In this month’s coaching update I took the group through a great tool from Google (which is, of course, free) that makes the task of testing web pages a breeze. You can now set up a test between two web pages in minutes. Your question is – what should I test, and probably more importantly, why should I embark on a testing program anyway?

So, like most people who only have a limited amount of time to spend on their website, it helps to appreciate the core commercial benefits a successful testing plan can produce before allocating some valuable time to it. The table that follows helps explain these benefits. I have borrowed this from a recent Google presentation I sat in on. It covers what they call the multiplication effect of testing by showing how someone who is using AdWords (Google’s paid advertising option) to generate a lead flow could choose between three options when considering creating more customers: they could increase their advertising spend, raise their conversion rates by 50% or, the best of the three, do both.

So, by choosing the third option (remember this was a Google presentation so increasing the ad spend was always going to be part of the solution) this marketer is able to increase sales by more than 2-fold (from 200 to 450 customers) while only increasing their ad spend by 50%. Not bad for just taking a conversion rate up by a mere 1%.

The news on testing is even better with lead-generation websites. Here, unlike a shopping site where achieving a 1% increase in conversion requires a reasonable amount of work, with lead generation you can easily push rates upwards by 5-7% without breaking into too much of a sweat. It all comes down to the level of convincing/persuasion you need to apply. Persuading someone to fill in a form to retrieve a free report is a lot easier than asking them to part with their money and buy something from you for the first time.

So we can now see that testing makes good commercial sense – the question still remains, what should I test?

Well, for lead-generation websites the three ingredients that make up the ’conversion cocktail’ are: a) the headline; b) the image used in the header; and c) the first piece of copy below the headline. When all three are in sync then good conversions follow.

So, we recommend you get the whiteboard out and start brainstorming some options. If the discussion becomes a bit heated over what version is best, then just hand the decision over to your web visitors and let them choose. Your creative folk may have a convincing argument on why their image should be used at the top of the landing page instead of the ’boring’ one that is currently there – just test the two and see which one works the best.

This brings me onto to the types of testing you can use. The header image test example I spoke about before would be set up as a simple A/B split test. This is where the web page traffic is evenly split between the two examples. Multiple versions can be tested against each other (A,B,C,D,E, e.t.c) but to start we suggest you go with a couple of obviously different pages to see what happens.This last point is quite important. One way to decide if the difference is obvious to your visitors is to print each of them out, hold them at arm’s length and see if you can spot the difference.

The next step up from A/B testing is multivariate testing. This is where you test more than one change at a time on a single page. You may want to test three headlines, four images and two different copy introductions. I recommend you head into this territory once you have cut your teeth on its simpler cousin. It’s a bit harder to set up and requires a lot more traffic for the test to run.

The number of conversion actions each of your tests achieves will determine which one is the winner. For lead-generation websites, these conversions are usually represented by prospects successfully submitting a form to receive some sort of report. However, the success of E-commerce websites will be gauged by sales. This last point could be cause for concern. For example, most tests will require about 30 conversions per option before you can start to reliably predict a winner. If you have two options of a shopping cart page running on an e-commerce website that only produces 20 sales a month, you could be waiting three months for your test to run. This is obviously too long. One way to get around this is to test actions that lead towards a conversion action rather than the conversion action itself. For instance, you could test the state at which a person either registers or logs into their account to start a transaction. There will be a lot more visitors doing this than completing sales, and by driving more people successfully through this stage it can only help build your final sales volume.

So there you have it. I hope these few words have gone some way to convince you to allocate some of your ‘website time’ to putting in place a testing program to drive your conversion rates upwards.