Hi there,
I read an older thread with a similar topic, so I post anew:
We have a network of websites promoting basically 3 shops which all sit on the same domain. But the promoting websites are 2 other domains. We’d like to know which keywords are worth our while to optimize and create backlinking campaigns for.
But any software we use so far doesn’t tell us how profitable a keyphrase is. It tells us by which keyphrases people arrive at our promotion domains, fine. But then these people are supposed to buy from our shop domain and we only see the referal in the statistics but from then on nothing more regarding the keyphrase they had used in the first place.
So, basically our dream would be a list saying:
People who arrived through phrase “abc” on domain xyz.com purchased to a percentage of x% a product in one of our shops. Even better: they purchased product-a to x%, product-b to y% and generated a turnover of A EUR in month august 2010 etc.
Now I know that we can use GA for all of our websites. The question is: is there any way to unify the data from different GA-identifiers into a statistics showing clearly the required data. Perhaps using the API?
Does anybody here know a piece of software that could do this tracking?
Many thanks
Marie Francoise
Really useful table you got here. Can tell right away whether it is worth buying ads for a keyword or not, by the last column. Cool.
It looks like handmade. Did you write SQL to get the data out of X-Cart’s database?
Not yet. My scripts are pretty basic and they most probably won’t work on other sites because I used functions from other scripts on my site. And the reports are quite ugly 
LOL Piwik’s API doesn’t have a method to track a goal You can only track goals with piwikTracker JavaScript object:
<script type="text/javascript">
piwikTracker = Piwik.getTracker( 'piwik.php', 12345 );
piwikTracker.trackGoal( 999, 123.45 );
</script>
Of course you cannot use it in a IPN catching script - it’s not happening in user’s browser.
BUT, if you look in the inspector at what happens when the trackGoal method is called, you will see that it actually loads a tracking pixel with many parameters (see screenshot). So I assume that if you do the same request from PHP or other server-side script, it should work
Hat off for using open source Still, Piwik is like any other stats app - very general, and you have to add so much yourself to get the data you need.
AFAIK you can send a “Success URL” to PayPal with order data and then PayPal sends an HTTP reguest to the URL when the payment is OK. So you must catch that request and ping Piwik through API to report a goal success. (I did not do that, this is just an idea.)
you can beautify your screens as a last step
What counts are really valid parameters.
What I did in connection with our 3 shops running with xcart is really meaningfull since it takes the real order values from the xcart data and associates these summed up to keyphrases.
Attached a screen shot.
And this one also covers those PayPal orders where people didn’t click for returning to our shops.
The top keyphrase “vitamine und mehr” doesn’t say a lot since this phrase is only used by those who already know us but forgot the exact domain.
But I am pleased to see that our efforts in ranking high for “potenzmittel” really is worth our while. But look into the last column (value per visitor)! If you pay 1 EUR for each click on this keyword (I guess the adwords price is even higher) you can never ever make a profit from this keyword!
This keyphrase list reveals some gems for us we haven’t even been aware of before. So, we have a strategy for optimization. We didn’t notice f.e. that “Lysin” is such a promising term…
What this kind of list naturally cant tell you: if you optimized for a keyphrase which doesn’t yet deliver any visitors currently, it might boost your profit a lot (or not) - you can only decide on a rational basis once you get a decent amount of visitors through this phrase.
Interesting stuff but a lot of work involved before you come to this data!
Correctly reasoning, these results apply to organic searches. Drawing a conclusion you coud bid, let’s say, 30c on ‘Potenzmittel’ and would likely have a profitable campaign is very far streched. People who click on advertisments behave differently I have seen this several times.
But it tells us in column “Tot-Wert” (total value, and this is already broken down to the bottom line profit for us) how much we could easily spend on a SEO campaign for lifting our sites even higher for one of these terms if we’d like to aggressively dominate the field.
Piwik allows only for a fixed value per goal. This is nonsense from a realistic point of view. So, I am getting the real values by sql requests from our xcart shopping cart databases.
The system as it is now rarely misses a keyword which converted to a sale. Omissions happen if a visitor lands on one of our pages by a searchphrase and then waits a few days or hours before ordering. Then it depends: is he coming with the same IP (good)? Or is he still “under” the same visitor_idcookie (good)? In these 2 cases we can assign the keyword he used a few days ago to his purchase, if not, we miss this relation and book his visit as a direct browser entry and count the keyword as not converted.
I am kinda excited that I succeeded to write this script to this perfection. It cost me 2 weeks of my life. 
… and some more days, I guess, since I just discovered another hole in the system which can be filled. It never ends once you start. Yes, it’s awesome!
I agree completely. I wonder why there isn’t any really good stats software available. Perhaps nobody is really interested in knowing “the truth” Perhaps the setup is always complicated since there are so many ways to do biz on the web and different strategies to gain visitors.
Once more you are right regarding the PayPal success reply. I also thought in this possibility but then found it absolutely confusing what they describe as piwik API. I am simply too dumb to figure out on how to insert a line into the paypal-called script in order to notify piwik on the goal.
So, let me ask you: you can provide a better solution?
mariefranc
GA sucks at referral tracking. It won’t give you the accurate data.
Basically you need to save the Referer string of each new visitor who opens any of your domains. Then assign the visitor an ID. Then read the visitor’s ID on checkout success page and save the ID along order data.
Then you will be able to code fancy reports about performance of keywords and domains.
Really, it’s not that complex, you can code that yourself.
Hello vassilevsky,
assign a visitor ID, you mean by placing a cookie on his PC?
I don’t know if anyone follows this thread but here is my solution which I developed meanwhile:
I’m using piwik which I installed on the server where the SSL area of our shops resides (who wants an alert pop up on the order confirmation page?). Then I put all our websites into one PIWIK site account (site id).
Then I placed the piwik tracking code of this domain in all the other domains pages.
Now this results in a mountain of data which PIWIK doesn’t display in the best possible way. F.e. piwik misses all goals from payments with delayed effect or if f.e. PayPal customers do not click back to our shop.
My stats show that this makes more than 10% of all orders which are missed by piwik goal detection. Merde!!
What I did: I wrote a script starting at the confirmed orders. Made the connection to piwik by visitor IP and get the keyword data now in relation to the orders. The details are very, very complicated I can assure you. If you want accurate results.
Because it doesn’t help in any way to know order xyz came with keyphrase abrakadabra. Instead you have to know: abrakadabra searching people landed x-times on our websites and y of them purchase with an average value of z EUR. Makes more sense. Then you know the exact keyword value and can calculate better how much to spend to push one of the sites in the top position for this phrase.
Quite tricky if you dont want to draw false conclusions.
All further complicates if you have a newletter.
Sometimes people buy from newsletters they received 2 weeks ago. I see people purchasing from newsletters from May! So, how did they come to the website signing up??
The more you think the longer you sit in front of the PC.
Ok, off to the beach!!
Greetings
