Advanced analytics - cross-campaign, channel and media tracking
Display advertising intrigues me - the best bits (like the Sony ad below) have the capacity to inspire, but measuring their success can be harder.
- Tracking the spread and effectiveness of viral, buzz or online PR campaigns
- Online metrics for measuring the effectiveness of offline and / or display advertising
- Measuring the mutual uplift between natural and paid search, online and offline display advertising and PR
So, I'm going to talk about some of the ways that we do these things (or that we are thinking about doing these things) but I'd also like to hear your thoughts and experiences....
Tracking the effectiveness of online PR
When I'm not writing here, I'm one of the directors at Distilled - and one of our products is Reputation Monitor which is a tool for monitoring what is said about you, your company or your brand online. None of us do very much plugging of our stuff here on SEOmoz, so I hope you'll all excuse a self-promotional paragraph. We built it because it was impossible to build a workflow around Google Alerts and watching loads of different RSS feeds of different sources of data meant loads of overlap. We also wanted to be able to add more filtering based on how confident we were that a particular story was or wasn't about the target being monitored (you can see how this would be a problem for us - most mentions of "distilled" on the internet are not about us!). We are working on a bunch of enhancements at the moment, but in my sole self-promotional paragraph, I'm going to end with a call to action to try it out (there's a free trial).
With it in hand, we discovered that there were other applications to areas outside reputation monitoring (such as the link building tips Tom wrote about a while back and tracking the success of viral / buzz / online PR campaigns). The way it is set up makes it particularly good at following mentions of unique phrases in viral content or in tracking links to particular pages.
When running campaigns for traffic, you would obviously track the effectiveness via analytics - and when doing linkbait for SEO purposes, you would be mainly looking at the links gathered, but when you are doing it for branding / PR benefit, the spread of the idea can be positive even if they don't link (and therefore don't send you traffic / SEO benefit).
Obviously knowing raw mentions is a pretty clumsy metric, so we are working on two enhancements to this figure:
- Visibility: how many people are likely to see / have seen that page
- Influence: essentially a measure of who is in the list of people that see the page
The first of these we are approaching via the aggregation of multiple public data points - we are hoping that with clever multi-variate regression, we can get somewhere close to a decent estimate. The single best indicator that Rand found a couple of years ago had a Pearson correlation coefficient of 0.74. By combining the data from a few sources (and perhaps throwing mozrank into the mix) we hope we can do significantly better. Watch this space!
For the second, we have a much harder task - we are considering a variety of different metrics on a range from easy to automate to very hard to automate... Various inputs we have thought of include:
- Links (especially from powerful sites) to that page - note that this requires waiting a bit - nothing gets links immediately
- Links to the site generally - from other influential places (to do this properly would require a kind of 'influenceRank' much like the trustrank in linkscape)
- Regularly being at the root of spreading stories (i.e. temporal link analysis)
- Number of comments on the story in question and other pages on the site
- RSS subscribers (if data is available)
Offline PR companies often talk about metrics such as "advertising equivalent" - i.e. how much it would have cost to buy the space that their editorial coverage achieved. If we can turn our visibility and influence metrics into an advertising cost benchmark, then mixing this with data on coverage enables us to build a money-based evaluation metric for comparing online PR campaigns. Does that sound like a good idea to anyone else?
Online metrics for measuring the effectiveness of display advertising
I was recently impressed, confused then saddened (in that order) to see the mobile phone operator (cell network for the north Americans) Orange running a massive advertising campaign in the UK (outdoor, display, newspaper / magazine, TV, transport network) with a call to action of Search Online for "I am". I was impressed that they would have something so measurable as a call to action, then confused that they didn't pick a phrase they ranked organically for (when you pick the phrase, surely SEO is easy?!) and saddened (for them) to see that it appears to have failed. For another example of a Google Trends fail, check out this comparison of hd-dvd and bluray.
With some advertising campaigns then, you can measure something specific like unique searches.
Other times, you might include a different kind of unique call to action (such as a unique URL or phone number). If you go down this route, don't forget to give people a reason to go to the URL you pick - if you use www.yourbrand.com/letmetrackyou no-one will do what you want them to. You either need to tell them there is a special offer available or (better) direct them via a dedicated URL (don't forget to 301 redirect though).
What about measuring the less direct elements, however? There are a lot of advertising campaigns that have huge branding goals. Here are our thoughts on measuring this:
- Branded search spike (either for the company, product or ad strapline)
- Tracking social network mentions (e.g. Twitter, Facebook Lexicon)
- General discussion (on blogs etc. - see tracking online PR / buzz above)
Mutual uplift between natural and paid search, online and offline display advertising and PR
At ad:tech in London last month, I attended a session where Ogilvy were talking about their 'real women' campaign for Unilever-owned Dove (I love this ad, despite not being in the target market). P&G were there and mentioned that they have complex econometric models for decoupling the effect of large numbers of inputs to estimate the mutual uplift from the interplay of their various ad channels. If that last sentence is gobbledegook to you, it is to me too...
I asked them if they could clarify, but sadly they declined.
So I'm left wondering how we should go about doing this kind of thing. How can we measure the effect on our PPC advertising of display (online display is hard enough, but what about offline display?).
This is where I would love to hear your thoughts and experiences in the comments... Any of you who work at more integrated online and offline agencies care to comment on best practices?
I feel that those of us with deep online and search experience should be able to add something to this process, but there is no point reinventing the wheel - it's far better to know what's out there before we start.
www.seomoz.org
published @ October 21, 2008