January 17, 2013
Yesterday we had an amazing opportunity to talk to Andrew Goodman who talked about Randomness, Black Swans and Fragility & How They Affect SEM. ________________________________________
PPC Hero: Could you start by telling us a little bit about your background in PPC?
Andrew: It’s nearly as simple as having tinkered with AdWords Select in Mar 2002 & being blown away by its power. I had written an article in 2001, “Paid Search is Here to Stay” (about GoTo). Few believed it at the time. As a fan of Permission Marketing (1998), I also wrote an Op Ed piece in 2000 called Stay Out of My Inbox. Within weeks of AdWords’ 2002 launch I created the first Ebook on PPC tactics & released several editions. We won consulting biz from there. My book, Winning Results with Google AdWords (2005, 2008) came later. I toiled on a PhD before that, so tend to seek theory, science, models & research to improve operations. Ironically, the model du jour is Taleb’s suspicion of top-down interventionism by the “Soviet-Harvard elite,” haha!
PPC Hero: This is a bit of a heady concept. Would you be able to distill for us your overall thoughts on this topic?
Andrew: The concept of Black Swan events connects to Taleb’s “rational sounding turkey with the convincing Powerpoint.” The turkey can show “increasing statistical confidence” that “being a turkey is safe,” right up until Thanksgiving. How does that map to PPC? It isn’t as bad as the global economy, but an account can still be a ticking time bomb. It can be worse when it isn’t battle-tested and when key information is opaque. That’s fragility. Once you get the hang of it, you see examples in nature, civilization, etc. — rational plans fail, experience prevails. Even how a home is situated in relation to a hill or flood pattern can show how experience beats “smart ideas.”
PPC Hero: It’s interesting stuff. You’ve said that PPC’ers often fail to grasp the law of small numbers. Could you explain?
Andrew: I suppose the conventional statistician would point to the law of large numbers, so let’s start with that. Statistical confidence increases with larger sample sizes; that is, the probability that the result is not due to chance. That still doesn’t prove causation, but arguably makes the result much more worthy of a response.
PPC Hero: What ways should people change their approach to get a better handle on these small numbers?
Andrew: Daniel Kahneman (Thinking: Fast and Slow) argues we’re poor intuitive statisticians. Even statisticians are! We “know” not to base decisions on $25 worth of clicks, then proceed to try to “plug” just those “holes.” Counter this instinct. Same kw & 4 match types? Mentally group conversion data together. Small data is more random. #Heroview
PPC Hero: Do you have specific strategies you’d like to share to avoid missing anything major?
Andrew: I think this gets into the area of handling the unseen and unexpected: preventing Black Swans. Here is precisely where your 10,000 hours of experience help understand the mechanics of PPC. Newbies might manage all dysfunction with bids, instead of knowing that high volatility comes from broad match. A deeply experienced manager will check SQR for eg. negative opportunities as well. Same data, different outcome. Multiple automated sweeps for nonperformers can counter human frailty, too. You don’t always have to act on the data.
PPC Hero: Those are great ideas. Could you also explain the concept of black swans to those that are unfamiliar with them?
Andrew: Taleb says we confuse absence of evidence w/ evidence of absence. 2008 collapse was coming. But there was no “proof yet”. Black Swan events are more harmful when interconnected, “optimized” systems are built by “experts.” That’s fragility. Economies with less debt, boring stuff, less complicated & less opaque loans & transactions, don’t cause the big meltdowns. Our world: Google’s algorithms are top down & out of our control; G = huge source of leads. Your marketing mix may be fragile! Take the model of paying a 3rd party $x per lead, not knowing “how to PPC fish” yourself. If they fold, you’re in trouble. One of the most important services some of us provide to clients is fair transparency into historical, granular performance data.
PPC Hero: Since they are impossible (by definition) to predict, what can people do about black swans?
Andrew: Make operations more robust, even antifragile. Diversify, value experience over theory, simplify, avoid opaque systems. As an aside: Taleb puts your whole life on the table. Own a share in your cousin’s bistro, not just paper assets. Think about it. People who lost life savings to Madoff would have done well to invest in a goat farm instead.
PPC Hero: Do you have any thoughts on the future of PPC, especially considering any future black swan we may encounter?
Andrew: Google could radically “simplify” advertising packages and take over total control over PPC management, tomorrow. That would create a top-down “expert” offering that would be bad for advertisers and agencies, but it could happen. Has this occurred to me? Have I ever done anything about it? Not as much as I would like. Most of my focus is on clients. Ad revenues and advertiser reach could be gravely threatened by a full shift to mobile; 50% of what we do could be invalidated. Draconian legislation on privacy, etc. could eviscerate attribution; bomb us back to the stone age of “spray and pray.”
PPC Hero: In addition to black swans, we also have to deal with fragility. Could you explain a bit about this concept?
Andrew: Fragile systems are dazzling, “optimized,” sophisticated and speciously risk-averse, trying to shield us from all harm. They may also reside on the wrong side of the convexity of risk curve. Small upside, big potential catastrophe. Taleb notes Thalidomide in the 1960’s given to mothers to avoid morning sickness, massive damage caused. Systems and bodies thrive on adaptive responses to many small stressors. Else evolve into stingless jellyfish.
PPC Hero: In your experience, what elements of an account are more fragile than others?
Andrew: To take one example, any black box “smart” CPA bidding. Granular data is opaque. “Autopilot mode.” You learn nada. By testing no further, you consign a campaign to much lower growth and risk ceding your fate to your elite masters. Someone else benefits from adaptive learning, diverse DNA pool. It’s kept from you. You become increasingly fragile. #Heroview
PPC Hero: You’ve said that sometimes our style of managing accounts can hinder their evolution. How so?
Andrew: The best systems & strategies make a great many little bets with some chance of failure and lots of upside. The real risk is never getting over the hump due to overcaution… you wind up behind, and take larger risks on longshot gimmicks… without admitting that’s what you’re doing. In Great by Choice, Jim Collins calls it “bullet, bullet, bullet, bullet, calibrated cannonball.” Many folks are reluctant to tinker and “miss” (the “bullets”), yet wind up blowing their wad on low % hail Mary’s. The typical tassel-shoed executive, logical planner, and “best practices” guru overplans, under-experiments. Don’t misapply this advice: an “experiment” should be with *higher intent* keywords, *tighter* match types. Many failed strategies constantly look at reach potential, impression share, and other fragilista metrics.
PPC Hero: How does all of this play in with the current agency ecosystem?
Andrew: Today’s agency ecosystem is robust b/c it is not monolithic and rocks a lot of old-school tactics that work. By contrast, some of today’s behemoth darling firms like Facebook, Salesforce, LinkedIn etc. could 100% disappear. It’s all a matter of perspective, in a way. Local farmers’ markets are on the rise. Big Fake Food: fragile? Maybe. Folks make “logical” arguments for consolidation, and rail against business “siloes”. But the great big consolidated firms often suck, and fail. Think Adobe, HP, Aquantive. Maybe great relationships and motives to help clients don’t “scale” neatly. Fragmentation may be robust
PPC Hero: Given all of these external forces working on an account, how do you know when to act?
Andrew: I won’t “lecture birds on how to fly,” but if you have intimate knowledge of an account that you built carefully based on bidding to performance and knowing your client… you continue to tinker and seek growth opps while using several (hands on and automated) means of alerting you to performance issues then you should have a good idea of how to blend growth with caution at each stage in life cycle. Jim Collins cautions against hypergrowth delusions. Just grow every quarter, & it all adds up. The “20 mile march.”
PPC Hero: If it’s even possible to distill it down to one tweet – how does one deal with all of the randomness in PPC?
Andrew: It might be a simple as a solid understanding of the relevant statistics. Fight “narrative tendencies” in your mind.
PPC Hero: Last question: If people are now hooked on these concepts (like us), where can they go for further reading?
Andrew: Lots of places!Fooled by Randomness; The Black Swan; & Antifragile (Taleb). Thinking: Fast and Slow (Kahneman). The Opposable Mind (Martin); Great By Choice (Collins); Innumeracy (Paulos). Survival is Not Enough (Godin); Ignore Everybody (MacLeod) More importantly, expose yourself to diverse ideas! Art! Flesh and blood things. The unexpected. Broader philosophies that favor “experience over theory” tend to be conservative (George Grant, Edmund Burke) But I think that oversimplifies things. Many people of all political stripes fear the fragility of tyrannical “systems.” Take care. The ‘fragilistas’ at the top may risk your future, despite their being shielded from consequences.
Well, that’s all from us. Thanks for tuning in!