A Funnelweb World

Last week I discovered Google Website Optimiser. It’s a great free service that allows you to perform some simple A/B and multivariate tests on a website. And it helped me to understand more about what these types of tests are, and how to set them up.

First, A/B Testing.

Put simply, A/B testing is where you test two complete pages in order to see which one has the greater conversion rate to your desired target page.

To try this out, I created a copy of my original London for Free homepage and simply removed some text and links, saving it as index2. Then, using google’s wizard, I added some javascript code to each version of the home page as well as to the target conversion page (which I chose to be my bus tour page).

The idea was to see which page was the best version when it came to encouraging people to click the link through to the bus tour page.

All well and good, except that the quality of the google code was awful. Even though it validated for google’s purposes (allowing my test to run), the W3C validator hated it. I managed to fix two elements of the code but everytime I edited the third part, the optimiser rejected it.

So I persevered with buggy code just to see how this test would work. After running it for a few days, it became clear that the page with less options encouraged greater conversion to the bus tour page. Probably fairly obvious, but it was an excellent way to get my head around how to set up and run A/B tests.

Multivariate Testing

This is where, instead of having two complete pages to test, you select only one page and indicate the sections or elements of that page that you want to vary and test. You still need a conversion target but you can test say several different images and see which one encourages greater clickthroughs to the conversion page.

I removed the A/B test code from the home page and decided to run a multivariate test on it, varying the text in one particular style element, the subfeatures section. I created two versions of the text, both aiming to encourage users to click through to the bus tour page.

Then, using the google wizard, I dropped some javascript code into both the home page and the target page, and copied my “variant code” directly into the google dashboard. The variant code is effectively a second version of the “subfeatures” ID code snippet.

(Yet another reason for using div IDs in page designs).

I’ve now been running this test for over a week and though the results are interesting, what has been of most value to me is to be able to understand how easy it is to use the google tool and how multivariant testing works.

I need to try and validate this google code and will probably also extend the tests to include multiple variants and see how it goes.

That’s tomorrow taken care of, then

Well, the redesigned site has been live for 4 months now and it seems to have been worth all the pain of recreating it (note to self - it takes longer than you think!)

In those four months its had over 70,000 visits and over 335,000 page views: comparable figures for the same period last year were 41,000 visits and 156,000 page views. (Have I mentioned lately how I love Google Analytics for making all this metrics stuff available to webmasters for free.)

In addition, since April the site has generated over US$500 in Ad Sense pocket money - not bad given I pay not a cent to advertise it.

I like this melarky of creating sites and then generating earnings from doing nothing, well maybe not nothing (see above.) If only I had time to create a few more sites…….. hmmmm, I feel another web project coming on

The site’s been live now for 11 weeks, and yesterday’s visitors helped take the adsense earnings for July alone to over $100 - in only 9 days.

Amazing! Google is one cow I would gladly kiss.

So I thought I’d have a look at the visitor stats so far.

Since 24th April when the site was relaunched, it has had over 43,000 visitors and 208,000 page views, with average visitor time on the site being around 4 minutes.

The number of click throughs to google adverts which have led to earnings is just over 1,100.

Which looks pathetic at first, but doing the calculation of click throughs to visitors, that gives me a conversion rate of 2.6% which probably isn’t actually that bad. From what I could find on the web, the average conversion rate for retail sites seems to be around 5 - 6 %.

And now that the site has had some time to settle down, it gives me a simple benchmark to use to see whether I’m achieving my aims for it or not.

My goal for the site was to provide a place that people could come to for free stuff about London, have a look around and then click through to advertiser sites.

Seems I’m getting people to the site, giving them just enough to look at - but I need to do more to get them to leave via click throughs.

And I have some ideas on how to do it - but that’s after I’ve finished the mobile version of the site, speaking of which …………… <launching dreamweaver>………..

The site’s been live for four weeks now and trends are becoming clearer.

Though site traffic is slightly down from the old version of the site (when comparing April to May) the number of pages visited per user is far higher.

Less visitors but more in depth visits - I’m happy with that. The lower number of visitors overall could be to do with easter being early this year, and I guess a true traffic comparison won’t be possible for 11 months or so.

And the earnings from google ads have shot up enormously. The figures for the past 28 days since the new site was launched have already well exceeded the earnings from the old site for all of 2008 (Jan 1 to April 23). I think most of this increase is due to the automated optimising by google of the adwords displayed, and by displaying more adwords throughout the site.

Time to take stock and think about how to improve these trends even more.

Something that links these two factors together are google keywords. Most of my traffic comes via the google search engine - and my google adsense earnings are generated by people clicking on adverts on my site.

So time to tighten up the use of keywords by improving their relevance and value: ensuring that my site appears as high as possible when people type in relevant search terms in google, and also ensuring that the most relevant (and highest value) adwords appear on my site.

Meet google’s keyword tool. Though it seems designed more for advertisers than for sites displaying their adwords, its excellent for showing two key pieces of info:

- the keywords most relevant to my site’s content that aren’t currently displaying my site in their search results

- the value of those keywords in terms of advertiser competition

Running the tool for my site, I have now identified a number of high value and relevant keywords that my site’s content isn’t optimised for. e.g. “London sightseeing tour”. Testing it in google search confirms that my site appears nowhere (well, not in the first 4 pages of results, anyway).

So as its a high value search term and deemed by google to be relevant to my site’s content, its time to do some optimising.

I now see a morning of content editing ahead of me.

Writing page descriptions into the metadata of each key page of my site which include the high value, relevant keywords, and lightly peppering the keywords through the body of my site content.

It will be interesting to see in another month what impact if any these changes have had.

The site has now been live for just ten days and I’m already seeing a significant increase in affiliate earnings.

Some broad facts and figures:

- on my old site, google adsense adverts were generating earnings of around $5 a month simply based on the single link sitting neglected on my old home page

- over the past ten days alone, google adsense earnings from the new site have reached over $30, which should mean monthly earnings of around $90 once things settle down. Incredible! Earnings are already eighteen times more than they were from the old site, before I even do any further tuning of it

- the only thing I’ve done differently is making the google advert links a little more prominent and targeting them for different site sections

- earnings from the other affiliate links via Commission Junction (ie. the banner adverts) are currently £53, again simply in ten days. And yesterday I received my very first cheque from them.

- site visits are roughly the same as they were for the old site. In the first ten days, the site received 5,757 visits and 27,936 page views, averaging 4.85 pages per visit.

So it seems as though visitors are doing what I was hoping they would do on the new site - visit, take a look around at some of the ideas I’ve suggested for free things to do and see in London and then follow an advert to leave.

The purpose of the new site is somewhat different to what I wanted to do with the old one.

Whereas the main purpose of the old site was to enable me to try out new web technologies as I learnt about them whilst also feeding my other hobby - my interest in London’s history - the focus of the new site is less so.

Though I will still use it to test out new technologies and tools that I read about, this time around I also want to use it to try out things as I learn more about affiliate advertising and visitors.

Web analytics is an area I’ve never really dug deeply into and I need to build up my knowledge in it for my new job, if nothing else.

And if the site brings in a bit of extra cash along the way as a result of my tweaking, bargain!

The re-designed London for Free site is now live.

After spending every weekend this year - and many evenings - working away at it, it feels strange to suddenly not have to write content, design pages and fix bugs.

The good news is that after only 2 days since it was launched, I am already seeing double the amount of revenue from Google Adsense. Very encouraging. Not that the revenue is major (note to taxman), of course.

There are still one or two bugs left to fix - in particular, the ajax form that worked so well on my testing server now fails to work on the “production” one unless I disable ajax from it completely. The code is exactly the same - but the hosting companies are different. So there’s another weekend’s work lined up, including emails to support desks, I guess.

I already have a list of things to add to the site for phase 2, but I’m going to take a break from developing it further for a few months or so, while I sort out changing jobs and getting my bike license.

Apart from trying to fix the bugs and monitoring / tweaking the affiliate adverts to see the effect on revenue.

More later on the impact of the re-design on advertising revenue and traffic - I need to give it a few weeks to build up some stats.

Hats off to Google - they definitely provide some excellent website tools. And of course they are totally free.
I’ve been using two of them for some time now on my current site: Google Webmaster Tools and Google Analytics.

The webmaster tools are invaluable for helping to highlight missing urls within your site e.g. for those occasions where you once thought a particular page was a great idea, created lots of internal links to it and then later deleted the page from your site.

And I have now discovered a new feature in the tools - (ok, maybe its been there for a while and I’m slow on the uptake).

They provide a section called Content Analysis. In particular, it lists the pages in your site that have duplicate short and long meta descriptions. Very handy as I update the meta data on my new site pages.

Google Analytics has always provided more information about my site than I think I will ever need to use. But in checking for the tracking code to use on my new site, I noticed that they have updated it.

Its hidden in the depths of their app and wasn’t easy to find (you can locate it under Analytics Settings > Profile Settings > Tracking Code). But in there they provide a new tracking code called ga.js.

This is Google’s explanation about this new code:

Use this tracking code to gain access to a wide range of exciting new features as they become available.

I’m not sure how long its been there - and whether the “new features” are actually available now but I was pleased to have discovered it: another thing to add to my new site.