Nassim Nicholas Taleb, author of “The Black Swan,” has been an unexpected and much-appreciated influence on my view of the complex world of PPC keyword advertising.
I’ve searched somewhat in vain for analogies in our field that map well to the scenarios in fragile, global, interconnected financial systems described and prophesied in “The Black Swan.” In most obvious ways, our marketing campaigns don’t face “black swan” events in the same way as financial systems do, with their capacity for ruin. (If you look harder, though, you see looming storm clouds on the horizon.)
Better Decision-Making by Understanding Basic Statistics
Probably more important to me were Taleb’s contributions in the earlier “Fooled by Randomness,” which hammers home the point that we tend to read causation into pieces of data that are far too small, taken on their own, to have any meaning whatsoever.
Humans are poor intuitive statisticians. No, not just “other people.” All of us. Daniel Kahneman, in “Thinking, Fast and Slow,” cites studies proving that trained statisticians are just as bad in their intuitions regarding probability as anyone else.
Following Kahneman’s and Taleb’s lead, I’ve taken to insisting that advertisers spank themselves with a wooden spoon whenever they jump to conclusions based on limited data. It happens constantly and is a tendency we need to deliberately combat.
A bad analogy (here, I suggest you inappropriately blame Kahneman) would be to see the folly in any attempt to divine special attributes inherent in Anisette (one of 18 blonde, 27-year-old, blue-eyed women sharing similar characteristics; all of average degrees of loquaciousness, unusually entertained by reality TV, fans of Danish cheese, from Stockholm, Sweden) that caused her to receive a marriage proposal from Bob, a wealthy entrepreneur from Billings, Montana…other than the simple fact that the laws and/or customs in both countries prohibit Bob from marrying all 18 women. The point is, as insulting as this hypothetical example might sound, in a group of like individuals that were unusually likely to attract marriage proposals from Montana, one of them had to “convert,” and that meant the others didn’t. Statistically, it could mean that Anisette was “special,” but it could also mean that the specific result was just random, though undoubtedly – hypothetically, as established by past data in this hypothetical example – Montana and Sweden have a better than random connection. Think of Anisette as a “keyword,” and the group of 18 women as an “ad group.” To the keywords that did not convert (yet): there is nothing wrong with you. Be strong. Be proud.
Now to the task of summarizing an important 500-page book – Taleb’s influential new “Antifragile: Things That Gain from Disorder” – for our purposes in SEM.
A key ingredient in Taleb’s epistemological stance is the healthy weathering effect of experience, and the “inoculating” and “stress testing” effect of exposing systems to small stressors on a regular basis. Top-down “designed” systems – cities, a hypothetical organism, financial systems, etc. – tend to carry on “perfectly” until black swan events come along and make mincemeat of their overly-optimized, “planned” rationality.
This is life-and-death stuff in realms other than ours (for which we should be grateful). Fragile systems lead to too-severe fallout from black swan events like the Fukushima nuclear plant disaster. They are fragile because of the planners’ stance toward risk. They don’t have to be that way.
Whenever feasible, it can be wise to expose systems (or people, or teenagers) to small but survivable stressors. When you try to avoid that by “coddling” organisms or systems, you create fragile entities ill-adapted for survival, like a jellyfish without a sting. In our physical health (and carrying the analogy over to systems), Taleb draws on hormesis, the concept of small stressors on a body that lead to its strengthening.
What Is Fragile? A Model (or Useful Caricature)
To sum up, highly fragile systems exhibit this prominent hallmark:
They are planned, “top-down,” highly “tuned and optimized,” and often seem quite impressive. They are also sophisticated and speciously risk-averse, trying to shield us all from “harm” – or, for that matter, any effort, exertion, or thinking. Fragile systems are like helicopter parents.
The opposite of the defenders of fragile systems (“fragilistas,” or what Taleb sometimes calls the “Soviet-Harvard elite”) is Taleb’s fictional character Fat Tony, the non-formally educated, Brooklyn floor trader who once made a killing in oil futures due to his superior (but “street”) understanding of probability.
Insofar as systems are starved of small stressors, experience, trial and error, adaptive responses, etc.; insofar as all the trust and faith is placed in rationalistic leadership whose systems are supposed to lead to beneficial outcomes, without a hitch, they are fragile. Ideally, we’d instead build robust systems that do allow for such ongoing strengthening and development. As a curiosity, Taleb also explores the uncharted waters of anti-fragile systems that actually make more money and get better when things get volatile; these are beyond robust.
Risk Models and Convexity
This whole conversation is underpinned by the all-important (technical, easily graphable, and described in depth in Taleb’s “Antifragile”) concept of the convexity of risk. Great mischief is caused by those who seem bent on taking massive risks where the maximum possible benefit is trivial. Some of the worst examples are in the field of health care.
Some well-insulated professional Internet marketers make a sport of exposing clients to extreme risk despite the prospect of only minimal gains. They should adopt the exact opposite stance: being relentless in their pursuit of experiments where the potential maximum loss is small, the learning opportunity great, and the total profit accruing from the aggregate of many small wins and losses adds up and snowballs over time.
Despite AdWords being created by Googlers, the system is primarily robust. It’s perhaps no accident that AdWords exploded in 2002 to 2003, in the wake of the dot-com bust. AdWords rose from the ashes as a performance medium and kept attracting ad dollars because of the broad-based, variegated nature of the measurable benefits flowing to corporate advertisers of all shapes, sizes, and bents.
Proof that humans are bad at prediction? The consensus among founding Googlers and their investors in the early days of the company was that they’d make the majority of their money from licensing or enterprise applications. They didn’t think the advertising would take off. They never expected Google to become such a financial success.
Much of that anti-fragile virtue is still baked into AdWords today. But not all. Next time, I’ll describe those virtues in more depth, and also warn about those seductive elements of the system that are indeed contributors to fragility. These are best avoided, as tempting as it can be to believe in systems that cruelly appear to promise the complete absence of risk until the black swan reckoning comes.
PPC image on home page via Shutterstock.