So, I just finished reading Nicole Rawski's extraordinary article on the Top 10 Signs that You Don't Understand Web Analytics . It's absolute gold in terms of what it's highlighting. With that said, it's certainly no secret that data and our ability to measure is becoming a big piece of the practice of online marketing and advertising, and perhaps this blog post might make a nice companion piece to Nicole's signs. Where she focuses on how to better understand the analytics coming from your web site - there's also a growing trend for using data we collect to make our decisions for us.
Now, I see that the New York Times is reporting on how agencies are looking at the struggling financial services industry for mathematicians to start driving differentiation through data gathering and analysis. That's right, your new online campaign delivered by the same algorithms of credit default swap analysis.
As the capability for us to measure every single thing about our visitors grows, even to the point where we are able to make assumptions on their interests from past web site visits and/or clicks, two things come to mind.
1. Measurement Is A Double-Edged Sword
Measurement done correctly illustrates a benchmark that provides insight into the relevance of your message and your tactic in order that you might improve the results by changing the creative or the tactic. This includes, of course, everything from the (old school) method of just test, iterate, publish, repeat - to (as the NYT article points out) using data to optimize ads and content on the fly. The other edge of that sword is of course that it shows how woefully wrong some campaigns are. And, as the NYT times also points out, the culture at some advertising agencies (and I'd argue in most marketing departments) just aren't ready to "handle the truth" of how inefficent their current campaigns really are.
To address the double-edged sword - agencies (and again marketing departments) that deploy deep data usage will need to get permission to fail. They will need to shift from a culture that uses analytics to "prove a point", and instead use analytics to "improve content". That seems easy on paper (or rather on screen) - but it's an extraordinarily large attitudinal shift for most organizations.
Just as an example, in one of my favorite books, Call To Action, the Eisenberg brothers talked about "aiming for the gutter" on conversion rates. So, an "industry average" conversion rate is 2% to 3%. The Eisenbergs argue this is aiming for the gutter - and what a 2% to 3% conversion rate means is that 98% aren't converting. The question will be how long after the agency hires the Nobel Laureate mathemetician to derive optimization strategies will clients/management teams expect to flip this. It's just variables right? Everything can be broken down into behaviours and measurement right? And how long will it be until we comb through the numbers to get to a result that we like.
That brings me to my second point.
2. Analytics Can Be Wrong - Or More Accurately Mis-Interpreted
This is something I think, over time, is a bit more dangerous. It's a trend I see with data becoming so predominant in marketing and advertising strategies - that in some instances it's not only tearing the heart out of content and creative, but actually hampering the success it's supposed to improve. This is best exemplified in Douglas Bowman's blog post where he discusses his departure from Google. They really did test the 41 shades of blue, and also asked him to "prove his case" that a border should be 3, 4 or 5 pixels wide.
The article in the NYT also seems to be suggesting this trend with the idea that "quantitative analysts" from the financial services industry are being hired by ad agencies in order to derive strategies for optimized marketing and advertising.
Now, I want to be clear, I'm a marketing guy - and I suck at Math. But I think that recent events in the financial services sector would tell us that analytics, algorithms and data in general doesn't eliminate risk. And, in fact, and an over-dependence on them can even fill our heads with a delusional (we can't fail) mentality to the point where now even average people NOT in bond markets have heard of a Gaussian copula function.
I've blogged before on how I believe we should be content-centric and not data-centric in our marketing strategy. I think an extraordinarily interesting trend to follow is whether the practice of marketing and advertising will ultimately breed a new split practice of marketing data management and analysis.
I do believe in measurement, but I believe the KPI's I measure provide me a benchmark that gives me insight to continually improve my performance over time. Nicole makes this point much more eloquently in her piece when she says "testing is only one part of optimization, the next -- and most difficult part -- is implementation. Optimization is a continual process to make sure you are putting your best foot, ad creative, and messaging forward". In short, analytics are not a proof point that one method, tactic or piece of creative is superior to another forever - but only in as far as you use it to improve the message then and there.
I've personally seen landing pages designed by amateurs beat those designed by "best practices". I've seen a call to action work wildly in 2004 and fail in 2006. I've seen great leads from a landing page that failed to convert directly, but instead came from a conversion on a secondary call to action (why do I suddenly sound like Roy Batty at the end of Blade Runner).
Anyway, it's the balance of the art and science, and of the heart vs. head that makes the magic of marketing. And, I would caution that an over-reliance on math to determine our marketing and advertising strategy is a dangerous game - and one we've already seen play out disasterously.
