If you’ve had a conversation recently with anyone involved in digital marketing, you’ve likely heard “optimization” mentioned enough times that it could be easily converted into a college drinking game.
It’s a totally diluted term that has morphed into marketing fluff over the last few years. But with the increased importance of marketing investment accountability surfacing, thanks to a down economy, better technology, and increasingly more demanding clients, “campaign optimization” is making a comeback. This time, however, it’s doing so in a more mature and inherently complex manner.
The three main areas of optimization you should currently be aware of are:
1. Ad server-based optimization
2. Optimization through advanced data mining
3. And, optimization based upon attitudinal data.
In the early days (circa mid-nineties), advertisers handed their ad creative over to publishers and waited for publisher-generated reports thirty days later. Changes made to campaigns were typically done “after the fact” and while still more flexible than traditional advertising, not quite as efficient as the medium proved to eventually become. Today, most large advertisers and agencies rely on their own third-party ad servers to deliver and measure campaigns. Some continue on to track “conversions” or KPI’s (key performance indicators) and realize the old standby, clicks and impressions, turned out to not be all that meaningful.
This brings us to the first level of optimization – ad server optimization.
Campaigns optimized at this level, can be done so in real-time at the creative level, publisher level, campaign placement within a publisher or web site and by channel (e.g. search, email, banner, etc.). Campaigns utilizing ad server optimization can be modified based upon a variety of factors such as click-through rate, conversion rate, raw conversion count (direct or latent), dollar amount returned from advertising spend and a number of other variables. Campaigns optimized utilizing ad server optimization cannot be optimized using some other equally important variables such as bounce rate, user navigation path and exit points, time spent on site, specific pages visited, and so on. To be sure, ad server optimization is a great thing, but not the holygrail, especially in an environment where data and thus, knowledge is king.
Enter, web analytics and data mining.
Thanks to companies like WebTrends, Omniture and even Google (Google Analytics), we now have more access to the happenings on our web sites and a ton of raw data to boot. The double-edged sword of course…all that data! Numbers are just numbers and the real value of all that data is having someone or in most cases, several some ones who can dissect and manipulate the data and more importantly ask the right questions of the data in order to glean real insight that can be acted upon. Internally, we call this actionable intelligence. More formally in statistics and geek-speak, this is known as regression analysis and includes a myriad of techniques for modeling and analyzing several variables. Regression analysis provides a framework for understanding how the typical value of a single variable changes when any other variable or variables is modified. Being skilled in this analysis, allows for more accurate predictions of outcomes and thus, allows for the manipulation of the appropriate variables that most favorably alter outcomes to be more desirable. A very important and often overlooked fact, is that regression analysis does NOT answer the why or infer causality, but simply finds correlations between certain variables (both positive and negative). Understanding that variables are correlated allows for variables to be manipulated, but again, does not mean that one variable causes the other or the particular outcome.
Consider this:
People get wet in the rain.
Mike and Joe are people.
Mike and Joe are wet.
Mike and Joe must be in the rain.
Certainly one can deduce that getting wet and rain are variables that are correlated. But, Mike and Joe could be swimming on a sunny day. Robust analytics solutions and talented individuals can spot correlated variables and capitalize on seemingly invisible patterns and trends. There currently seems to be a large disconnect with most companies and getting the most out of the data…more and more companies have web analytics solutions and yet fewer and fewer have teams that can properly mine the data and use it in the most optimal manner. Why buy the tool, but stop short of buying the talent (whether internally or agency-side) to properly use and maximize the potential of the tool?
Maybe a simpler question to ask these decision makers who control the purse strings is this: do you think it was Jimi Hendrix or his guitar that was more important?
It will be interesting to see how costs and pricing structure evolve over the coming years in this area.
We’ve covered ad server optimization and the much more comprehensive intelligence from mining the data with the right tools and the right talent, but what if you are one of those marketers who still wants to answer the why?
This is where attitudinal research (in-house or third-party) can be beneficial.
One can, through a variety of methods, collect data from a sample large enough to be statistically relevant and attempt to answer the hidden motives of the consumer. Certainly there are some biases that can taint the data, but, there is great value in better understanding consumers’ motives, actions, and the reasons why they do and do not make purchases. These insights can typically only be gleaned through a well-structured research study. Not to diminish the importance of this type of insight (because it is certainly valuable), but I would argue that regression analysis is more powerful and ultimately, more beneficial to most advertisers. In a world where survey responses often reflect our “projected reality”, the data reflects a more truthful reality. And until we reach a day where we post pictures on our Facebook profiles that actually look like us and not the best case scenarios of ourselves, I’ll side with what we actually do versus what we say we do.