The AdWords advertising system utilizes an auction-like process to decide which advertisers’ ads get to show and in which order. By bidding higher or lower, you can appear at higher or lower positions — depending on what your competitors are bidding.
Account managers use many approaches for setting and tweaking bids over time, and bid optimization is one of the oldest features in the SEM software industry. Automated bidding can work great, but many account managers (particularly agencies and consultants) have neither the budget nor the inclination to pay for an additional system to help them manage their AdWords account, so often the account managers must set bids themselves, which can be a daunting task.
I recently audited a large enterprise account and saw that the account manager had been very busy making bid changes, every three days or so, for the same keyword in many cases ($2.00 –> $12.00 –> $8.00 –> $15.00). Frankly, it looked pretty crazy. This got me thinking about one of my pet peeves: the push that account managers feel to take action, any action. After all, they’re supposed to be managing the account, right?
Certainly, if an account is making money hand over fist, one of the best things you can do is… don’t touch anything! (It may seem crazy, but you’re getting paid the big bucks to make the big decisions — and doing nothing is a pretty big decision.)
I would liken this to the debate over “always be testing.” Half of your tests will do worse and degrade the performance of your account during the testing process; a client once told me, “I don’t want to do any testing this next quarter because the campaigns are performing well, and we are under particular pressure to hit our numbers right now.” This was an extremely wise business decision on her part.
How different people approach manual bidding
I’ve seen various takes on how to bid over the years. Some like the approach of going in daily, weekly or monthly and comparing performance over time to see if there’s been a change. Others simply like to examine positions, and for keywords that are showing too low, bid them up. The more sophisticated will actually examine cost per lead or Return on Ad Spend (ROAS), and adjust bids based on that — keywords with CPAs higher than the account average are bid down, and keywords with CPAs lower than the account average are bid up.
I often do a combination of all three approaches myself; keeping an eye on position, for instance, is pretty important in order to ensure volume these days, with only one to four ads showing at the top of the page.
The problem: Sparse data
If you have a keyword with 200 impressions, eight clicks and one conversion that cost you an average of $3.00/click when the account CPA is $30.00, what does that tell you about whether you’re bidding too high or too low for that keyword?
Answer: absolutely nothing. Don’t even bother calculating what CPA you just achieved; it’s not even worth your time. Basically, you have NO IDEA what is up with that keyword. (Sorry, a single conversion is not enough to tell you anything!)
From time to time, there has been talk in the industry of “real-time bidding,” where bids are automatically and constantly adjusted. This concept is, frankly, ridiculous to me; changing bids based on super-sparse data is silly. Think about it. You get a click… should you change bids now? Obviously not. One click on your ad doesn’t mean anything. A bunch of clicks, though… well, now you’re talking!
The other problem: Forgetting what the heck you did already
People who religiously adjust their bidding on a weekly basis are particularly susceptible to this. Since you’re disturbing the bids, you’re mixing data from different bid levels together. So if you examine performance over the last month, you have maybe three weeks at one bid and one week at another bid. Did you write down when and what bid change you made? Right, me neither — it’s too much of a hassle.
So often, what you’ll see is someone making bid changes, thinking they have enough data when they are really just muddying the water and making decisions based on blended costs per acquisition from two different periods without realizing it.
On top of it, there’s a tendency to jerk bids around. You change a keyword’s bid one week, then you forget and change the same keyword’s bid the next week… in the opposite direction. Or you may be making the same decision twice — bidding up, then bidding up again the next week before there’s been enough time for data to accumulate at the new level.
Two solutions: Space and time
You can get more data in two ways.
You can associate that keyword with other keywords in the same keyword space — i.e., you can analyze at the ad group level, or the campaign level, or associate the keyword with others in other ways, in order to aggregate and get more data.
Or you can simply… wait.
Account managers hate waiting; they like to tell their boss they are in there tweaking the account, making changes left and right, and so forth. It feels bad to not make bid changes. What if someone looks at the account history and sees you haven’t made any bid changes in months? Won’t it look bad?
But guess what: If you don’t have enough data for a rational decision, you shouldn’t be making changes!
How often should you be bidding?
Prepare to hear some sacrilege! If you can hold off and try to do bidding across the entire account once a month or once a quarter (or even less frequently, if possible), and you make sure you only examine data since the last time you changed bids, you’re going to have much more solid data and will make better decisions.
How often you adjust your bids depends on how many clicks and conversions you’re getting, of course. But here are my general guidelines:
- E-commerce B2C
- Low average selling price (ASP) products with a high conversion volume: weekly or monthly at most may be best.
- High ASP/low-volume products: monthly or even quarterly.
- B2B: These markets are much more stable, and quarterly is usually just fine.
A caveat is that you may find yourself in a very dynamic market requiring much more frequent bid changes (e.g., the short Medicare signup season, where bids sometimes have to be changed daily), or you may have a single competitor come in and disturb the equilibrium of certain keywords and throw everything off. For this reason, you should monitor keyword positions, and if you see a big change on a high-volume keyword, don’t be afraid to take action.
In addition to getting more data and a more reliable history to work with, there are other benefits to waiting. Instead of spending 10 minutes here and there tweaking bids, if you aggregate all that time together and really do a focused, in-depth analysis based on CPA or ROAS across all the keywords, you’re going to do a much better job of making changes to bids across the account.
Coping mechanisms while you wait
Certainly, you should be analyzing bids on a more frequent basis than monthly. Tweaking bids for the top 10 or even 20 keywords if you really can’t resist the urge (try hard!) is okay, provided there are enough clicks and conversions to justify making a change. Adding negatives and positives, and setting up and analyzing ad or landing page tests, are things you should always have going on with any account.
Additionally, you can spend your time implementing more advanced optimization techniques like dayparting, or mobile or geographic bid modifiers; go ahead and do the account some additional good while you wait. That way, you can improve performance in other ways and have plenty of “evidence of industry” for your boss or client, besides just lots of bid changes.
Water those tomatoes, but don’t flood them
Really good stewards tend to be their own worst critics. The next time you feel pressure (usually just from yourself) and are tempted to unnecessarily tweak some bids, just take a step back and realize your account may do better if you simply have patience.
Instead, put your time into analyzing your data in other ways and strategizing while you wait just a little while longer; the results of good decisions based on sufficient data will be worth the wait.
Opinions expressed in this article are those of the guest author and not necessarily Search Engine Land. Staff authors are listed here.