A big user of Big Data are the advertising industry, so it’s worth understanding how advertisers go about running online advertising campaigns and analysing the results.
Placing the Advertisements
Decide how you whether you wish to run your ad campaign:-
1. In-house. Advantage is that you will have full control of the content and placement of the advertisement.
2. Via an advertising agency. Advantage is that the advertising agency specialises in the placement of ads and knows the best way to target your desired audience.
3. Use an affiliate network– generally used where the advertiser is happy to pay for advertisements placed by publishers. Somewhat of a scatter gun approach.
Capturing Ad Statistics
Reporting captured Ad statistics
You can then choose to either:-
1. Develop in-house reports
2. Rely on reports by the affiliate networks (if you’ve chosen that route)
3. Rely on a 3rd party analytics tool e.g. Google analytics
Generally, if you’re a large organisation running ad campaigns both in-house and via affiliate networks, you’ll want to receive the statistics from the affiliate networks and report yourself. Google analytics is suitable for small companies, and affiliate networks is ok for companies who just use a single network.
What are you trying to achieve and how do you achieve it?
Some example questions that an advertiser asks to determine the success of an ad campaign:-
How many users have clicked on the ad – “click throughs”
These statistics can be captured by adding parameters to the ad hyperlink. The parameters can contain information about the campaign, the type of ad, and the source webpage and/or any bespoke parameters you may wish.
For example, this is an ad hyperlink provided by tradedoubler on behalf of an advertiser:-<a href=”http://clkuk.tradedoubler.com/click?p(256096)a(1704967)g(22300378)” title=”My Anchor Text” target=”_blank”>My Anchor Text</a>
The parameter information – p(256096)a(1704967)g(22300378) allows tradedoubler to lookup details about the ad, and increment the correct click-through count.
The parameters can be read by the web server when it receives the target webpage request and the information stored in a database.
A name for website visitors who clicked on an ad hyperlink to get to the site is a referral.
How many times has a particular advertisement been presented to the user – “impressions“
The Click-through rate-click throughs/impressions can be compared with click throughs/page views to provide information as to how changing the number of ads on the web page generates more attention or not.
Impressions can be captured by the web server which stores the Ad image requested by the web site where the Ad is placed, by incrementing the impression count for the campaign whenever the image is requested. The Ad image contains a parameter for the campaign id. For example, An Ad image controlled by the tradedoubler affiliate network looks like this:-
The parameter information after the ? allows tradedoubler to work out which campaign the ad is associated with.
How many times has a page been presented to a user on which an advertisement is located – “page views“
Click throughs/page views will provide an indication as to how successful your ad was at receiving attention.
Page views can be worked out by aggregating impression counts for all objects on the same webpage.
Note: Impressions can be more than page views as multiple ads could appear on a single page.
How many users came directly to your website as opposed to referrals?
Direct users are usually more valuable than referrals as they’ve found your site generally from a search engine or word of mouth, rather than via an ad, and are more likely to register/buy a product. A user whose come directly, won’t have any associated referral parameters in the url that they used to arrive at the site, so the server can spot this and record information about the direct visitor in a database.
How many unique visitors have there been during a defined period of time?
This is useful to eliminate skewed results e.g. where somebody could multiply click on an ad, in order to generate ad revenue from your company.
The main method of identifying a unique visitor is to store a tracking cookie, the 1st time that they land on your target webpage with an expiry period of your choice (typically 24 hours). Each time a request for a target webpage is made, the target webpage will run a script to retrieve a valid tracking cookie, if 1 exists or to set one if it doesn’t. If there is no valid tracking cookie, the script can inform the server which can log a unique visitor event. The drawback of tracking cookies is that users can disable them. Another method of tracking unique visitors is to log the http header and parameter information passed to the server when the visitor first gets to the site.
How many users who clicked through to your website actually went on to register or bought a product – “Converted”.
This can be further split down by those who immediately converted or those that converted at a later time.
Converteds can be captured by checking whether a visitor who subsequently registered or bought a product was originally a referral or not. This can be done by tying the information in the tracking cookie against the information captured about a unique visitor when they first visited the site. An immediate can be deduced by comparing the registration time against the time that the unique visitor was first detected on the website.
Where’s the best place for the Ad to be placed and which type of AD generates the most clicks?
The advertiser will want to know if a banner ad worked better than a skyscaper ad, and whether a pop-over ad worked better than a static ad, for example.
This information can be captured by embedding parameters in the ad hyperlink url. The server will extract the parameter information and store the results in a database.
Which advertising channel generated the most traffic?
For example, did more visitors arrive due to an email campaign, a mobile app or a web ad?
This information can either be deduced from information sent as part of the HTTP Header with all web page requests submitted to the server, or as a parameter added to the ad hyperlink url.
Which is the best geographic location for the ad
Ads can do better in certain countries, so this helps target the audience.
Can work this out by writing code which does a lookup on the location of the ip address. Prone to error, however, as a user can often be located in a different country to the referring web server.
Determine profile of users who converted
This can be established by asking questions of the user when he first registers on the site. The information is then used to try to improve conversion rates by better targeting the audience who are likely to be interested in your product.
For most organisations, it would be very labour intensive to manage a network of publishers who are advertising your products (many of whom are home workers) on your behalf. For this reason, there are specialist organisations referred to as affiliate networks who do this on your behalf e.g. TradeDoubler, DoubleClick. The publisher signs up with the affiliate networks and are offered various rates of commission, typically based on pay per click (PPC), to place an advertisement. The advertisement could be placed on the publisher’s own website or they might purchase a Google Adword if the commission is cost-effective. The ad contains a hyperlink which redirects the user via the affiliate network to the target landing page. The affiliate network can then record that a click through has occurred. The ad also contains a link to an image stored on the affiliate network server, so when this is served up, the affiliate network increments the impression counter for the ad, and for the publisher.