Untangling the Gordian Knot of campaign tracking – Part 2

Last time I started to look at the knotty problem of campaign tracking. Since the birth of offline marketing we have been using simplistic views of how campaigns work and which channels produce results and which channels don’t. The classic approach is that the “last click gets the sale”. In other words the channel or campaign that generated the last visit before the conversion event gets the credit for that conversion. When the internet was viewed purely as a direct response medium and campaign plans were simple, this approach might have been sufficient. But in today’s multi-channel, marketing mix optimisation world, this approach is naive.

One of the main problems here is the technology landscape that most advertisers are faced with. First of all there are a number of different systems involved. Each marketing channel will tend to have its own system for the deployment and management of marketing activities. For example, advertisers use bid management systems for PPC (Pay Per Click) search campaigns, an ad-serving systems for display ads, an email system for managing emails and so on. They may be using these tools in house or an agency may be using them on their behalf.

Each of these marketing systems will have its own data capture and reporting capabilities built in. This is important so that the channel activity can be optimised against the performance of the campaign in terms of clicks to the site and/or conversions. But it also means that whilst campaigns can be optimised within a channel (ie PPC search) it is difficult to optimise campaigns across channels because the data is sitting in different places. One of key issues then is to be able to get all the campaign response data in one place. The options here are to use a single campaign management tool across all channels or to collect all your campaign response data in one place, like your web analytics system. One you have all the data in one place you can then at least begin to look at optimising campaigns across the different channels that you use.

Campaign management tools like Doubleclick and Atlas can increasingly be used for multiple channels. Although originally developed as ad-serving technologies, they have expanded their capabilities by adding on bid management capabilities. It is also possible to track activities on other channels either through redirects or through the universal tags that are beginning to appear such as Doubleclick’s Floodlight tag. One of the advantages of the ad-serving technologies is that they can measure “post impression” effects of display advertising as opposed to just click-throughs.

Post-impression effects are where someone is served an ad impression on a site but they don’t click through. The ad impression is recorded and if that person subsequently arrives on the advertisers site, within a certain period of time, and converts, then that conversion can be attribution to the “post-impression” effect of that advertising.

For some advertisers the ability to understand post-impression effects is very important, particularly for branding campaigns. However, there are a number of issues to take on board about measuring these post-impression effects or “viewthroughs” as they are also known. First of all, advertisers need to ensure that the post-impression data is also discounted against marketing channels. For example, post-impression effects of display advertising need to be discounted against search activity. If display advertising drives increased search activity, then there is a risk of double counting the sales effect if the two channels are measured and analysed independently. This comes back to the issue of having the campaign data in a single repository.

The second issue about measuring post-impression effects is what time window after the ad has been served do you allow for the visitor to come to the site? The point of measuring post-impression effects is that ads do not always generate a direct response and that (in a similar way to TV ads) exposure to display ads build awareness and consideration which indirectly leads to conversion. But what time interval should you allow between exposure to the ad and a sale to say that there has been an effect? Although it does vary, typically advertisers or their agencies might use a window of up to 30 days. So, they are potentially attributing conversions to the fact that a visitor was served an ad up to 30 days earlier.

I think that advertisers need to consider the difference between association and correlation when evaluating post-impression effects, particularly for large brands running large campaigns. In determining the effects of marketing activity we are looking for correlation. We are looking for cause and effect. Just because someone was on a page where my ad was displayed and then they came to my site up to 30 days later, I can’t be confident that there is real correlation. It may just be a coincidence. However, if they saw my ad 5 times and then came to the site a couple of days later, I can be more confident that I am seeing some sort of real effect.

Next time I’ll be taking a look at the differences in measuring campaigns using campaign management tools and web analytics tools why they won’t necessarily be telling you the same thing. Till then…

More from Applied Insights

See more: Articles
See more: Web analytics
See more: Campaign analysis

Learn about our web analytics consulting Discover our innovative marketing applications Get to know Applied Insights