It’s not easy to find the bid level that is most profitable for a given keyword. It depends on:
a) your income per conversion
b)your conversion rate
c)the impact of different positions on your click and conversion rates (i.e. lower CTR in lower positions, etc.).
d)what your competition is bidding
A) and B) are easy to find out. C) is difficult but just possible.
D) is impossible in adwords (though not in Overture).
Therefore in Adwords one cannot work out exactly which bid level is most profitable. (and in Overture you can only do so by looking up the competition for every single bid one at a time). However, there are some hard rules that can be calculated nonetheless that give a lower and upper ceiling for this elusive “most profitable bid”.
I’ve managed to put together a spreadsheet that can help a lot with this. You enter your income per conversion and your conversion rate, and it then gives you guidance about how profitable different bid levels are. You can get it here:
[awakenedheart.org ]
The most profitable situation is always to be in position 1 paying the minimum bid; the profitability of other bid/position situations can be expressed as a percentage of the profit per click you would make in that ideal situation. The spreadsheet tells you the profitability of each bid/position combination.
The "lower ceiling" comes from the fact that if it is more profitable to be in position A paying X, than it is to be 1 position below A paying the minimum bid, then you should bid X. In other words, there tends to be a bid level such that gaining 1 rank of position is worth the extra spend (above the minimum bid) because of the extra clicks/conversions it brings. Cells that show a situation like this are colored blue in my spreadsheet. The "lower ceiling" is the column furthest to the right (highest bid) that is mostly blue. You should always bid at least at the lower ceiling; to bid any less is to pass up on some profit.
The "upper ceiling" comes from the fact that if it is less profitable to be in position A paying X, than it is to be 1 position below A paying 1 cent less than X, then you should NOT bid X. In other words, there tends to be a bid level such that gaining 1 rank of position is NOT worth even 1 cent. At some point the extra clicks just cost too much. Cells that show a situation like this are colored orange in my spreadsheet. The "upper ceiling" is the column furthest to the left (lowest bid) that is mostly orange. You should never bid more than the upper ceiling; to bid any more is to lose profit.
The area between the 2 ceilings is one where it is unavoidably uncertain whether the higher bids are justified or not. It all depends on what the competition is bidding.
A note on the impact of position on impressions, clicks and conversions. This is a difficult area to research: you need very large volumes of data. If you just look at your own data in one sector, you’re likely to have your results distorted by arbitrary differences between your keywords in the number of competitors. (e.g. if you are #1 on your best CTR terms, and #3 on the rest, you might think that being #1 was really good for CTR; but actually all you are seeing is that you chose to bid highly (or had less competition) on those terms that had the best fit with your offering.
I used the published research of the “Atlas Rank Report” which had huge data samples from many different advertisers. Reading their reports will help you understand what I’m doing in the spreadsheet. See their reports here:
[atlasdmt.com ]
[atlasdmt.com ]
I adjusted it in 2 ways though. Firstly, I left out their results on differences in conversion rate between positions. The findings were small; in opposite directions depending on how popular the keyword was; and my offering is unusual so I didn’t trust the relationship to be the same as for anyone else. You can very easily factor this data in though, just by changing that conversion potential column.
Secondly, I didn’t trust their assertion that for Adwords position #2 got only 70% of the impressions that position #1 gets. So I also give in the spreadsheet some alternative datasets (Adwords modified to remove this 1 phenomenon, and Overture) to give some additional perspectives. Generally, it doesn’t seem to matter much.
Your logic is good at the keyword level. The problem with your approach, though, is that it doesn't take into account the fact that your keywords collectively constitue an keyword portfolio whose overall returns are more important than the returns on any one keyword. If you optimize keywords individually you'll be doing what all of your 'optimized' competitors are doing.
Optimizing the collective keyword portfolio involves building a mathematical model that
a)encapsulates the historical performance of the overall keyword set
b)incorporates near real-time cost and revenue data to update its model
c)continually examines how best to allocate spend *across* the keyword set to best meet the portfolio-level ROAS goal
d)clusters sparse data from individual keywords to make statistically meaningful inferences about how to bid those low-volume keywords
With regards to Google Adwords, you're right in stating that a rules-based approach like the one built into your spreadsheet won't work on Google because of its CTR*maxCPC model. With Google typically accounting for 60-70% of a ppc campaign, being able to accurately model and optimize AdWords is a challenge that very few companies have worked through.
I would like to hear more opinions (especially from people used to these figures), and tour model could be improved.
This is useful even for beginners; they can see what is max CPC they can afford.
Well done and waiting for the "PROs" to give their opinion.
Impressive spreadsheet. I'm still digesting it but my 1st feedback is that your best bid amount applied to one of my oldest and most optimized campaigns is the average CPC I have now. This of course can be a coincidence :)
For Income per conversion I used my "maximal amount I'm willing to pay for a conversion". Is that the number you mean?
Keep up the good work,
Rob
"Income per conversion" means the profit you make on a conversion, but NOT taking into account click charges.
I.e. it's the amount you sell your widget for, minus the cost of the widget.
A lot of my observations have matched what Atlas found.
For each position you move up, you increase your conversions by about 20%. For instance, 24% at position 7 is 20% better than the 20% at position 8. From position 2 to position 1 is a larger increase.
After position 8-10, things drop way off.
What I ended up doing was just bidding to a certain ROI. I know the profit I make per conversion and my conversion ratio for each keyword, so that gives me my earnings per keyword. From that I can calculate the bid needed to return a specified ROI. For instance, if I earn $2 per conversion and get a 25% conversion, that's $0.50 that I earn per click. If I want a 150% ROI, I bid $0.20.
I'm very interested in modifying my algorithms to maximize my profit, but I just can't find a reliable way to determine what position I would get for various bids.
There are some keywords where I have tons of data and can see what bid will get me what position. I noticed an interesting thing with many of those. The total net profit increased slightly as I increased the bids (up to a certain point, after which it declined rapidly), but in many cases it took several times as high of a bid to get a very marginal increase in profit. For instance, with a $0.20 bid I might get 100 clicks ($20 spend) and $40 in profit. Bumping it up to $0.50 would get me 500 clicks ($250 spend), but the profit would only increase to $50. The diminishing returns aren't worth the larger capital outlay to me.
Hmm, I think I might focus on this from a different angle. Rather than try to figure out the position that I'll get for each bid, I think I might change my target ROI each week (50%, 100%, 150%, 200%, 250%, 300%) for about six weeks and see which ROI produces the highest total profit.
I intuitively feel it has worked for me, but proving that mathematically is a bit challenging.
one to thing to always remember though, at the end of the day, ROI doesn't pay for the groceries. It's total profit.
Hmm, I think I might focus on this from a different angle. Rather than try to figure out the position that I'll get for each bid, I think I might change my target ROI each week (50%, 100%, 150%, 200%, 250%, 300%) for about six weeks and see which ROI produces the highest total profit.
I'm still testing, but it looks like my "sweet spot" is about 150-200% ROI. When I decreased my bids so that my ROI was higher than that, my clicks went down enough that my total profit decreased. When I increased my bids so that my ROI was lower than that, my costs went up enough that my total profit decreased.
