Announcing Changes to the TrafficEstimatorService

[AdWordsAPIAdvisor]

In an effort to provide more consistent and accurate keyword data to
our developers and advertisers, we will be changing some of the fields
within the KeywordEstimate object on Thursday, November 3rd.

We will add the following field to the KeywordEstimate object:

· clicksPerDay - (float) The estimated number of clicks generated
per day for a keyword in a given ad group

Similarly, we will remove the following fields from the KeywordEstimate
object:

· impressions - The estimated number of impressions for a given
keyword
· ctr - The estimated click-through-rate for a given keyword.
· notShownPerDay - The estimated number of times that the ad would
not be shown, despite a keyword match.

These changes will take effect on the afternoon of Thursday, November
3rd (Pacific time zone). At that time, we will update the
TrafficEstimatorService WSDL which may disrupt the operation of your
programs. Therefore, we have posted a temporary copy of the "new"
TrafficEstimatorService WSDL at
http://www.google.com/apis/adwords/developer/wsdl_preview.xml so you
can begin making the necessary adjustments now (This temporary copy
will be removed on November 4, 2005, after the normal WSDL location is
updated).

We are in the process of updating the AdWords API Developer Guide
(http://www.google.com/apis/adwords/developer/adwords_api_intro.html)
and plan to have these new fields added on Thursday.

--Rohit Dhawan, Product Manager

[Darrin]

Hi Rohit,

Thanks for the advance notice.

You mention "in an effort to provide more ... accurate .... data".
Does this mean that the algorithm behind the estimator is also
changing?

Thanks,
Darrin

[runge]

Hi

Given that the placement of an ad, the CTR, and the number of clicks
per day for the ad depends on many things (like how good the text is,
how well targeted it is..) it seems that removing the only hard fact,
impressions, will make the tool less usefully and not more.
Though I guess if we can get numbers on what Google assumes will be the
CTR for a certain position, we can get back the impressions value. That
way we can calculate the number of clicks with our own CTR value and
get much more accurate results.

So in short, can you at least give us a hint what CTR the tool thinks
will be on say position 1,2,and 3?

Thanks in advance

[Ben Michelson]

Runge,

I believe this may be just the first phase of a new
"less is more" concept.

I expect subsequent versions will alternately snip out
or merge previously inaccurate fields, until finally
(AdWords 1.0) the TrafficEstimatorService will be void
of inaccuracy by providing no information whatsoever.

Cf. The Bellman's Speech (stanza 1-4)

Ben

__________________________________
Yahoo! Mail - PC Magazine Editors' Choice 2005
http://mail.yahoo.com

[Kostas]

> it seems that removing the only hard fact,
> impressions, will make the tool less usefully and not more.

That was my thought as well, impressions is the only figure that can be
accurate. Perhaps the reason for removing it is that it gives us too
much information (how many searches there are on Google for a specific
phrase).

[Robert]

Well done, Google. I just want to release my first Adwords program -
partly based on the ctr value. I work about two month for it. Why we
should develop programs for Google, if Google changes the API every two
month (see also KeywordService)?

Thank you, Google :-(

[Craig Webster]

Hi Robert,

While I can understand that you're upset about your program it should be
noted that the API is fairly clearly marked as a *beta* release and as
such you should have taken steps to decouple your program from the
backend API.

This is exactly what the Facade pattern should be used for. See
http://cs.ucsb.edu/~cappello/50/lectures/patterns/Facade.html
for more info.

Yours,
C
--
Craig Webster | web: http://xeriom.net/
Xeriom.NET | tel: +44 (0)131 516 8595

[fahad]

Support,

Any ETA for consolidated and stable release of Google API soon?

Fahad
fha...@webdiversity.co.uk

[Nana Gilbert-Baffoe]

Hi,

I'm also quite disappointed to hear this. I depended on all the data
being removed from the API to offer estimated keyword search volume
data and other extended stats to my customers. Without this info my
tools are not as useful.

Unless I'm missing something, I really don't see the improvement in the
TrafficEstimatorService after the change. It looks more like a
downgrade. Can you please explain your full reasoning for this change?

I think it would help us more if you add the new clicksPerDay field and
keep the rest as well. That way we can choose what to use.

Thanks
---
Nana
http://www.technobloggie.com

[Patrick Chanezon]

No.

Thanks,
Darrin

--
Patrick Chanezon, AdWords API evangelist
http://blog.chanezon.com/
http://www.google.com/apis/adwords/

[Patrick Chanezon]

Sorry but no, we do not plan to disclose that data through the API.

P@

Thanks in advance

[Patrick Chanezon]

Fair enough.
Stanza 10 is pretty good fit as well:-)

" The Bellman perceived that their spirits were low,
And repeated in musical tone
Some jokes he had kept for a season of woe -
But the crew would do nothing but groan."

I'll spare you the jokes and would refer to Coleridge's "Rime of the Ancient Mariner"
"Facile credo, plures esse Naturas invisibiles quam visibiles in rerum universitate. Sed horum omnium familiam quis nobis enarrabit ? et gradus et cognationes et discrimina et singulorum munera ? Quid agunt ? quae loca habitant ?" - T. Burnet, Archaeol. Phil., p. 68 (slightly edited by Coleridge).

Translation:
"I can easily believe, that there are more invisible than visible Beings in the universe. But who shall describe for us their families? and their ranks and relationships and distinguishing features and functions? What they do? where they live?"

P@

[Patrick Chanezon]

I am sorry that this change disrupted your program, but as Craig mentions, the AdWords API is still in beta.

P@

[Patrick Chanezon]

I am sorry that this change affects the usefulness of your tools, but the change will happen next thursday.

P@

[Robert]

Ok, I'm just realize that I'm a beta tester only. Please tell us, when
do you think Google releases the final version? (I think, it is better
to wait for the final version, then I can see if it make sense to
finish my software.)

Robert

[runge]

Hi
Thanks for your quick respons, though it was not what I was hoping for.
Oh well, I guess I can calculate the CTR from the keyword history I
already have on Google and compare it to clicksPerDay to get
KeywordEstimate hidden CTR.
And then we also have Overture to the rescue with their estimates (
http://inventory.overture.com/d/searchinventory/suggestion/ ).
Both are of course moving targets but I think comparing those two
values and getting a value on how much a keyword gets searched on
Overture vs Google and then doing your own calculations on clicksPerDay
will be the best way to move forward.

What do you others think?

[runge]

Hi Ben
I am reading The Bellmans's Speech now.
Fun read :)

Thanks

[Patrick Chanezon]

Sorry but we don't have any official date for the end of beta yet.

P@

[Patrick Chanezon]

I second that:-)

P@

Thanks

[inasisi]

That the Google Estimator accessible through the UI is mediocre would
be to say the least. In many of the adwords forums, people have
recounted, how hopeless those estimates were. In my own case, I have
seen an estimated position of 4 when the actual position turned out to
be in the 30s and in other cases it has been far lesser. I have not
seen even one case in which the estimated position was more than 8. The
pleas from various adwords customers to improve the estimator seems to
have fallen on deaf ears. And now Google wants to stop showing the only
metric that was close to reality, namely the Impressions without
improving the algorithm.

My paranoid theory is that, Google is underestimating the position and
overestimating the clicks on purpose. This way the advertisers would
bid on more and more words lured by the high estimated clicks.

[Patrick Chanezon]

As you write it, this is a paranoid theory, that I want to contradict.
Google would never risk its reputation by skewing the results one way or another.
The algorithms used in the TrafficEstimator may return some results that do not match your quality expectations, but they are not skewed in any way.

P@

[inasisi]

If it is not on purpose, I don't understand why Google is not
correcting the huge skewness in its estimates and further remove the
only good statistics that we had to access to. If Google felt the need
to be consistent to both the API users and the advertisers who use the
UI, then they should have provided more information on the UI instead
of having to strip them from the API.

Just to prove the skewness of the estimates, I did a quick experiment
with the help of the Overture's Bid Tool and the Google Estimator. I
used Overture, as the bids are visible to everyone and I have found
Google and Overture to be equally competitive in most of our keywords.
Here is the result.

Keyword Google Estd Overture Actual Overture Price
Position for Position for for first
a bid of $ 0.1 a bid of $ 0.1 position
------------------------------------------------------------------------------------------
mortgage 7.6 173 $ 3.61
mesothelioma 7.0 33 $42.57
credit card 4.1 101 $ 4.72
life insurance 5.0 112 $ 8.72
auto insurance 7.5 134 $13.00

So what this shows is that for the keyword "mortgage", the Google
Estimator predicted a position of 7.6 when I bid with just $0.1. On
Overture I would have been in the 173rd position. All the other
keywords tell similar story of the huge underestimation by Google with
respect to the position. To keep the experiment clean, I chose keywords
that I have never used in my campaigns and also my overall campaign CTR
is in the normal range and nothing out of the ordinary.

If this is the skewness with respect position, the skewness with
respect to CTRs was also equally bad. Before when we were able to see
the expected CTRs, I had seen many cases in which the first position
was expected to get a CTR of 6% while the 7th position would get a CTR
of 4%. Come on give me a break, I have used Adwords enough to know that
the fall off in CTR with respect to the position is much more
precipitous.

So if Google is not doing it on purpose, it is guilty of wilful neglect
of the estimator or of hubris in thinking that the Estimates are good
despite so many users complaining about it. If Google is not able to
provide reliable estimates atleast give access to the Impression count
which is much more reliable. Overture probably realizes that estimation
of clicks is a tricky business and hence only shows the impression
estimates.(This should in no way be construed as a recommendation for
Overture as it has a whole bunch of other issues)

[inasisi]

Since the the tabs are converted to spaces the above table did not show
up properly. So I am reposting the above table with spaces instead of
tabs. Hopefully, this shows up better