Creating and maintaining a bidding strategy is one of the most complex areas of AdWords and there is no unifying theory. If your goal is to find the best way to bid your keywords you are reading the wrong article because bidding formulas are not better or worse, just different. There is no one solution to bidding but there are some shared concepts. You have to find the one that you believe in the underlying concepts. So let’s explore some of the factors.
Deciding on your target CPA (Cost per Action or Acquisition) is probably the first step in most formula development stories. People often have problems with this. When you talk about CPA most just say they want it to be lower. That is very short sighted. CPA is a balance between the cost of the advertising and the value of the business activity it creates. The higher the target CPA the higher the volume of the leads is likely to be. We have seen many cases where people have saved themselves out of business by only paying attention to the cost per lead with little or no consideration for the volume of leads.
Going the other way with this is also dangerous because if the CPA is too high you will lose money on every sale. Trust me, you will never make it up in volume! Oftentimes we are dealing with actions that are not the final outcome so the quality of what flows into the pipeline has to be considered in this process. The value difference in the response type is a factor as well, one client I am working with gets 90% closed sales from his form responses but less than 30% for phone calls.
Conversion Rate is the next factor in the formula since it connects the CPA to the click. For example let’s say we have a target CPA target of $50. If we have a conversion rate of 12% then we know that we need 8.33 clicks on average to get one conversion. This means that our target CPC (Cost Per Click) is $6 ($50/8.33 clicks = $6 CPC). The challenge here is that conversion rate is not a constant since it is impacted by ad position, work intent, competitive environment, and many other factors. You need enough data that the peaks and valleys average out to what is likely to recur.
What you bid is not what you pay and the auction discount is the factor that gets from the Target CPC to the actual bid. You can figure out this discount by taking your traffic for some period of time and extending the clicks by the Average CPC and the Maximum CPC (This is what Google calls your bid). Down total these two and divide them and you have your Auction Discount. For our example here let’s say that the Auction Discount is 20%. You bid $1 you pay .80.
So here is the simple formula
Target CPA/(100/Conversion Rate)*Auction Discount = Target Bid
AdWords is not this simple!!!
If a bidding formula were really this simple I would be unemployed and every business using AdWords would be making money hand over fist. The problem is this formula does not consider the variables and there are many of them. Target CPA, Conversion Rate, and Auction Discounts are all variables that change based many factors. Market demand, Competitive Landscape, Product Presentation and many other factors come into play here and they vary by the time of day, geography, and circumstances. Statisticians have long understood that the way to lie with numbers is to average them. That is what’s happening here. We have the average CPA, the average Conversion Rate, the average Auction Discount. So we have a lie multiplied by a lie multiplied by another lie and we wonder why it does not work perfectly.
The last variable in the formula is common sense applied by a person who understands the audience and the market. You can want to bid a certain level but this is a competitive world and your competitors may not let you get away with what you want. Sometimes you need to modify the formula because your gut tells you something different than the math.
For all its problems, developing a bidding formula can get you closer to maximum leverage. The challenge here is that low volume conversions act like random events and it takes a long time to get enough data to make good decisions. Most businesses lack the patience to let their data build but to some degree you just have to find the strength to do so. What we have found over the years is that keywords should be separated into dependable converters and all the others. The dependable converters should be managed by the formula at a word level but the others have to be treated as a group.
When you have a word with a single conversion and another with no conversions you have the same result. In low volume words the next conversion is just as likely to come from the zero as the one. As the volume of events (conversions) increases this statement becomes less and less true. Depending on the volume of conversions we generally start to treat a word different after is has 4 conversions because at that point it’s starting to build statistical validity.
Formulas are great things and they help you think through the impact of your bid on the process but formulas are not as smart as you are. Words and what they mean are far above the ability of current software development so you need to keep people in the formula to bring the common sense factor to the end of the calculation.