Friday, 18 December 2009

Riding the tsunami - optimising an online marketing campaign



In the last three months we've taken web optimisation in a new direction. To date we've been trying, (emphasis on 'trying') to follow the essential principles of mvt and split testing. Putting in test content, wait a period of time, wait for visitor volume to do it's work, seeing what works in getting more customers to convert, seeing what doesn't, etc, etc. This is all well and good and will always be the core of what we do, but recently I've decided to just go hell for leather on getting an uplift and cashing in on an already positive proposition.

Our bank pioneered the cash switching incentive. This is where you can switch current accounts and the bank (us) will give you £100 cashback for the pleasure. This has been an extremly good promotion for us and we will routinely serve this promotion on the website to drive current account applications, and boy does it work. I cant give you the monetary figures for obvious reasons, but there's a very good reason other highstreet banks are now offering this same incentive. Anyway, after seeing the sudden peak in traffic that you get to the site during the campaign I decided to perform an optimisation test during the promotion period on the customer journey or offer page. I basicaly came up with half a dozen alternative page designs that focus on the promotion and were ran directly against the default page design using our normal testing tool (Maxymiser). Because of the nature of the promotion period it is definitly not a conventional MVT or split test. I had no intention of seeing what would emerge as the outright winner over time I just wanted to serve the best performing content for that moment in time. So you run the 'test' and due to the massive traffic volumes you can just sit there and watch the customers come in and crank through the test variants very quickly. You soon see what works and what doesn't. I then engage in some aggressive culling of the negative performing variants.

So after an initial trial I already know what the best page design is for this particular campaign and will serve this content again during the next campain period, knowing it will be sucessful and do it's job. This means I can just challenge this 'champion' at my leisure. But in terms of results the first run gave me a 44% uplift over the default page design in getting people to apply for the product. Wow!

UPDATE: 15th August 2011.
Coming late to the party, but getting there in the end are Get Elastic ( a very good web optimisation site) endorsing this approach (after a bit of soul searching) with an article titled 

Should You Avoid Testing During a Traffic Spike?