conversions, which is going to be a lot more subtle, so my measurable is actually going to need a lot more visitors than a simple test should)

Unfortunately with high volume is coming high overhead in terms of getting anything done - to make the big tests needs to go via a project manager, the creative team, client sign off and then through development. As an extension of that I didn't get to choose the MVT software which is an add on to the CMS which is used. Looking on the bright side, they tell me it can do 'everything' so I will spec what we need to test and claim "not my problem" about how they make it do it- once we have some stuff rolling I will get more involved though, just to get a feel for it.... so i'll report back then in case anyone ever actually wants to pay money for a CMS specific testing package.