Another bundle of advertising flyers just landed on my front porch, only to go straight into the recycle bin. Although I merely dream of the distant day when such lazy and wasteful advertising methods will be punishable by death or a day spent shopping with Paula Abdul, I hold out some hope that search marketers will soon learn to clean up their acts, guided by less draconian incentives.
The easy availability of keyword generation tools, pumping out cookie-cutter keyword lists, has led many marketers to emphasize keyword list length over smartly constructed accounts. Wading through endless, ambiguously-performing long tail keywords (Yes! they might be cheaper! Or not! On the rare occasions when they convert…) shifts focus away from keywords that really matter, crisp and accurate bidding decisions, ad copy performance assessment, content performance assessment, and more.
Keyword intent misjudgment shouldn’t be considered a rookie mistake. It happens to the best of us. It takes years of experience to distinguish “dumbass keyword choices” from high-intent keywords.
Why not dump everything in and just test? For a couple of reasons. First, it’s unwieldy. A single assessment of long tail data for conversion patterns in a large account could take you the better part of a day. If your day is worth, say, $750, I hope your boss or client (or yourself) is factoring this into the P&L scenario. The second reason is that lazily constructed accounts don’t garner high quality scores. Keeping low-CTR and low-intent keywords in high ad positions for too long can hurt your entire account, and could result in high minimum bids creeping in.
A decent way for an intermediate user to address the “dumbass keyword problem” is to go into the game with some presuppositions about user scenarios, and then confirm them with a reasonable amount of data. There’s nothing wrong with getting creative and adding a number of offbeat keyword suggestions into a group, just so you can definitively rule out the bad ones. But go easy. Don’t fill the account with loosely-relevant keyphrases. And once you confirm for yourself that some of them don’t pan out (for reasons you should have already been thinking about), don’t keep blithely loading them in every time a keyword tool suggests them for other parts of the account.
[Note that this is where continuity in campaign management is a real plus. Delegating the account to a junior person here, a variety of agencies there, and generally bouncing around in your strategy often leads to reinventing the wheel and remaking the same mistakes. If you’re working in a team: take notes. Don’t be afraid to tell your colleagues clearly what doesn’t work. Step on toes if needed.]
To work with an example, let’s look at the potential difference between the phrases
“wild bill’s waterpark franchise”
“wild bill’s waterpark location”.
A keyword research exercise will certainly show you both of these, and many other phrases besides.
The first query sometimes leads to a valuable business-to-business lead generation event for this client. The second almost never converts. All this is going to do – and you can, of course, confirm this through observation – is attract a steady stream of consumers looking for a map or directions.
To a newbie charged with the task of using keyword tools to find “all the relevant keywords,” it’s a source of pride to dump in a passel of phrases much like that ill-advised “wild bill’s waterpark location” above. A seasoned pro keeps such longshot phrases on a shorter leash, testing them only sparingly. The old incentives (hey, it’s free to try) have been phased out. In a “Quality Score World,” hit-and-miss targeting tactics don’t perform as well.
The brute force method (lots of keywords, post hoc interpretation) is a favorite of low-skill players and eager rookies. In hockey, we call that “dump-and-chase.” I’ve always been a fan of the “nifty pass, pass, shoot, score,” methodology. Let’s hope that the “dump-and-chase” analyst is actually skilled at interpreting subsequent patterns, and that this analysis can make up for the lost time and money that resulted from blithely dumping low-intent words into the account at the start.
Lest you think that’s a problem easily solved with automated bidding-to-ROI bid management, consider that in a highly granular account with upwards of 500 ad groups, a bid management system may be working with insufficient data for any given keyword for a long time. On top of that, going with bid-to-ROI technology in a highly granular account is a very blunt way of keeping your ROI in line for whole campaigns. But to keep you in the right range, the system may be acting overcautiously in many bid situations, and overzealously in many others… because it just doesn’t “know.” Autobidding works where there are high volumes and easily-recognizable patterns. Many accounts don’t look like that. Your own interpolations can do a better job than an autobidder. The autobidder can get your ROI numbers into range, but it doesn’t care all that much about your total profits.
So, you need to be manually taking your bids on high-intent words to relative extremes to get those ads in top ad positions. And you need to be lowering your bids on low-intent words relatively extremely. In short, unless your bidding system is more sophisticated than any currently on the market (with the possible exception of Google’s intent data, which they use in non-transparent ways to set your initial quality scores), for many individual business situations, it’s your brain that is going to have to understand keyword intent and filter out statistical noise to the best of your abilities and convictions. And getting it closer to right from the start will save a lot of time and money.
Although fun to use for their entertainment value, “commercial intent predictor” tools like the one available in Microsoft adLabs won’t help here. The phrase “wild bill’s waterpark franchise” has commercial intent of one sort; the phrase “wild bill’s waterpark locations” has completely different commercial intent. Not least, the second version might be a cue to Google to show a map in the blended search results, possibly skewing or worsening your ad performance. If your campaign is B2B, you should turn down that “locations” bid, or just eliminate the keyword altogether. In this case, a broad match for the business name, bid correctly, would be a significantly better bet than a bid on a keyword with high-volume and completely wrong intent, because the broad match would at least give you a portfolio of exposure on the range of high-intent and low-intent queries.
A great way to build up a file of useful historical data is to look at what keyphrases actually converted for you in both your paid and organic search. On the paid side, you can build up a nice file of data by keeping some broad matched keywords turned on. You may soon come to learn that certain “buy words” are quite common, and need to be bid on even higher. You might also find that, beyond that, the patterns are difficult to read. In which case, the answer is to leave broader matches and broader keywords running at the appropriate bid, rather than to get too excited about any particular, obscure longer phrase that some keyword tool happened to dredge up for you. Make more detailed use of your analytics package to drill deeper into what types of phrases are leading to conversion events on both the paid and organic sides; and on broad matches in paid, look at what actual phrases converted (not just the fact that the broad match converted).
As a general account orientation, there seems to be a temptation of late to build very large accounts with the expectation that you can “always scale back later.” I find it’s truer than ever that it’s advisable to start out with a tighter ad group structure and replicate that through any similar ad groups in your plan. Build a strong account history and an account structure whose performance you can easily assess. You can always expand later.
Opinions expressed in the article are those of the guest author and not necessarily Search Engine Land.