Google Analytics v Microsoft Adcenter Analytics - Gatineau

I was going to write this post yesterday, but Microsoft Adcenter Analytics was down, so I decided to postpone it until today. Actually, this one has been brewing in the pipeline ever since I decided that I was going to put Microsoft Adcenter Analytics on my blog. Here I am going to do a bit of analysis and show you the main differences between Google Analytics and Adcenter Analytics (Gatineau). Specifically I want to look at differences in data quality.

I think the first thing to point out is that one bit of code sits immediately before the other one. Have a look at the source of this page. Go on. Right click and select view source. You see almost at the bottom of the page that there is the Google Analytics code followed immediately by Microsoft's code. One immediately above the other and hence there shouldn't be any discrepancy in terms of one having time to load and the other not.

Lets start by setting the date range. I put it on the site on the 3rd April, so lets choose to look between the 4th April and 27th May (yesterday). And lets start at the top and look at how many visits and page impressions there were:



And here we have the same stats in Google Analytics below:



So Google is telling me that I have 1,036 page impressions, whereas Microsoft is telling me that I have only 991. That is only 4% difference. I don't think we can be too worried about that, can we? Similarly we have a 3% difference in the number of visits that we have to the site.

Lets go down a bit more of a step then instead. Below is a table showing my top content sections for this blog. I am willing to admit that the 'unique page views' that Google Analytics gives may not relate in real terms to 'visits' that Microsoft Analytics gives. The 'unique page views' seems to just be a summation of all of each pages visits. This is why Google Analytics is giving me such large figures for visits compared to Microsoft Analytics.


That actually brings me on to a very good point. Which of these is the better metric? Surely visits to a certain section would be far more useful than knowing how many unique page views there had been. Why would Google put this figure in at all - it doesn't actually help me at all as far as I can see apart from misleading me. I know my site structure is fairly poor, but say for example I had my site set up in, for example, /jobs as one folder and /articles as another. We know that people look at more jobs pages on average than article pages, so this would just indicate that my /jobs section was more popular. In this instance I think the content hierarchy works much better in MS Analytics.

Having said that, the layout for GA is much better for content. If I want to look at most viewed pages in GA, it is one report. If I want to look at most viewed pages in MS analytics then I have to look at the tree map page and see if I can work out what is going on (the other route just allows you to drill down in sections). Actually each of the pages themselves are all very similar


Next we want to look at the referrers to the site. And this is where the two systems differ, although there is a reason. The reason is that, according to Google I had 118 visits coming from stumbleupon. I've talked about Stumbleupon before on this blog (hey - don't forget to give me the thumbs up!!!), but it was interesting how Google and Microsoft dealt with this referrer in different ways. In theory the person doesn't come from stumbleupon (they click on stumble in the browser and arrive at your page), but they do have stumbleupon as their referrer. Google Decided to put this in as a referrer, whereas Microsoft decided that this was only 'Direct Traffic'. Interesting comparison.

The other thing I like about Microsoft is that it gives you a breakdown of the location of your search engines:



Especially given that quite a lot of traffic is starting to come from blog search, news search, image search. Microsoft also gives you the breakdown of which traffic is coming from Google home pages, mail pages, etc. Unfortunately it doesn't have easy access to a break down of search terms, which is a tad annoying. It does have the tree map view that I commented on last time to give you all your keywords, mind you.

Two final points about Microsoft Analytics (for the moment) is that the figures they were giving me for unique users were a load of gobbledygook (they were in the thousands - far more than page impressions and visits). I'm suggesting there is probably a bug in there somewhere that will be fixed soon. Also they don't have an export function yet (still in beta) so you can't export any of this into excel (or even send it by mail). These would be nice to have features.

Moving from HBX to SiteCatalyst - Part 1

Ok, we're finally at the point where we are about to start thinking about moving from HBX to SiteCatalyst. I think I'll keep an updated version of this going every week just as a catalogue of what is going on. Also as we move site by site we'll be able to see all the things that we did wrong (yep there'll be lots of them) and all the things that we did right and wanted to keep on doing (yep there'll be lots of... what? Eh? Oh). I think I've said on here before that moving analytics solution is not an easy business.

Ok this is how you start:

1. Select your sites for the pilot
2. Set up your SiteCatalyst Accounts
3. Set up your cookie
4. Give Developers code for pages and requirements document
   
5. Test vigorously
6. Release running at the same time as HBX
7. Check that you are collecting everything you want to use
8. Give Developers new requirements document
9. Test vigorously
10. Compare to HBX

That almost looks like a ten step plan. It isn't because some of these steps will run ad infinitum (specifically 7, 8 and 9). Tagging a site isn't a one off thing with HBX or SiteCatalyst. You have to do a series of steps so that you can get your tagging right (make sure you are collecting everything you'll need to use). Even then you will have to continue monitoring the tagging because the site will change and hence so will the names of pages (either deliberately or accidentally).

For us it was a relatively simple set of choices for our sites. Firstly we have a number of different platforms, so we needed one that spanned across them all. Our obvious choice was Farmers Weekly Interactive. Why? Because it has:

• A section on the main server for all our articles
• A section on the main server for all of the classifieds
• A section on Movable type for the blogs
• A section on Community Server for forums and photos
  
• A section on our custom built Job Service
  

This means that in theory if we can come up with a solution for this, we should be able to copy it across all of our sites.

Remember that theory's are all too easily proved wrong. All of our sites are very customised and have large sections which we haven't taken into account. We have a large image gallery that sits separately on one site and several large directories on others. Top tip here is make sure you don't think that you could do this just on one site and then copy it across all others. Make sure you test all pages and make sure they have all got the right (and unique) tags on them.

We've also got a couple of other types of site sitting around that we are going to tag as well. Firstly we have a subscription site which we will probably use Gazetteers. Although this is yet to be decided. It will allow us to work out what we want to do with our users who log in and what sort of data we want to get out of them (if any).

Finally because they are a big search B2B engine, we are going to look to see what we can do with KellySearch. They also have a nice little section on their site under each advertiser allowing them to see the usage of those pages, so we are going to have to come up with a solution for that, although that may not be in the first take.

This sounds simple enough, but remember this:

Each of these sites has a set of priorities for things they want developed. If you run in a Agile or scrum methodology which seems popular these days, then this is something that you'll need to get added into a sprint. That means bumping something else out and making someone unhappy. The trouble with web analytics implementations is that they do take developer resource away from real changes to pages and when you already have a web analytics system working, people often wonder why the changes are being made.

I've got around this one by claiming to the Business stakeholders that the move to SiteCatalyst is enforced. This is a slight white lie - we're not being forced to move from HBX (yet), but this is something that is advantageous for us to do now rather than later.

The next two steps are critical. We've identified sites, we've looked at their development schedule and worked out when we can get this pushed in and now we are going to start the whole process. Steps two and three in the list above should not be underestimated. Creating the cookie and the CNAME change associated with it (more on that next week) will take time and needs to be scheduled in before the sprint for the developers start. More importantly we need an account to play around in so that we can see the changes. Getting your Omniture rep to set these up is key to getting this working on time.

I think the last point I want to make is that this is not just a simple process of moving tools. We have a system set up at the moment that looks at metrics from HBX and has a structure about it. I've done numerous training programmes on how to use HBX and how to use the data that comes out of HBX so colleagues can do their job. It was part of my reason for setting up this blog 8 months ago. The steps after this are going to be doing part of that process again - showing the staff where to find things, what it means, how to use it. And for those that only wanted to see unique users and page impressions it's also going to be a case of setting up a series of dashboards and scheduled reports.

As one of my colleagues put it this morning "Jesus - you've got such an interesting few months ahead."

Using Hitwise to Help Drive Search Strategy

Not long ago I wrote 6 tips for using Hitwise and a bit of an introduction as to how you can use Hitwise to help your sales teams sell. The next step is to look at how you can use Hitwise to drive your search strategy in terms of which keywords you go after organically and which ones you go after in a paid search sense. Remember when you are looking into your search strategy for a site, you need to come up with a list of keywords that you want to rank for. Using Hitwise can help you come up with those keywords and monitor how you are performing compared to your competitors. Use your web analytics tool to find out if it is driving the right sort of traffic though (ie is it converting, consuming, generating leads, etc). I'm going to use MarketingWeek as my example, for no other reason than I can.

1. Find out your own search terms from Hitwise.

These will be similar, but won't match your web analytics tool. It's a good place to start, because you'll discover what Hitwise is actually telling you. Make sure you choose an appropriately long time period. It'll look a bit like the one below.


You'll notice that it cuts off fairly early in the list to terms that are below a certain threshold. That's ok, because we're not really going to pay too much attention to this list anyway. These are the terms that you already get traffic for, you probably work on these anyway, but make sure you do in the future.

Note that big box on the right of your search terms. That tells you how many terms on average the users are typing in to get to your site. It is almost certainly more than one. Here we have around about half are three or more. This is good, this means you have a long tail.

Next thing is to click on 'view engines' for a couple of keywords. This will help see if you are ranking well for terms in one engine, but not another. You'd have probably picked this up out of your analytics tool anyway.

2. Create a Custom Category of your competitors (and yourself)

Step one is to create a custom category for your key competitors. Recommendation here is to start small and work your way up. More importantly work out who is in your online space that you want to compare to. Don't include offline competitors. Now I don't know much about Marketing Week, so this is going to be a bit haphazard from here, but if I was working there I'd choose a couple of online Marketing news sites as my competitors.

There are two ways of creating these custom categories, but the one I find easiest is to pick out the main category on the site (in this case Business and Finance - Marketing) then select your sites using the little tick boxes next to the name. Then from the drop down at the bottom of the page select 'Add to a custom cat: A NEW custom Category'. The pop up will ask you to type in the name of your category, I've called mine 'Marketing' for obvious reasons.



That was all really easy. Lets look at those search terms then. Choose to 'Profile' from the top menu and then 'Search terms'. On the new screen lets choose the 'Industry' radio button and select our newly created category from the drop down. Now we have a list of search terms that drive traffic to our mini category that we've just created. This is very bespoke to our needs.

This way of picking up search terms that you can attempt to rank for allows you to use many sites to give you a 'seed list'. This seed list can be looked at, filtered and then used in your content creation plan to help drive traffic to your site. This is a much more effective way of looking at yours and your competitors search terms than doing a site by site comparison.

3. Look at your competitors Paid search

This is where we head a bit from the beaten track. Hitwise has the ability to look at the search terms that your competitors are getting traffic from by paying for it. This might mean that they intend to use it as part of their organic search effort at some point in the future, it might also mean that you can attempt to usurp them by bidding more money for those keywords (assuming it gives you a return on investment, there is no reason not to).

In the case of a publishing company like MarketingWeek, it seems to me that they are unlikely to spend money on anything other than jobs related searches and so that turns out to be the case. In this situation our competitors from the custom category don't have many jobs sections, however the ones that do sit in in a separate domain in Hitwise. It is often a good idea to be careful of this, so we can look at the search terms going to the domain jobs.brandrepublic.com and then we'll get a better competitor set.

It's vital that you don't bid on search terms unless you are going to get return on investment. If the return (ie the money you get back from the transactions) doesn't meet the expenditure in the first place, then you are wasting your money and you need to rethink your strategy.

4. Look at your Specific Search terms

Step four is to look closely at your search terms in terms of where they are driving traffic and the long tail terms.

For our site we have the term 'Marketing Jobs' being very important, so we profile Marketing Jobs on Hitwise.


Here we have the results. We do need to put this all into context though, so the first port of call we need to make is to go back to Google and look at what happens if someone types in 'Marketing Jobs'. Our favourite site is at number two at the moment in the organic rankings (in the UK), but it only receives 12% of all the traffic. I would have thought that it would want to be higher in this list and achieve more traffic - this should be one of MarketingWeek's KPIs.

To see why, lets look down a bit 'Marketing Jobs' actually only accounts for 17% of all the searches involving 'Marketing' and 'Jobs'. That is a huge tail. Lets look at that tail and see if we can get those terms into our site so that we can rank for them more efficiently. It'll probably turn out that we get more conversions for those terms anyway.