Big data giving insight into consumer behavior. Aivars Lode Avantce
Why Big Data Marketing Needs To Get A Whole Lot Bigger
In the old days, marketers sought to identify a target consumer and then would spend millions to catch her at the right time, in the right place, with the right message. Success was like winning the lottery, you were never quite sure what you had until the results were in.
In the digital age, we identify a target market; bombard them with banner ads, online videos and tweets. If we get a good response, we bombard them some more. Has anything really changed?
The truth is that while media has been transformed, marketing practice has not kept pace. We throw budgets into different buckets, but the decision-making process remains much the same. You develop a theory of the case, test it in-market and then, if it goes well, do it some more. A true digital revolution in marketing has yet to take hold, but it has begun.
Who Is The Consumer, Really?
An often-repeated lament in marketing, has been that we waste half of our advertising budget, but just don’t know which half. It continues to resonate because we all know that increasing marketing efficiency is a great way to improve profitability.
Conceptually, the simplest way to increase efficiency is to prevent wastage. By targeting the right consumer at the right time, the right place with the right message, we can get the most out of a marketing budget. In other words, fish where the fish are. Put your time, effort and money where they can do the most good.
In practice, however, targeting becomes more problematic. If 60% of your consumers are women, should you ignore men? If 35% of your consumers are 18-24, does that really mean that you should spend all your money on college students? A recent Catalina study found that over half of brand sales come from outside the demographic target.
We need to stop thinking about target consumers and start thinking in terms of consumer networks. Just because the daughter buys it, doesn’t mean the mother (or father or brother) won’t and beyond consumers themselves, there are advocates and detractors that can affect a purchase as well. They all matter.
Consumer Journey Or Drunkard’s Walk?
Another popular marketing concept is the consumer journey. (Here’s McKinsey’s version). In this view, prospects begin totally unaware of the wonderful brand experiences that await them until they are led on a fabulous adventure in which they consider, evaluate and purchase on a never-ending quest to becoming advocating consumers.
In reality, our behavior looks nothing like that. I might plan on having a hamburger for lunch until my friend mentions that she’s on a diet and we opt to go for salads. Then we hear a colleague rave about a new Tex-Mex restaurant and decide to go there until a client emergency has us hunkering down in a conference room and ordering pizza.
So our path-to-purchase looks less like a guided tour and more like a drunkard’s walk, in which we stumble around, bouncing from innate preferences to brand impressions to peer recommendations to personal experiences before we land on any particular purchase decision.
The Limits of a Statistical Approach
The use of simplistically blunt methods such as target consumers and sales funnels wasn’t so much a product of self-delusion as it was a marriage of convenience between available technology and the need for accountability. We never thought our models were a perfect depiction of reality, but developed techniques to suit the tools we had.
With a small, but controlled sample, you can extrapolate out to large populations and the error will be somewhat manageable and measurable for a limited amount of variables. The problem is that in acomplex system, different factors interact with each other in often unpredictable and counterintuitive ways. Micromotives often result in macrobehavior.
This issue isn’t exclusive to marketing, but occurs in many fields. The study of epidemics, for instance, has historically used statistical models with some effect, but much was left to be desired. More recently, they have begun to use new models with an agent-based approach, where whole populations are simulated.
One company, Concentric, is now applying similar techniques to marketing that incorporate a wide range of data sources, including qualitative and quantitative factors, in order to simulate the marketplace. This creates a far more accurate depiction than using econometrics to optimize one KPI or another. It also allows more flexibility.
Testing “What Ifs”
In the future, this type of agent-based modeling approach will become standard. Instead of spending hours in conference rooms arguing the merits of targeting techniques, the “consumer mindset” or what the Marketing Director overheard his daughter and her friends say, we will come up with “what if” scenarios and test them virtually.
The models will never be perfect. They will not account for emerging factors that we haven’t encountered before, nor will they calculate existing ones with absolute accuracy. What they will do is help us weed out failed approaches before we spend money on them. They will also alert us to significant market changes through post-analysis.
Most of all, an agent-based simulation approach will increase our understanding of how the marketplace works. By continually asking “what if” and testing our notions in simulation, we can continually run experiments and learn from them, at negligible cost.
Flipping The Funnel
Amazon has taken a different approach to big data and simulations. Instead of worrying about the consumer journey, its enormous scale and heavy IT investment make Amazon a market simulation unto itself. The insights they gain are then deployed to offer you what you want, when you want it, through on-site optimization and email marketing.
The strategy has paid off and Amazon dominates online retail, accounting for 45% of desktop visits and almost 60% of mobile traffic. It’s becoming more and more difficult for any company without a strong big data effort to compete in e-commerce and even those that do make the investment have a hard time matching Amazon’s data quality.
BloomReach is looking to close that gap. It offers offers its own big data solutions to companies ranging from Drugstore.com to Nieman Marcus to Crate and Barrel. By monitoring search engines and social media, BloomReach’s algorithms can identify consumer intent and can even create pages that match that intent with retailers’ inventory.
Much like agent-based simulation, this represents a fundamentally different approach to marketing. Rather than trying to surmise what’s going on by extrapolating from a small sample, big data solutions allow you to track the marketplace and adapt to changes in real time.
From the “Big Idea” to the Big Simulation
Throughout its history, marketing has always been driven by visionary ideas. A big, bold concept, backed by significant media investment, could mean the difference between a hit product and a flop. It was exciting, but risky. No amount of research or rigor could change the fact that you were, to a large degree, taking a leap of faith.
However, in the information age, we are no longer required to believe, only to imagine, test and observe. Instead of dreaming up big ideas and testing them in-market, we can test them in a virtual marketplace built by real-world, real-time data. If our wild hunch falls flat, all we lose are some bits and bytes.
Concentric has put together a nice set of case studies which report a 90% model accuracy andForrester found that BloomReach delivers conversion increases of 60%. We are, in effect, entering a new simulation economy that looks very much like the real one except that the cost of failure is negligible while the rewards of success remain massive.
And that’s the beauty of marketing driven by big data simulations, it allows us to dream bigger than ever. We can now go and test our wildest ideas, tweak them and then turn them into realities.