Hello!
The comments on this profile are unclaimed and thus are unverified.
Do they belong to you? Claim these comments.
SF Greg
Is this you? Claim Profile »
1 month ago
in More Thoughts on Revenue Allocation / Attribution on The ClickEquations Blog
In other words, Google Analytics resets cookies with each visit, so only if the current visit is from AdWords will a conversion be reported as an AdWords conversion. Even if someone visits every day for a week, each time via AdWords, if the eighth visit is by an organic click and that's when they conversion, Google Analytics reports that as a conversion for an organic visit, not AdWords.
But in the AdWords interface and reports, if a person visits seven times, and only the third visit was via AdWords, then if they visit an eighth time via an organic click, AdWords will still report that as a conversion, because the AdWords cookies last 30 days and are not updated.
So in this sense Google Analytics is showing last click data. AdWords is showing equal click data, giving attribution to the AdWords click no matter whether it was the first click, a middle click or the last click, so long as it happened within the last 30 days.
Or am I missing how it really works? Thanks!
But in the AdWords interface and reports, if a person visits seven times, and only the third visit was via AdWords, then if they visit an eighth time via an organic click, AdWords will still report that as a conversion, because the AdWords cookies last 30 days and are not updated.
So in this sense Google Analytics is showing last click data. AdWords is showing equal click data, giving attribution to the AdWords click no matter whether it was the first click, a middle click or the last click, so long as it happened within the last 30 days.
Or am I missing how it really works? Thanks!
1 reply
Craig Danuloff
Greg: In this second comment you're confusing 'last click' of the entire string of visits with 'last click' of the paid search visits. Both GA and Adwords are 'last click' within the set of ppc visits. Actually this raises a whole new set of questions I'll have to go investigate about how GA really handles allocation across different visit types - not entirely sure - but I know who to ask!
1 month ago
in More Thoughts on Revenue Allocation / Attribution on The ClickEquations Blog
For AdWords, the web analytics applications usually report last click data, but AdWords is reporting equal click, right?
2 replies
Greg Moore
In other words, Google Analytics resets cookies with each visit, so only if the current visit is from AdWords will a conversion be reported as an AdWords conversion. Even if someone visits every day for a week, each time via AdWords, if the eighth visit is by an organic click and that's when they conversion, Google Analytics reports that as a conversion for an organic visit, not AdWords.
But in the AdWords interface and reports, if a person visits seven times, and only the third visit was via AdWords, then if they visit an eighth time via an organic click, AdWords will still report that as a conversion, because the AdWords cookies last 30 days and are not updated.
So in this sense Google Analytics is showing last click data. AdWords is showing equal click data, giving attribution to the AdWords click no matter whether it was the first click, a middle click or the last click, so long as it happened within the last 30 days.
Or am I missing how it really works? Thanks!
But in the AdWords interface and reports, if a person visits seven times, and only the third visit was via AdWords, then if they visit an eighth time via an organic click, AdWords will still report that as a conversion, because the AdWords cookies last 30 days and are not updated.
So in this sense Google Analytics is showing last click data. AdWords is showing equal click data, giving attribution to the AdWords click no matter whether it was the first click, a middle click or the last click, so long as it happened within the last 30 days.
Or am I missing how it really works? Thanks!
Craig Danuloff
Adwords is last click. Most web analytics (including GA) is last click too. The difference between Adwords and GA is the date the revenue is assigned to - on Adwords it's the date of the click, in GA it's the date of the sale.
5 months ago
in The Crazy World of Revenue Attribution on The ClickEquations Blog
Like painting a house, my experience with these projects is that 90% of the job is prep work. After six months of getting a data set ship shape, relatively simple statistical analysis will produce great results.
As I believe Jeff Bezos once said, Amazon's greatest asset is it's database. As the provider of an application, your greatest asset may be having multiple databases, one for each customer, with all databases structured identically.
Once you do the hard work for one customer, the prize will be a database that has been restructured many times, with additional database tables of derived data. With data that's just right you can run the statistics and show the results.
The payoff will be that all customers will have the new database, so the usual need for a custom job and expensive prep work will be gone. Each customer will only need to let their account run and their database fill, and the built in statistics formula will present some wonderful attribution reports.
The results may be surprising. Rigorous analysis may reveal that if the first click goes to a specific ad, the person is 20% less likely to buy. Pausing that ad may boost sales by 5%.
I'm skeptical of CRM - Conference Room Marketing. Things don't need to make sense when presented to the group. Ideas that are clear and quickly accepted may actually hurt the business. Ideas that everyone rejects as nonsense and flat out wrong may actually work quite well.
I like This Old House because in this complex world it is relaxing and comforting to watch skilled people succeed in an environment where reasonably simple rules, like measure once and cut twice, will do the trick.
My job involves imponderables, complexities beyond my comprehension. When the spreadsheets go up for the senior execs, I sit quietly, waiting to talk with the consultant at lunch, where we will talk shop and reveal how little we really know, where we will talk about what the spreadsheets don't contain, where we will acknowledge that our different approaches are often mere personal preferences and not the solid conclusions the black and white numbers on the spreadsheets make them out to be.
Months of work often results in incremental improvements. While glancing at a few customer records occasionally yields insights and it's fun to offer up a guess, I am skeptical, since this approach often produces results that give new meaning to your final statement - there are no right answers. When the data is so complex, how can it be otherwise?
As I believe Jeff Bezos once said, Amazon's greatest asset is it's database. As the provider of an application, your greatest asset may be having multiple databases, one for each customer, with all databases structured identically.
Once you do the hard work for one customer, the prize will be a database that has been restructured many times, with additional database tables of derived data. With data that's just right you can run the statistics and show the results.
The payoff will be that all customers will have the new database, so the usual need for a custom job and expensive prep work will be gone. Each customer will only need to let their account run and their database fill, and the built in statistics formula will present some wonderful attribution reports.
The results may be surprising. Rigorous analysis may reveal that if the first click goes to a specific ad, the person is 20% less likely to buy. Pausing that ad may boost sales by 5%.
I'm skeptical of CRM - Conference Room Marketing. Things don't need to make sense when presented to the group. Ideas that are clear and quickly accepted may actually hurt the business. Ideas that everyone rejects as nonsense and flat out wrong may actually work quite well.
I like This Old House because in this complex world it is relaxing and comforting to watch skilled people succeed in an environment where reasonably simple rules, like measure once and cut twice, will do the trick.
My job involves imponderables, complexities beyond my comprehension. When the spreadsheets go up for the senior execs, I sit quietly, waiting to talk with the consultant at lunch, where we will talk shop and reveal how little we really know, where we will talk about what the spreadsheets don't contain, where we will acknowledge that our different approaches are often mere personal preferences and not the solid conclusions the black and white numbers on the spreadsheets make them out to be.
Months of work often results in incremental improvements. While glancing at a few customer records occasionally yields insights and it's fun to offer up a guess, I am skeptical, since this approach often produces results that give new meaning to your final statement - there are no right answers. When the data is so complex, how can it be otherwise?
1 reply
Craig Danuloff
Thanks Greg - Interesting thoughts well expressed. We can and will do the analysis, but I worry that there aren't universal truths even in deep analysis - in other words there are all kinds of visitors who behave all kinds of ways, so there is no right answer. The real 'right answer' is to be able to show multiple angles on what the attribution means, so that a keyword doesn't have one number - the revenue it gets credit for - but many numbers which taken together give the impression/weight of that keyword and they are all considered when a person or an algo decided what to do next about that keyword. Much left to do.
7 months ago
in Quality Score - The Preamble on The ClickEquations Blog
Do you think the fourth function of Quality Score is to make money for Google?
I've heard people puzzle about why going along with some aspect of Quality Score increases the quality of the ad or the users' experience. Yet it makes perfect sense the other way around. Quality Score is what Google thinks is a quality ad.
A "high quality" ad is one that Google can put up and make lots of money with, while at the same time users will be reasonably happy -- happy enough to keep clicking.
So CTR rules because there's only so much space on the page and ads that don't get clicked on much wreck Google's revenue stream. Then, if users have a bad experience they will stop trusting those ads and clicks will decrease over time, so Google has to enforce some minimum standards.
Yet the profit motive is so strong that Google will not put up an exact match ad if there's a broad match in another ad group and the ad in that group has a higher CTR, regardless of the fact that the exact match ad has more search terms in the ad text and the landing page has more of the user's search query terms too.
To an extent, Google is paying lip service to Quality Score and relevance, hiding the search queries that broad match, knowing that if people saw them they would flip. The opportunity for using a data driven approach to negative keyword discovery is not fully available using Google's native tools (104 other unique queries...), yet rather than offer advertisers an easy way to diagnose and optimize this aspect of their account, Google hides the ball, talks the talk about Quality Score, privately knows about the wildly inaccurate matches, yet somehow calculates that they'll take the money as long as people keep on clicking.
Savvy advertisers solve these problems, most pay too much, relevance suffers, long term success at Google suffers as greed tips the balance in favor of short term profits, and even the advertisers who are aware of and solve for these problems pay the price of increased resource use to deal with these problems.
And consultants don't have a motive to lobby Google, because it's these very complexities that allow them to optimize an account and have clients pay through the nose for their services.
When you say that a function of the Quality Score is to provide a preferred customer program to reward top performing advertisers, what does that actually mean? The most interesting aspect of the webinar you're having on Quality Score is that it will "cover the more philosophical issues of transparency and fairness." This and the pragmatic tips will be great.
I've heard people puzzle about why going along with some aspect of Quality Score increases the quality of the ad or the users' experience. Yet it makes perfect sense the other way around. Quality Score is what Google thinks is a quality ad.
A "high quality" ad is one that Google can put up and make lots of money with, while at the same time users will be reasonably happy -- happy enough to keep clicking.
So CTR rules because there's only so much space on the page and ads that don't get clicked on much wreck Google's revenue stream. Then, if users have a bad experience they will stop trusting those ads and clicks will decrease over time, so Google has to enforce some minimum standards.
Yet the profit motive is so strong that Google will not put up an exact match ad if there's a broad match in another ad group and the ad in that group has a higher CTR, regardless of the fact that the exact match ad has more search terms in the ad text and the landing page has more of the user's search query terms too.
To an extent, Google is paying lip service to Quality Score and relevance, hiding the search queries that broad match, knowing that if people saw them they would flip. The opportunity for using a data driven approach to negative keyword discovery is not fully available using Google's native tools (104 other unique queries...), yet rather than offer advertisers an easy way to diagnose and optimize this aspect of their account, Google hides the ball, talks the talk about Quality Score, privately knows about the wildly inaccurate matches, yet somehow calculates that they'll take the money as long as people keep on clicking.
Savvy advertisers solve these problems, most pay too much, relevance suffers, long term success at Google suffers as greed tips the balance in favor of short term profits, and even the advertisers who are aware of and solve for these problems pay the price of increased resource use to deal with these problems.
And consultants don't have a motive to lobby Google, because it's these very complexities that allow them to optimize an account and have clients pay through the nose for their services.
When you say that a function of the Quality Score is to provide a preferred customer program to reward top performing advertisers, what does that actually mean? The most interesting aspect of the webinar you're having on Quality Score is that it will "cover the more philosophical issues of transparency and fairness." This and the pragmatic tips will be great.
1 reply
Craig Danuloff
Thanks Greg - Lots of stuff to reply to in there:
- Certainly QS drives Google profit. That's the main point of all three really.
- Improvements in 'user experience' and 'Google profit' are generally correlated, which is why they use that to defend and justify everything, but the trouble is at the margin when they make decisions which drive profit at the expense of user experience. They do it all the time but not surprisingly don't point those out. Minimum first page bids are a good example - there should be a quality level (not score, but concept) where no bid is high enough to make the first page - but that doesn't happen.
- What I mean by Preferred Customer Program is that High CTR, both in terms of complete account history, and at the AdGroup level, both help successfull clients do better with new and less powerful keywords. They get a boost on that keyword that a new or less successful client would not.
- I too find the issues of tranparency and fairness interesting. I'll do another full blog post about it this week, and certainly some will be covered in the webinar.
- Certainly QS drives Google profit. That's the main point of all three really.
- Improvements in 'user experience' and 'Google profit' are generally correlated, which is why they use that to defend and justify everything, but the trouble is at the margin when they make decisions which drive profit at the expense of user experience. They do it all the time but not surprisingly don't point those out. Minimum first page bids are a good example - there should be a quality level (not score, but concept) where no bid is high enough to make the first page - but that doesn't happen.
- What I mean by Preferred Customer Program is that High CTR, both in terms of complete account history, and at the AdGroup level, both help successfull clients do better with new and less powerful keywords. They get a boost on that keyword that a new or less successful client would not.
- I too find the issues of tranparency and fairness interesting. I'll do another full blog post about it this week, and certainly some will be covered in the webinar.
