|Bidding for maximum profit|
I've been struggling with this for a while. An ROI system tells you how much a click is worth, but it doesn't tell you at what number of clicks your profits are maximized.
0 clicks = 0 profit. x clicks = 0 profit. Clicks greater than 0 and less than x = profit greater than 0. But at what number of clicks (and thereby CPC) is profit maximized?
If CPC increases in a linear fashion along with the number of clicks, I think profit would be maximized at x/2 according to the example above. The graph of profit vs. clicks would slope up to the maximum profit and then slope back down to zero with the same slope on either side of the point of maximum profit.
If CPC increases exponentially along with clicks, I think profit is maximized at a number of clicks greater than x/2. If CPC increases logarithmically along with clicks, I think profit is maximized at a number of clicks less than x/2.
I looked at some macro data in my account and it looks like CPC doubles when the number of clicks quadruples. At this point my lack of mathematical ability really takes its toll and I'm pretty much stumped.
Am I on the right track? Can I use a method like this to figure out what my maximum CPC should be in order to maximize profit?
"Our Chief Economist, Hal Varian, explains the how to adjust your bids to maximize the profit from your marketing investment on AdWords."
Google AdWords Bidding Tutorial:
GREAT video, thank you. So in order to go about this in an automated way, I should have my system bid small on each keyword, then increase the bid after receiving a certain number of clicks, then calculate the Incremental Cost per Click (ICC) after receiving a certain number of clicks under the new bid, if the ICC is less than the value of a click, increase the bid and repeat, if the ICC is less than the value of a click, decrease the bid and repeat?
I wonder if this would work better on a larger scale. Like if I do all of this on an aggregate basis for the entire campaign?
BTW, any other video of Google's (or somebody else's I suppose) you would particularly recommend?
Yes, you've got it. Do keep in mind that the curve's apex is rounded, and where you are on that curve is constantly in flux. So don't get lost on debottlenecking your optimum profit point on that curve, it's a moving target - get in the sweet part, and work on the PPC foundational aspects, it's rare for your offset on that curve to be your point of largest returns for your work you'll do.
|Do keep in mind that the curve's apex is rounded, and where you are on that curve is constantly in flux. So don't get lost on debottlenecking your optimum profit point on that curve, it's a moving target - get in the sweet part, and work on the PPC foundational aspects, it's rare for your offset on that curve to be your point of largest returns for your work you'll do. |
Since it's a moving target like you said, I plan to have my automated system test this indefinitely. Once the apex has been approximated, I think bids would usually move up and then back down after each test period on either side of maximum profit. If the apex moves significantly, the system will follow it and move bids up continuously or down continuously until the apex is found again and then go back to going up and down around the apex again. All automated of course. How does that sound?
What do you think of using aggregate data from my entire campaign for this instead of trying to get really granular?
Focusing on the cruise control is most often misleading, work on the engine, wheels, exhaust, suspension.
Meaning, developing tools to automate the Profit / ROAS curve leads you to focus too much on a flat curve, you want to find steep curves to work on.