First time here. I'm doing some volunteering for a medium-sized non-profit and I've taken up analyzing their online strategy. I'm pretty much a novice, but they are even more so. I'm hoping you might help me fight what I think is small-scale (or possibly large scale) click fraud.
This is the scenario: they have a large contract with a referral site which supplies 15% of their traffic. They pay that site for referrals through search results and now and then a prominent banner on the homepage.
It started when I noticed the banner didn't convert much at all. Or rather, I'm not sure because the tracking was not set properly (transaction was done at another site). I had to model the conversion statistically based on the sales for each day. The banner came out very low, like 15% of the rest. That was my first suspicion.
I looked for other strange patterns and found unusual activity from a tiny town close to the referral website owner, for years. Almost every day a few visits, both via ads and organic/direct. This stopped in October last year, when the non-profit I worked at implemented an online ticket shop.
Recently I've seen bot activity. Organic search results came in at 100/hr from a certain city, with the same OS/browser, whereas before that city showed very little activity. This bot activity coincided with the second day of a recent banner campaign. It looks like they are "propping up" their ads with organic search results, thinking people won't delve into the channels.
Is there any innocent explanation for all this?
This is still pretty small theft, if it is that. But my statistical model suggests much lower conversion for traffic via the banner. It may just be wrong. I don't see any odd location or server clustering of the sort described above. Is there any failproof way to check whether this activity is from bots? Or maybe it's just a bunch of real people clicking for cash, I don't know.
First thing I want to do is close the measuring loop. But I need approval from the website operator for that, so that may take some time.
Thanks for any responses!