AdWords and the Clicks-To-Visit Ratio

Questioning the Google billing of AdWords clicks

[Ben_Graham]

Here's a question for the AdWordsAdvisor, and the community at large. I've been doing some analysis on our web logs, and have noticed that there are some real inconsistencies between the clicks I pay for, and the visits I receive from Google AdWords. First, some definitions...

I define a visit as tightly as I can, and use a session ID to do so. This means that if the user closes out their browser I will consider that a new visit. I have no persistent cookies.

In analyzing my traffic I have created a new (or at least new-to-me) metric: CTV, or Clicks To Visits. What I've discovered is that different advertising mediums charge me consistently at different ratios. So Google AdWords, for instance, will consistently charge me for clicks that I never see. I advertise in large volume (certainly statistically significant) through several systems. What I've come up with is as follows:

Overture charges me for a set percentage more clicks than I ever see. It is very consistent from week to week (we'll call it X). So, if I was to produce 1,000 visits from Overture it would result in a bill for 1,000 * (1 + X). The X represents over-billed clicks.

Google Premium Sponsorships consistently charges me for four (4X) times as many over-billed clicks as Overture.

Google AdWords charges for 5 times (5X) as many over-billed clicks as Overture.

LookSmart charges a whopping 10 times (10X) as many over-billed clicks as Overture.

Now, the response of the CPC companies would be that a certain number of people double-click, and that will cause the problem. But shouldn't consumer behaviour through Overture mirror that at Google or through LookSmart? These numbers match week-in, week-out.

Just in case I had an error in my reporting, I went back over our logs using WebTrends, and the numbers hold true there.

My question for the community is to ask whether or not they notice similar behaviour amongst the different CPC providers. My question for AdWordsAdvisor is how do you define a click, what do you do to ensure I do not double-pay for clicks, and what do you consider an acceptable CTV?

[mcavic]

I haven't used any PPC - just signed up for AdWords today. But this doesn't seem right to me at all. I'd expect the PPC providers to filter out duplicate clicks within a certain time frame, and given that, I'd expect the click-to-visit ratio to be exactly 1.0.

It sounds to me like they think they can get away with overbilling you.

[Ben_Graham]

I would agree that they shouldn't charge for double-clicks, etc., but they certainly do. In the past I expected them to make claims like "it's impossible to detect double-clicks" or "we're working hard to prevent click fraud". What I can't quite swallow is why Overture is much, much better at correcting for double-clickers than Google AdWords or LookSmart is. Even Google Premium, which charges on a CPM basis, reports more clicks than I ever see in visits or visitors.

[Clark]

Overture has been doing this longer. What I would be curious about is the bots. Do you have a filter to keep out bots? Is your filter keeping out the bots more robust than Google's filter not charging you for bots?

[Ben_Graham]

My ads all have URL tracking in them. In order to skew the results above the bot would have to go to to [----?source=gaws&term=-----....] Of course, that would generate more visits than clicks, so they'd have to do it through www.google.com in order to generate more clicks than visits.

I do remove known bots from my web logs when doing analysis of this type, so if a known bot goes through Google to my site it would throw off my numbers. But I feel that I shouldn't pay Google to run Googlebot through ads, if that's what's happening. If it's not Googlebot generating the clicks then I think everyone would have to agree that it's click fraud from another bot.

My guess is it's none of these, it's just very poor accounting on the publishers' part. I'm sure they'd fix it if it went the other way, and I only paid for 80% of the clicks they worked so hard for.

[mcavic]

Ah, that's true. Google could be letting a bot through, only for it to be blocked at the site, which rightfully would be a click, but not a visit.

And, Google might be better at keeping the bots out than Overture, which would be why Overture would be billing more visitless clicks.

I don't block anything, and I'll definitely be checking my logs.

[Clark]

Ben,

What kind of bot blocking do you do? Is it something dynamic? Or just the main ones? Or you just check from time to time for misbehaving bots and add them.

I'm sure googlebot does not click on their own adwords links. But the other bots may be something else entirely and I really wouldn't discount this possibility.

[Ben_Graham]

Clark:

I just re-ran the numbers and calculated the percentage of my whole traffic that comes in with the source= URL tag that I have identified as bots or internal traffic. The bots / internal traffic total 0.1% of all traffic with a source= tag, indicating no significant volume.

And I don't block the bots, I just exclude them from my internal reporting. Heaven forbid, blocking bots!

Ben.

[Clark]

Ah, ok. I was thinking more along the lines of a "bad" bot, the type that go after email addresses.

[Furmanov]

so far found google to be quite accurate in calculating the number of visits but there were indeed few (but really few) clicks that didn't show up in logs, we use?source=...&search=.... type of tracking, the web-server wasn't down (100% sure about this since we are monitoring it from outside the network etc) so I have no idea why this happened

also I had this impression that google can filter out bots, at least they don't seem to charge for googlebot "clicks" :)

what I hate is paying for multiple clicks from the same visitor, I guess this is about 50%-60% of what i pay to google

and hate paying for searches my ad wasn't supposed to be shown for

and the adwords' support response on both issues was kind of frustrating

[Ben_Graham]

In order to ensure there is no lack of figures, I've done even more analysis on the traffic in order to clear up any confusion about possible double-clickers. Careful double-checking indicates that approximately 5% of users may leave the site and click on another ad without closing out their browser (and being issued a new session ID). This is still below what Google is over-billing me according to my logs. Even if we say that they can't / won't police their users and they are allowed to double / triple / quadruple click to their hearts' content they still don't generate nearly as many page requests as they do billed clicks.

Anyone else able to replicate my ratios?

[Advertiser 123]

Ben,

I also sent you a personal message and hope you got it. I would also like to understand the count. Let me know if you got the message please.

[bttmfeed]

So far this month we're seeing the same consistent ctv as we have over the past 14 months months, which hovers around 1.015. So paying out on 1.5% more than we receive.

Not enough Overture or Looksmart data to comment about though.