This is the subhead for the blog post
Bid management is probably the most difficult and least understood aspect of search engine marketing. Indeed, it is so complex that I am confident that I won’t be able to truly do it justice in one short article. So consider this article an introduction to the basics of bid management, not the be-all, end-all document. In a coming print issue of the magazine, I’ll go into more detail about some of the more advanced bid management techniques (rules-based bidding versus positional profit optimization, for example).
It All Begins with Tracking. Whether you are Walmart.com or selling out of your garage, your bid management is only as good as your tracking. There is simply no excuse not to have keyword-level tracking for all of your paid search campaigns – without it, you might as well consider your spend on the search engines a charitable donation to the Larry and Sergei private jet foundation.
The simplest way to quickly get keyword-level tracking is to install the search engines’ conversion pixels. These are small bits of code that have virtually no impact on your site’s load time and can be implemented by anyone with a basic knowledge of HTML. Install these tracking pixels and you’ll be able to see which keyword and often which ad text, search query, and referring site drove your conversions. You can then immediately start reducing or pausing keywords/ad text/referrers that aren’t bringing in the appropriate return.
Rules-Based Bidding: The Industry Standard. The easiest and most popular mode of bidding is “rules-based bidding.” Rules-based bidding means that you establish the value of a click to you, then bid a portion of that value based on a business objective. I realize that the preceding sentence makes it sound pretty complicated, but trust me it isn’t. There are three parts to rules-based bidding: A) determine your business objective; B) set minimum data thresholds; C) calculate the appropriate bids.
A. I have generally seen three business objectives occur over and over again in SEM campaigns:
1) Maximize revenue within a given budget. If you are start-up trying to grab market share, you may look at SEM as a means to grab as many customers as you can as quickly as possible, without regard to profit dollars. In such a case, your objective is to drive maximum revenue/leads/sign-ups/whatever within the constraints of your marketing budget.
2) Maximize profit. Most lead generation companies and many mature businesses are concerned with profit and profit alone. In such a case, a business would rather drive $5 of revenue with profit of $4 then drive $1,000,000 of revenue with profit of $1. Of course that’s a bit of an extreme example, but the point here is that a profit maximization strategy does not look at revenue or margin as a goal but rather as a means to drive the most profit.
3) Maximize revenue with a margin constraint. This is probably the most popular business objective. The goal here is to get as much revenue as you can, provided that you don’t go below the company’s margin objective. This enables a business to grow revenue/customers/market share while at the same time ensuring that this growth is not coming at a rate that will bankrupt the company. Basically it’s a hybrid of the aggressive growth approach of revenue maximization and the aggressive profit approach of profit maximization.
B. The next step in rules-based bidding is to determine your minimum data thresholds. The concept here is that you don’t want to bid adjust every keyword every day, simply because most of your keywords don’t have enough statistical data to make an informed bid. For example, if you have a keyword with one click and zero conversions, you wouldn’t want to reduce the bid to zero, nor would you want to make a dramatic increase in bid price on a keyword that received one click and one conversion.
I recommend a two-part approach to establishing minimum data thresholds – one based on time and the other based on your historical keyword performance. The time-based threshold basically suggests that you run data over multiple time periods – for example, you could run a 7 day report, a 14 day report, and a 28 day report. By doing this, you’ll quickly see that some keywords that didn’t get many clicks over a 7 day period are suddenly very significant over a 28 day period. You may also find the opposite to be true, where a sudden spike in clicks over a short-time period suggests that something has changed in the bid landscape that requires your attention.
For your historical keyword threshold, I recommend looking at three factors: clicks, cost, and conversions. Determine your historical conversion rate and revenue per conversion for your keywords, and then use this as a benchmark to determine when it is time to make a bid adjustment. For example, let’s say that you expect a 2% conversion rate and each conversion is worth $50 of revenue to you. Using this data, a simply threshold rule of thumb would be to say that you are going to bid-adjust any keyword that has gotten 30% more clicks than your average number of clicks required for a conversion (2% would mean 50 clicks to a conversion, 30% more would mean 65 clicks) or has cost you 30% more than the average revenue without a conversion ($65 – 30% more than $50). To account for opportunity keywords, you also want to identify keywords that may not have reached your minimum click or cost threshold, but have already gotten a certain number of conversions.
C. Now that you’ve got your business objectives and data thresholds established, it’s finally time to begin bidding (you thought it would never come, didn’t you?). A rules-based approach is defined by the following formula: Bid = RPC(1-MG) where RPC means “revenue per click” and MG means “margin goal.” Let’s put this into action. Say that you have a keyword that has received 100 clicks and $25 of revenue. Your goal is achieve a 20% margin. First, calculate your revenue per click. In this case $25/100 = $.25 RPC. Our margin goal is 20% which we express in the calculation as .2. Thus Bid = $.25(1-.2) or .25X.8 or a bid of $.20. Note that revenue per click doesn’t have to just be revenue, it can be “cost per conversion” or “cost per lead” or whatever your business metric is.
Combine Filtering with Your Bidding. If you’ve installed a conversion tracker, or better yet installed an analytics package (like Google Analytics, Omniture, or CoreMetrics), you can combine your rules-based system with additional data about your users to get even better results. I’ll be writing a separate column on filtering, so I am not going into a lot of detail on it here, but suffice to say, understanding day-parting, geo-targeting, match-type, and referrer conversion rates can have a huge impact on your bidding strategy.
Just to show this in practice, I’ve found that for some of my clients, conversions spike substantially between 7 and 9 AM and then around lunch time. Between 9am and noon, however, conversions plummet. Why is this? I suspect that people browse for products before work, then get into the office and work hard for a few hours, then go back online and start shopping during their lunch break. By increasing bids during peak shopping hours and reducing them when people aren’t serious about buying, I can achieve huge profit increases. The same is true for understanding all other elements of your user behavior.
Cluster. If you’ve built out a list of thousands of keywords, the odds are that there are many keywords in your mix that get a few clicks here and there but never get enough clicks to make it on to your 7 day, 14 day, or 28 day reports. This can often be a real problem for you, since these little keywords can collectively cost you a lot of money and margin. Consider, for example, if you had 1000 keywords that each cost you $5 a month and never converted – on an annual basis you’d end up spending $60,000 on these seemingly harmless keywords!
A good solution to stop these losses (on keywords I call ‘slow bleeders’) is to cluster groups of similar keywords together and bid en masse. If you’ve organized your Ad Groups into tightly knit groups of related keywords, the easiest way to do this is to simply create an Ad Group level default bid. For example, if you have five keywords in the Ad Group that have met your minimum data thresholds, but 200 that have not, you bid individually for the five keywords, but then aggregate click, cost, and revenue data for the remaining 200 and bid these together. If you want to get even more sophisticated, you can try to cluster keywords based on semantic similarity (words that are like each other) or behavioral similarity (users interact with the keywords in the same manner), but frankly I think creating very basic clustering rules is the best strategy to save your sanity.
Keywords are Zeros and Ones. It’s easy to fall in love with keywords. I know I sound like an uber-nerd just for saying that, but it’s hard not to get excited when you find that ‘secret’ keyword that your competitors haven’t discovered, or to just see a keyword drive in revenue day after day.
At the end of the day, however, to perfectly execute your bid strategy, you must remain impartial to all keywords. Although it sometimes hurts to bid reduce, pause or even delete one of your favorite keywords, numbers speak louder than your keyword affection. I tell my clients that I consider all keywords to basically be binary code – zeros and ones. If the keyword “coffee cake” sells diamond rings, I’m going to buy it, and if the keyword “diamond rings” doesn’t sell diamond rings, we’re going to cut it. Sometimes keyword bidding seems illogical – it can be hard to believe which keywords work and which ones don’t – but you have to listen to the numbers and not to your heart!