Monday, August 29, 2011

Tips for Building your Google Adwords to help Analysis

In a perfect world either all your prospective customers would know about you or they would be able to find you in a search engines because you’d rank for all those search terms that prospective customers (even those that don’t know it) search for. Of course we don’t live in a perfect world. So in an imperfect world you need to pay the search engines money to artificially promote yourself in Search rankings.

The idea of Google Analytics was always that you could integrate your paid search data to get a better impression of how well your advertising pounds with Google were working. So what I thought I’d do is show you a bit about how it all works. Because that’s what I do!

Creating an adwords account couldn’t be easier. All you need is a credit card and a website. In fact, you don’t even really need a website, you could spend your money driving your traffic to another website.

When you’ve created your account you are ready to create your adwords. Many people jump into this part without any preplanning, but it is important to see how campaigns are structured in Adwords so that you don’t regret some of your early decisions.

So the way that Google groups is adwords is described using the diagram below. You can create many campaigns with many ad groups associated with a campaign . Keywords are then associated to ad groups, of which you can have many. To help with your future optimisation you really want to build this structure as flat as possible so you have the greatest control over keywords, however in a practical world it is difficult to manage large numbers of keywords in a flat structure, so you’ll need to think about how you can sensibly group these keywords together to manage them on a useful manner.

It’s also important to think about what each of these groupings allows you to do, so that you can manage it on a more efficient basis.


At a campaign level you can do the broad level of interesting stuff:

  • Targetting of location
  • Choosing types of ad (content network, mobile devices, Google Search, etc)
  • Set a budget for each day of your campaign
  • Ad Extensions (eg locations, extra links to the site, product details, phone numbers, etc)
  • Start and end dates for your campaign
  • Optimisation for clicks or conversions
  • Demographic bidding for the content network
Ad Group

At the Ad Group level you can get into a bit more detail about your adverts and start building them:
  • Create your ad text (headline, description, url displayed
  • Destination url
  • Cost per click (if you didn't choose that before)

This is the stage where you get the least options of things you can do. You're essentially just choosing the keywords you want to bid on!
  • Keywords bid for
  • Broad or exact match
  • Exclusions from the list
  • Cost Per Click (if you didn't choose that before!)
Of course putting it like this it can be fairly obvious how you'd want to set up your campaigns! If you're in a need of creating lots of geo targeted adverts, then you're going to need to split them between campaigns (for example). If you're creating adverts with lots of different destination urls, then you will split them between Ad Groups within a campaign. If you're going to come up with a list of keywords that you want to specify exact match or exclusions, then you're going to do it at the keyword level.

Goals and Events

The next thing that you are going to have to think about with your paid search is what you want your users to do. This is critical - you don't want to just think getting people to the site. To make it worthwhile you need to look at the return on investment. So you want to be looking at goals and events. What is it that you want your users to do when they get to the site.

This would be a useful point to mention the Google Adwords metrics. They look a bit like this:
  • Clicks - how many different clicks you have had on your adverts (note that they may not have got to your site, so they may not match your visits)
  • Impressions - this is a measure of how many times your advert was shown to users
  • Click through rate (CTR) - this is clicks divided by impressions (ie how many of those people who saw the advert clicked on it). It is a great measure of whether your ad wording is catchy
  • Average Cost per click (CPC) - this shows you how much it has cost you per click
  • Cost - rather self explanatory, this is the total cost of all the clicks
  • Average Postion - this shows you the average position that your advert has appeared on the page
As with all averages, you need to be slightly wary. If your ad appears nine times in position ten and once in position one then you are going to have an average position of nine and you'll probably have a click through rate of 10% (CTR = 10%, average position = 9). Whereas you could also legitimately be in position eight one hundred times in a row and only get one click (CTR = 1%, average position = 8)

Creating Reports

So you noticed earlier that you can put your destination url in your ad group. This means that you can use your campaign tracking at this level of the process. This is useful because it allows you an extra level of reporting that you may not have had at a higher up level.

What would be useful is if you came up with a way of naming your ad groups in a way that easily allows you to group them at the campaign level (whether you do that via SAINT or with Google Analytic's built in functionality or however your tool does it). This means you can create some very handy top level reports of visits and conversions. From this point you can choose which campaigns, ad groups and eventually search terms that you want to concentrate on for conversions!

Of course the beauty of analytics is that you can find out what the keywords people typed into Google! The url that you come from in Google contains a useful little query string with the keywords in it! Of course your analytics system doesn't know if you clicked on a paid advert or a normal link - this is why it is vital that you tag your campaigns in a way that allows the tool to know (usually with a campaign code on the landing page url).

Having this level of detail is brilliant if you want to work out if an individual keyword is driving enough revenue to be worthwhile spending it on getting there in the first place! If it isn't, then you can put it on an exclude list so that you won't bid on it. And if it is, then you can try and work out how to get more visits and more revenue by pushing your position upwards with a higher cost per click!

Really, you've got no excuse not to go and give it a go now.

Sunday, August 14, 2011

The ICO has its Data Protection Priorities Wrong

It's very difficult writing a 'let's get things in perspective' blog post, when half your country has been rioting, hence I've waited a bit before putting this together. I've been quite vociferous on the cookie legislation that has been introduced here in the UK and I've been keeping myself up to date on the other blogs on the subject. Whilst I agree with many of their sentiments, I thought maybe it would be useful to bring in a comparison with wider legislation around privacy.

For a bit of history - you can read a description of what cookies are and why they affect your privacy and other ways that you can track users. You can read about the reaction in the popular press to the new laws and you can read about my proposed solution.

If you want to see the impact of asking users to opt in to cookies, then the ICO has quite a good case study themselves. The ICO introduced an option to opt in to cookies at the top of their pages.Vicky Brock asked them to tell us about their visits using their analytics tool through an FOI request (the freedom of information act is a useful British law that allows you to ask the Government for some of their data for a modest fee - as long as it is anonymous). The result was stark - they had a 90% drop in traffic. Clearly this isn't representative of their user base any more - it'll be so biased that it will be unusable, especially for long tail keyword analysis.

Source: Vicky's Flickr

Following on from this, I've been reading some articles by other bloggers. Notably Thomas Baekdal who wrote a post entitled 'Privacy Smivacy' in which he states that Tracking is not a privacy issue, because:
Your privacy is being violated when other people decide to share what they know about you. It is not a privacy issue that people know something about you to begin with.
I think this is a fair point. And he highlights it by using a couple of examples in a follow up post. Although I think his examples are a little bit limited because whilst there should be no reason to worry about someone noticing what I do when I'm on their lawn, a better example might be having a party on someone else's lawn. You know the person, but you don't know who they have invited to the party and they're all watching what you do. Not only do you not know who they are, you don't know when they are next going to turn at a party on someone else's lawn.

However, this would be fine if it wasn't for very loud accusations of wrong doing by some organisations. Kissmetrics have been in the press because they were using 'etags' as the unique identifier for users. Kissmetrics of course were very vociferous in their response to this claiming that whilst they collected this information, they didn't share it across customers or share it with third parties. In fact there was nothing that you could do with the information. Spotify and Hulu, two of the biggest clients subsequently stopped using KissMetrics. It brings an interesting case for the 'analytics code of ethics' and I'm not sure that Kissmetrics passed this test. Users should always have the option of opting out if they don't want to participate.

But Thomas has a very good point - your privacy is being violated when other people decide to share what they know about you, so lets look at some of those situations when this has happened and the general reaction.

Council contractors lose Government data, including bank account details

In case you're not familiar with this case - a council contractor copied a load of data on to an unencrypted USB stick which was subsequently 'lost' in a pub. This is clearly against ICO's privacy guidelines - any personal information that is transported should always be encrypted. It's part of the ICO's guidelines.

What is the end result of this case? Well the ICO haven't fined anyone because they think that no harm was done as the USB was returned.

Sony Loses Customer Data from Playstation Network

Earlier this year Sony closed down their much used playstation network for several weeks. The rumour was that a beta version of the network 'accidentally' allowed subscribers to have access to all 77 million users details in an unecrypted format. 77 million users had their personal data breached and copied, including bank account details. Users were requested to change their passwords for logging into the network, but also told to check with their banks to ensure that they hadn't been charged.

What is the result of this case? Well it's still being investigated, but it would appear to be an open and shut case. Sony have even admitted that it was their fault. Will they get the maximum £500k fine?

Lush lose Customer Data

Lush, over a five month period, managed to lose 5000 customer's bank account details. There was some code that was added to the website of the company that allowed a third party to access their bank account details as users bought stuff on their site.

What was the result of the case? Well Lush have said that they'll add extra security to their site. The ICO have said that this is sufficient (obviously the 5000 who had their bank account details stolen will agree).

ICO only issues 6 fines

In fact, rather than looking at all the times that the ICO hasn't issued fines, it is easier to look at the situations where it has. In fact it has only issued six fines so far, since it has been allowed to do so. Four have been to councils and one of them was to an organisation that no longer exists.

So what we have here is a number of companies that have lost customer data and they haven't been fined. What we are going to have when they start instigating the cookie ban is an organisation that is going to fine people who collect information which isn't personally identifiable.

The ICO should be given more power to prosecute those organisations (however big they may be) through larger fines or prison sentences for those responsible through corporate liability if they do not store personal data in a secure way. They should not be looking at handing out fines to those who collect data which is not personal and which will never be shared with third parties. The priorities here are wrong.

Blog Directory - Blogged