keyword cost to click disconnect

[Bionic]

Tweet

Hi,

I use a db to track all my ad clicks and log relevant data. I also use keyword reports to try and match cost to those clicks. However, I have not yet discovered a highly accurate way of doing this, for two reasons.

One, it is difficult to keep the database clean of spurious clicks from bots and even the likes of Google, so therefore there are generally too many clicks in the db. I can filter out bots using the user-agent, but even Google don't always have the word 'bot' for instance in all their clicks.

Second, the keyword cost is only provided by keyword. The only way I have got a reasonable match is by averaging the cost per click per day for the keyword then matching it in my db. It's all a bit woolly, but I cannot see a more accurate way of doing this. By the way, I don't want to use conversion tracking in Google for instance (a combination of paranoia and having a system that works with more than one ad provider).

So what do others do to filter out unwanted clicks, and secondly to match costs? I use php, but generally it is the methodology that is interesting. Of course, any good resources would be welcome :-)

[D-Mac]

With regard to filtering out the unwanted clicks, you could try only storing referrers with a keyword query string. Bots tend not to have ?q=my+great+converting+keywords in the referrer string. Another method is to whitelist the referrers you want to keep data on, and ignore the rest (you can still store them, but filter them from your analysis results)

It's been a while since I've imported click data from adwords. Are you referring to the fact that they lump lots of keyword variations under one keyword, and say that you had 27 clicks that cost you £1.15? If this is the case, you probably know that you can still work out the actual keyword used, by reading the referrer string. This doesn't solve the problem that you have 27 clicks that cost a total of £1.15 though. As far a I know, the only solution is to do what you are already doing, and divide £1.15 by 27 to work out the average cost for each click, and then assign it to the actual keyword used (rather than the one they say it is). As I said though, it's been a while since I've looked at this, and it maybe that api has some nice shiny new options.

[Bionic]

Thanks for the reply D-Mac. I did wonder, but having trawled a lot of rubbish over the last few days, I have not seen a better way to assign costs to keywords. It is good to get a different view. Working in isolation, I often wonder if I am missing something obvious.

Saying that, making more use of the 'referer' is a good idea. I will take a closer look at that. :-)

Thanks for the ideas.