Alex Schultz: APIs

Adsense Google referrer test - the results

So in my earlier post about using php setcookie() for an adsense test. I set up a test to see whether traffic sent to my site from Google behaves different to traffic sent to my site from non Google domains. The results are out and confirm my thoughts that Google users are more likely to click on Google Ads.

The bars on the above graph show the impressions given to Google referred traffic and impressions to non-Google referred traffic (totals for both are >100k impressions and >1k clicks). The lines show the CPMs acheived by each type of traffic (the last point on the red line is skewed by a new test I just started). Quite clearly the RED line (google traffic CPM) is higher than the PURPLE line (non google traffic CPM) and so Google referred traffic eyeballs on Google AdSense are more valuable than non Google traffic (hardly the biggest shock in the world).

The data is a bit richer than this though when you dig just one level deeper and look at CTR% and CPC and what exactly is driving the higher CPM from Google.

The above graph shows CTR% for Google and non Google referrers. The Google user's CTR is significantly higher than non Google users.

This graph however shows the CPC for Google and non Google traffic to my site and the non Google traffic tends to win out (have a higher CPC) or be the same. So essentially clicks from traffic sent to my site by other search engines than Google, direct visits or links on other sites are equally or more valuable than those from Traffic sent to my site by Google however traffic sent to my site from Google is far more likely to click on Google ad's than other traffic. It would be nice to test this with YPN! but they keep turning me down for membership. So for now I may rotate in Valueclick banners rather than adsense ads in some places for non Google traffic.

There are more tests to come and I am getting quite excited about them since the latest test I am running on ad positioning seems to have shown me how to triple my Google AdSense earnings (thanks Keith).

Above all this test has taught me one thing. I am massively frustrated with only 200 channels in Google Adsense for Content. This is far too low to do the testing I want to do real time, all the time, I can only run 7 AB tests on different variables at the same time and only 3 tests with 4 variables or two tests with 6 variables. What about Geography, hour of day, colour, shape, size... I want to run complex tests and really understand the interplay of variables and if I was say New York Times I would want to do that even more.

I am exploring AdLogger as a solution and similar tools but come on Google, you guys know data, you guys are excited by the interplay of variables, why not let me get access to that data and really play with it?

It's like a pinata

Jon Stewart is incredible I loved his statement earlier this week on the daily show "It's like a pinata: Beat them hard enough with a stick and... sweet candy inside".

Advertising to Students: facebook's a win for banks

Personal banking is an amazing example of an industry that is embracing the net for all it is worth, becoming a better experience for customers by doing so and making more profits too. I am actually always blown away by how interesting an industry personal banking truly is. An acquaintance used to work at Barclay's bank in the UK and he suggested to me that on average an in bank visit (fully loaded with hr and real estate) costs the bank about 100GBP, a telephone contact costs 10GBP and an online contact costs 0.01GBP. The efficiencies are huge for shifting users online and for me the convenience of online banking is great.

The second interesting effect (for me) is switching costs. When you make the decision to join up with a bank the probability is that they will retain you as a customer for the rest of your life. Apart from the first few years of your custom with a bank you have a huge switching cost. With direct debits, payroll by EFT, mortgages, loans and so on you are really tied in tightly to the bank. For this reason a customer acquisition is hugely valuable to a bank.

Recently I have seen banks start to roll more of their advertising online and these guys are more canny than the new media guys. They understand targetting demographics in the right way, grabbing the people they want to recruit where they play/browse on and offline. The one really obvious thing I haven't seen is a bank really forming a kick arse facebook.com campaign or even a really strong strategic partnership/purchase and that blows me away. In the years when someone is first at university switching costs are much lower for their bank account, usually they don't have direct debits (and banks can now switch those for you), they only usually have mum and dad paying them money, maybe a government loan and then they usually have a bank branch in the university town. These students can switch (esp. when getting their first job), these students are using facebook tonnes and they are the most valuable demographic. Someone, somewhere get these banks to try it out.

Taking Brand Advertising Learning's Online

A relatively standard way of working out the value of a brand building campaign in old media is to advertise say Bold washing up powder in all regions but the Northwest and define that region as a control group for research. The advertiser can then watch the uplift in purchases of Bold in the regions with the advertising vs. the trend they would have seen based on the control. This is obviously tricky because it is hard to pick out local PR effects and so on from the overall trends and does the Northwest really have the same trending as the rest of the country in general but in broad strokes it works.

Something that shocks me is that I have not heard of large brand advertises using this trick online and I have been browsing blogs and searching for evidence of this happening. Geotargetting through Ads on Google/Yahoo/the big banner networks to a city level is totally doable. Why not just for a test spend a few hundred thousand dollars on a banner advertising campaign in say for example York in the UK or Boston in the US on a sizeable network like adsense or valueclick. Running this campaign for a similar period to an offline campaign and looking for uplift.

Maybe I am being foolish and this is already happening, I just don't know about it or maybe brand advertisers are convinced of the value of online already (as a site owner I don't believe this... if they did I wouldn't have smiley ads continually appearing through Valueclick, Casalemedia and Burst on my site) but I love to do things that I can measure and as such this idea really excites me esp. since it could convince more advertisers to spend money online.

Geotargetting, IP sniffing and behavioral targeting has a bright future

I saw some very interesting research from forester today about the growth of online advertising. Unsurprisingly (as in the real world) they feel that the growth of direct response advertising like PPC will at some stage be outpaced by brand advertising, banners and so on, online. As I looked at this I can believe it but if fraud in CPA networks is an issue, click fraud (and low value clicks) in CPC networks is a bigger issue then CPM fraud and to a huge extent the fact that not all impressions are equal is a totally massive issue.

Look at it this way, I am not a fraudulent webmaster (I don't have the time ;) ) but of my sites my cocktail recipes site has a smaller level of traffic than my paper airplanes site (by about 50%), the cocktail recipes traffic is predominantly adults in the 18-35 bracket which is definitely of more value to banner advertisers than kids. I am not fraudulent but I do know some of my impressions are more valuable than others. 

Right now advertisers are spoilt with CPC advertising in that it is relatively easy to pick the search engine and keyword which gives you the highest return. Even if you have no advance knowledge of the value of a keyword with enough money through a process of trial and error you will eventually reach the perfect demographic. The same definitely should be possible with CPM advertising but somehow from having once been ahead of the game in targeting CPM advertising has fallen behind since PPC became huge and is pretty spammy and untargeted for the bulk of impressions. Networks like valueclick, tribal fusion and casalemedia carry loads of smiley and hotbar type ads which certainly give them a sleazy feel to work with as a small webmaster (although I admit I still take their money).

Users unknowingly tell webmasters and ad networks a lot about themselves. A large number of environmental variables are passed to the server which is sending you a site. Operating system, referring domain, browser and version, JavaScript enabled, certain tool bars installed, what ISP they are with and so on and so forth. I am positive that someone who elects to use firefox is telling you something meaningful about themself and is in a very different demographic than someone who uses Internet explorer. That should be useful to brand advertisers and a canny brand advertiser could make a killing by targeting using these variables, in fact it should be useful for Google to expose this kind of targeting too if they could pass those details through to PPC advertisers and allow them to target on that the world would get more interesting very fast indeed.

Geotargetting is already huge (at a country level) and is the first step into more granular targeting of users. Google, Yahoo and the large banner advertising networks are aggressively moving into the local space which is really cool. There is no question someone who lives in rural Yorkshire (or Montana) has different needs to someone living in downtown London (or New York). This is powerful stuff we are only just tapping now but tv advertising and radio have been making Billions from for years. A typical way of monitoring offline ad performance is to have dark regions and light regions and then observing what happens in each region.

As a webmaster I am taking baby steps into behavioral targeting using Google channels and my adsense return by referrer test illustrated in the previous example is my very first test but I have plans to start doing some advanced channel naming so that I can test hundreds of variables at the same time and serve the ads based on that complex set of variables I have tagged the user with. Over 3months I should get enough data to have a million impressions against most of the big segments. I don't however think that many of the big money spinners in the Internet optimize in this way

Google has made a move in this direction with CPM banners into their adsense network which is of course huge but it is the nature of Google that they are thought of as a direct response solution. I am sure they want the branding big bucks (and they have said as much) but I wonder if their brand has the elasticity to stretch beyond being a highly targeted direct response advertiser. Personally I don't think so, certainly not easily so their is a space here for someone else to really nail this one to the wall... could it be Yahoo!? They have the network, they have the advertisers and they have the experience, it'd be cool to see it happen.

Skype users online draws adwords ad, why?

So I sometimes don't have the foggiest idea why some things happen on the net. I have the pleasure of having a top ten position for my randomdomainname.co.uk website for my graph showing skype users online (built using a little RSS webservice Skype open up to allow you to do this) for the keyword skype users online. Incredibly this is driving traffic to my site every day which is cool but surprising (esp. since I am only about 10th and the site is entitled Alex's Tag Cloud Fun - del.icio.us & eBay ). What was even more surprising is the screen dump below:

The blue ad at the top is someone paying for traffic to his site for essentially the same skype statistics that I have developed and placed on my site. This fella is definitely doing a great job (and luckily his graphs look exactly the same as mine) and is also making those graphs free but it totally blows me away that they want to pay for traffic, esp. with no adverts on their site. If anyone has any idea why you would want to pay for this traffic I'd love some suggestions.

Using PHP setcookie() for an adsense test

I have wondered for a long time whether people who come to my site via google behave differently in respect to ads than other users. There are lots of reasons for wondering this but specifically for now I am thinking that potentially someone who has just come from google may be more likely to click on an adsense link than someone who has come from somewhere else.

Thankfully I have 10s of k's of ad impressions daily to test this on and the technology isn't exactly challenging. I have set up a cookie on every page of my largest site that drops on a users computer once every 72hrs where they have been referred from. This is done using the following code contained in an include(''); file.

if ($HTTP_COOKIE_VARS["paperairplanesreferrercookie"] == '' )
{
$value = $_SERVER['HTTP_REFERER'];
$timenow = time();
$time = time() + 7200;
$cookieset = setcookie("paperairplanesreferrercookie", $value, $time); // expire in 2 hours
};

Then lower down the page on every page I have a similar include that powers my advertising. Within this include I read in the cookie and test whether it contains the string ".google." (not perfect I know but good enough for my purposes. The code which does this is below:

if (strpos($HTTP_COOKIE_VARS["paperairplanesreferrercookie"], ".google.") === false)
{

echo $googleadvert1;

}

else

{

echo $googleadvert2;

};

Note the === in the strpos() function is because strpos() doesn't always return false when it is false therefore you have to look for responses equivalent to false (===) rather than just equal to false (==). So for me this looks like it could be a really interesting test, I will of course paste the results of the test here and if there's any substantial difference in results I will of course make changes to capitalize.

I may choose later to also split out people with cookie blocking (since I love vinny's article) and see how they behave but for now I will just let the results tick over like this for a couple of weeks.

flickr typepad widget

Today I finished my second widget for the typepad developer program, am really proud of and pleased with it. I am however always looking for comments and suggestions for improvement so if anyone has any... let me know ;)

del.icio.us typepad widget finished

Today I finished my first widget for the typepad developer program, am really proud of and pleased with it. I am however always looking for comments and suggestions for improvement so if anyone has any... let me know ;)

All your base are belonging to Google Base

So this weekend I have been completing the user sign in functionality for my cocktail making website. Am not too nervous before adding it since it is stuff I would want and I will use so I hope other people will enjoy it too but if not, I'll use it :) so it's worth it. In the process of making all the current revisions traffic has been growing to the site esp. through natural search from Google. I have made some tweaks in the interlinking of the pages in particularly to make the site more palatable to google and I have more to come.

I have been monitoring the SEO results for key terms for my site such as "cocktail recipes". The results have really frustrated me, a competitor has uploaded 10k cocktail recipes to google base and as such google has added a cocktail search box to the top of the search results page for anyone who searches for the keyword cocktail recipe or cocktail recipes so they can dig into the cocktail recipes in Google Base. This is leading to an escalation, to retain traffic I am going to have to start adding content to Google base. I don't blame my competitor, he's got a short term boost in his share of cocktail related traffic and if he hadn't done it someone else would but where does this end?

As a publisher I want traffic from google badly and to get that traffic it looks like I am going to have to gift them some content... in fact not just some content but all the content I have taken 5yrs to get hold of through kind additions by my users. At what stage will google say thanks for the content, give me an attribution and then not refer the users to my site 100% of the time (as they do now)? If I were google that's where I would be going since webmasters are always going to need the traffic google can send (even if it's a % of what it is today something is better than nothing) and so will be incentivized to put content into base and google can monetize more effectively by just taking the base content and attaching ads to your hard work without giving you a share.

Then you come to Yahoo! who I prefer since at least they are asking users directly to give them answers and they aren't asking webmasters to give away their hard work. At some stage Yahoo! will have a huge list of answers to most of the big questions being asked and then why should anyone ever leave Yahoo? Of course microsoft has their own Base type product on the roadmap at the moment too. In all of this I see a threat to the small content publisher like me... although I am confident that I can stay ahead of the big boy's search technology and user features in my specific area, I cannot continue to rely on search (in 2-5yrs time) as the main driver of users to my site... I need word of mouth, I need links, I need gorilla and viral marketing but am I moving too late?

Should the search engines be competing with every niche webmaster at the same time for content while being their lifeblood for traffic? Right now I simply don't know but I'll tell you what I am scared they will kill my little sites with their own content within the next few years.