Could Data Modeling Make Marketing Easier to Plan and Price?
by Jonathan Salem Baskin
Published: November 08, 2010
|Jonathan Salem Baskin|
I attended the Society for New Communications Research’s Fifth Annual Symposium late last week in Palo Alto and got two days of research on the latest applications for social media. It was so refreshing to explore new media through probing questions and data-based answers that make the case for how new media work, not simply promoting them. (Full disclosure: SNCR Press is publishing my new book, “Histories of Social Media,” and I’m a senior fellow, but I’d be singing their praises if they were total strangers. Thankfully, they’re not.)
One evening over dinner, I had a really interesting conversation about predictive models and why they’re not used more in marketing planning. It seemed to us that the availability of performance and cost data, combined with computer-processing capacity, should allow for very robust predictive models of likely campaign success. Many of the larger consumer brands analyze past campaigns, and procurement officers use cost data to determine averages for ROI estimates (which are models, of a sort). But these efforts still can’t deconstruct or assess the value of creative content nor all of the variables of time, place, and context that ultimately determine the success or failure of campaigns. It seemed to us that CMOs would kill for better predictive models, yet we couldn’t cite a single brand that had fully embraced the idea.
It’s not like it’s new or anything. Eighteenth-century French mathematician Pierre-Simon Laplace imagined an intellect (think computer) that “at a certain moment would know all forces that set nature in motion, and all positions of all items of which nature is composed … nothing would be uncertain and the future just like the past would be present before its eyes.”
Modern society has never achieved such perfection, but he could have been describing the predictive models that drive technical trading in stocks and other financial instruments, or the scenario planning the armed forces use to imagine wars. Both applications aren’t just forecasts but dynamic systems that spit out recommended actions as change and circumstances dictate. The broad criteria that futurists use to contemplate trends are another example of modeling, though usually static. Weather forecasting is dynamic modeling on acid.
So why don’t we marketers possess more detailed models? Is it really that uniquely chaotic? Procurement officers don’t think so, and they appear to be gaining traction within organizations. There’s no good reason why we marketers couldn’t take responsibility for deciphering the forces and positions that accountants can see only indirectly, if at all. I think we don’t do it because any model would be imperfect and miss the ultimate value of creative, risk-taking or other things we pride ourselves on doing so well.
Instead, we let the procurement folks do it for us and gripe that they’ve discounted or ignored the true values of brand and marketing altogether.
Again, maybe you already do this, or something like it, but our riff constituted a far more explicit and detailed approach. Imagine putting a magnifying glass to attributes like:
Creative: Could you dare to categorize and sub-categorize creative content? For instance, are there X types of humor, and have some worked better or differently than others? Do different types of content yield different returns? If creative matters (and it certainly does), then risk defining it and assigning values. You’ll be wrong, but better you than somebody who doesn’t even understand it in the first place.
Planning: Your channels have values that are determined by your agency’s account planners, but do they specify deeper variables within and/or across them that affect those values? What channels are causally dependent on others, and how? There could be dozens of campaign roll-out strategies, each using different channels in different volumes and order, and every one yielding a ranking from your model.
Real-time: The real-time nature of campaign performance is rich in variables that could be identified and valued. Think the impact of foreseen or planned events (like the difference between a solitary blogger picking up your story vs. a headline in USA Today), and the potential of unexpected events (nobody picks up your story); each possibility could be reduced to numbers with meanings relative to one another. This would let you change program parameters as the thing was running.
Context: The qualities of context — what’s going on in the world, what is the likelihood of various events occurring — could be identified and added to the model, since your results are often the result of factors outside of the construct of your campaigns. Think how many zillions were wasted on marketing campaigns during the last economic malaise, for instance. Variables in your model could tell you that your campaign has problems, or opportunities, as circumstances literally dictate.
Again, it wouldn’t be perfect, but the deliverable content from your model would be a weather-like forecast of performance probability. The more details you put into it, the more likely it would be accurate. This could yield at least three benefits: Forecasts of possible performance would be a tool to more finely tune your investment decisions; variables of likely success or failure that you could track and manage from as campaigns roll out; and a standardized approach to mapping campaigns that could constitute a competitive advantage for your brand or a selling point for your agency.
It would also change the nature of the procurement debate. Imagine having the ability to make the financial cases for elements of your campaigns based on how they factor into your model instead of using crude “it worked then (or for them), so it should work this time for us” arguments or, worse, getting shot down because your fellow C-suiters simply don’t appreciate the subtle nuance of brilliant branding. Maybe this idea of modeling is playing a role in the latest ANA initiative to help CMOs and their agencies come to terms with the procurement process. If not, it should. Better models would be an alternative (or an antidote) to generic, cost-based assumptions. If you could bite the bullet and risk deconstructing campaign content, it could shift the investment debate from one of efficiency of delivery to that of efficacy of performance … which is where most marketers I know want to live.
I also know that the idea of literally building a giant computer to quietly run in the background of your department and run probability models isn’t sexy like the creative of a new viral campaign, but it’s hot stuff to me and my fellow geeks. The possible payoff of a better understanding of why things work is a powerful temptation and, as long as we could avoid the mistakes of Dr. Forbin that I witnessed in movie theaters when I was a kid, such models would transform the very conception of marketing.
If you’re already doing something like this, I’d love to know about it. If not, can you tell me why?
UT THE AUTHOR