On the surface, PPC keyword bidding looks simple and one-dimensional. Keep updating individual keyword bids until each is hitting your key performance indicator (KPI) target, and you should be fine. That’s a start. But in reality, doing that is harder than it looks. And doing this alone may leave your account well short of full optimization. Instead of thinking in one dimension, it’s better to think of the bidding puzzle more like a Rubik’s Cube. Learn to wrestle with it in (at least) three dimensions.
Remember, keywords themselves simply represent a set of user queries you expect to show ads against near Google Search results. The data we attribute to those keywords gives us clues as to the value of those keywords. But keywords aren’t profits; your job (in part) is to bid them predictively in such a way that you prioritize showing ads to searchers who have the greatest propensity to buy, highest typical order size, etc. Although some keywords may be highly correlated with profit, to quote the streetwise trader Fat Tony from Nassim Taleb’s Antifragile: Things That Gain From Disorder, “They are not the same ting.” (The trader noticed that the market had been lulled into equating the potential outbreak of war in Iraq with the price of oil, but he noted that they were merely predicting that the price of oil would further rise with the advent of war, but also believed that this prediction was likely to be quite wrong.)
Here are some typical bidding errors:
1. Overbidding at the top end.
Rule-based bidding (or hand-made versions of same) to KPI targets absolutely should include parameters related to ad position and/or impression share. A bid increase when a keyword is already averaging ad position 1.1, or when impression share metrics are already sky-high, is simply waste.
2. Mistaking ad position stats for delivery.
Keywords with mediocre quality scores, broader match types, etc., may appear in high ad positions, yet have plenty of room for more impression share (if you want that). There is often room to bid up in today’s auction because Quality Score today often creates a smooth curve of increasing exposure with increased bids. Ad Rank is computed on the fly for every single user, so, depending on the characteristics of a given user and query, your ad may show in all sorts of different positions to different users. Google also states that Quality Score not only determines ad position on the page, but eligibility. You can’t expect ad position stats alone to tell you whether you’re bidding high enough. Beyond checking impression shares or the Bid Simulator, consider setting up AdWords Campaign Experiments (ACE) to rigorously test the volume and return on investment (ROI) impact of higher (or lower) bids.
3. Mistaking keywords for queries.
A keyword’s numbers might look pretty good, until you check the Search Query Report and learn that brand queries, or broader or more specific queries leaching over from another part of the account, are being tied to this keyword for delivery (and thus statistical reporting) purposes. For more predictability in bidding to queries, you’ll need strong campaign structure, more exact match types, and best-practice use of negative keywords.
4. Yo-yo bidding.
Numbers change. That’s the nature of large datasets in dynamic behavioral environments. But if the ratios among your keywords are reasonably accurate, then chances are – if you take it upon yourself to make thousands of changes every week based on what could be random fluctuations and normal variations in the data – you’ll be no farther ahead than if you made virtually none of those changes. As a long-term strategy, think about the profound wisdom of ensuring that all of your enduring ratios are more or less accurate, including the ratios of one keyword match type’s bid to another, high ROI keywords as opposed to weaker ones, etc. Yes, you’ll always need to make ongoing adjustments, but how much is too much? If the account is at least a couple of years old, chances are you’re doing too much of this Sisyphean “chasing” of small bid adjustments, and diverting attention from other, equally or more important, elements of account strategy and management. (Related to this: although the sentiment isn’t without a grain of truth, we tend to pigeonhole keywords as being “nice” or “naughty” based on past performance data. With a slightly broader focus than the keyword – say, the ad group – we might manage and communicate differently about the effectiveness of the strategy as a whole, while still being open to making needed adjustments to discrete keywords within groups.)
5. Not including other predictive factors in bidding.
Based on past events and little else, we assign keywords a bid range and then adjust when we see more data come in. That’s not perfect wisdom, it’s just trying to get better and better at predicting the value of what will happen next in the searcher behavior universe. But don’t forget to inject other influential variables into your “predictions” (i.e. your bid on any given user query). Under Enhanced Campaigns, you can assign plus or minus bid percentages by time of day and day of week, as well as geography (country, region, or metro area). If you’re not taking advantage of these factors, your bids could be off by 30 percent or more at any given time and place.
There is a lot more to bidding, keywords, queries, Quality Score, and overall prioritization of effort in a PPC account than initially meets the eye. Thinking in multiple dimensions is a must.