Can You Hit Google’s Knuckleball?

While Google does throw many “knuckleballs” at marketers, there are several “fastballs” in there, too, if you know where to look.

Another year, another season of high hopes for my beloved Toronto Blue Jays. Our former Cy Young-winning knuckleballer, R.A. Dickey, offers hope and heartache. A strong seven-inning performance will be followed up by an agonizing start when opponents solve his mystifying pitch, sending numerous offerings flying readily over the center-field fence.

Still, at the end of the day, most of the agony and heartache resides at the receiving end of the knuckleball. Major-league hitters know how to handle a straight pitch, no matter how hard you throw it. They just adjust their timing accordingly. The knuckleball, on the other hand, offers them more of the unknown – or should we call it randomness, or opacity, or even… {not provided}? One pitch hangs up high; the next dives down and toward the batter. The following, seemingly similar, knuckler does the dive thing, but down and out of the strike zone.

This is why advertising – even in its digital form – is so difficult to master. Whether it’s about attribution (did my spend really influence a purchase?), a weak grasp on whether an increased budget leads to diminishing return on investment (ROI) (through cannibalization of effort, for example), or opacity in mapping the likely impact of a higher bid (and thus average CPC) on ad position or impression share, you run the risk of getting “called out looking” by failing to pursue opportunities where they exist. Or – more to Google’s taste – you may be induced to swing for the fences without really knowing what you’re swinging at; your ad campaign sees too many embarrassing whiffs and the occasional “mile-high pop-up”… a mighty swing whose energies don’t translate into victory.

Where Are the Fastballs? Everywhere, If You Look

Despite our often cynical observations, we seasoned advertisers do need to acknowledge that the reasons Google has felt urges to scoop extra profits at the expense of newbies stem from the fact that over the years, they’ve built so darn many “fastballs” into the system. If you’re any good, these are pitches you can hit.

  • Exact match keywords. If you know your typical return on a certain keyword in a certain ad group, your bid strategy and ad copy are dealing with a more stable, predictable, “known.”
  • Use the bid simulator. With this tool, Google has attempted to lessen the uncertainty implied by the question: “If I bid higher, what impact is that likely to have on the number of impressions and clicks I’ll receive?”
  • Since AdWords was invented, you have always been able to compare the performance of your ad creative versions, reasonably tightly mapped to a set of keywords that people are searching on. In the history of advertising, that’s like taking a split-test in a direct mail campaign and setting it at warp speed.
  • And the list goes on: full, robust analytics (including multi-click funnel information), call tracking solutions, etc. Whether it’s from Google, Bing, or a third party, there is plenty out there to help you swing hard at your known producers, aka fastballs.

Beware These Knucklers

  • Broader match types. Unlike exact match, the broad match type in particular might allow you to view performance at an aggregate level and bid accordingly, but it has significant opaqueness and volatility attached to it. That’s an advantage in a way, as long as you set expectations accordingly. You’ll swing and miss a lot, as the system puts your offer in front of a wide range of queries. You need to be mindful of this imperfect but broad targeting methodology, especially insofar as it may be cannibalizing other efforts. And would a more granular approach using more exact match types allow you to do even better?
  • Product Listing Ads (PLAs). Knowing where to set your bid (see “Bid Simulator”) above is vital to the art of swinging hard at fastballs. If you are afraid of needlessly bidding too high, then you may remain cautious. What if a higher bid won’t get you significantly more impression share? The first version of PLAs had a lot of us swinging too hard at this big, opaque knuckleball, or conversely, standing around being called on strikes because we were so cautious that we were below the auction. That has improved a bit in that impression shares and the Bid Simulator are now available in Shopping PLAs, but only for Product Groups and/or on sufficient volume. Result: We’re forced to bid too high or too low until enough tinkering finds us the right level. Google’s throwing us that knuckleball, so chances are, they’re profiting from it.
  • Segments that we attempt to bid on may not be particularly valuable information in themselves. If pure propensity to buy was a segment, we’d bid on that. Instead, we attempt to glean the same from demographic information and geo segments. Being from Arizona doesn’t cause someone to be a lower converter from a click on your ad, or to have a low average order size. Some people who live there have a high propensity to buy. But a lot of people may be too old for your product, or not have a household budget for the product, or some other reason. You can bid to that factor, but the effort feels like swinging for a knuckleball. Sometimes, you discover that an urban vs. rural split within a state is stronger. You get a bit closer to being able to make headway in terms of accurately bidding up and/or down. You get a nice change-up to hit, if not a fastball.
  • Small batches of data can be the biggest knuckleball of all. For some reason, even on a product with significant volume, we tend to individualize the analysis by keyword. If one of the keywords in the ad group suddenly sells $250 worth of goods, we get overexcited by that one keyword. The reality is, a great deal of randomness is a basic fact of life in terms of where we wind up attributing someone’s purchase. If that keyword didn’t get the credit, a closely related one might have. And who is to say, in an ad group that sells 40 products a year, how important or meaningful it is that we sold seven products this month? In any random distribution, events can tend to bunch up. So we tend to overrate recency in our calculations as well, swinging hard at the plate in a knuckleball environment. In our industry, we have to develop a more sophisticated understanding of what it is to be appropriately predictive, as opposed to just chasing after events that have recently occurred, but may not be likely to occur again in the exact same way for some time to come.

Keep swinging! Hope you find more fastballs than knucklers next time out.

Originally posted on ClickZ –  Can You Hit Google’s Knuckleball?