Click Tracking isn't accurate

I've had a discussion on Telegram with Victoria (@vika_royalpartners). The problem I have is very similar to another program that uses Affilka software - Wild.io. The problem is that in my system I track clicks for every program. Royal Partners tracking doesn't appear to reflect any of that (and same with Wild.io affiliates). I sent several images of my tracking so that they could try to correlate it with their own statistics - but they seem unwilling to do so.

Their stats for the entire month shows 6 visits when I contacted them, while mine shows 41. Mine continues on for the entire month while theirs only shows a spike in a 4 day span. The responses I got were a little strange to me...

Quote:

---

I’ve discussed this with our technical team and would like to clarify the situation.

There may indeed be some inaccuracies in how clicks are reflected in our system. However, all key metrics that determine partner payments - such as first-time deposits, deposit amounts, and other financial indicators are tracked and displayed accurately.

You are welcome to use your own click-tracking system - there are no restrictions on our side. You already have access to the API, so you can reconcile the data in any format that is convenient for you.

---

Towards the end I said the following:

Quote:

---

I honestly would rather your team look at my clicks and tell me why they do not show up in the Affilka report. That would be ideal and show good faith.

---

I had told them I could export the clicks with IP, location, device information. Their response again was strange to me:

Quote:

---

We have the system as it is as we pay more attention on metrics we pay for. Our partners are able and free to choose any other 3rd party to monitor clicks

---

What is the deal? Is Affilka the problem? I'm doubtful of that because I also promote Rainbet which uses Affilka.. and they appear to at least be in the ballpark as far as my click-tracking versus theirs (Mine says 193 Clicks, theirs says 215).

We hope this issue was resolved HERE

I have some data on this that might help someone diagnose a similar problem.

During the first two weeks of August one program recorded 690 visits from our site and zero registrations. At that time there were only about 160 users on the site. Program was counting four times more visitors than we actually had.

It turned out there were two separate causes, and they need different fixes.

The first one was our own caching plugin. Its preloader scans every link it finds, and it found the affiliate redirects, so every warm up cycle generated fake clicks.

The second one was external. We extracted the raw click log from redirect plugin and grouped it. Without the preloader traffic that left 697 requests from 647 unique addresses across 639 different /24 subnets, and no subnet with more than 7 requests. The referer was empty on every one of them, and the user agents were all normal browser strings. The most common one was the standard Chrome DevTools mobile emulation string, 103 requests from 103 addresses.

On the networks, since that is the part worth sharing. The datacenter half came out of Tencent Cloud ranges, 43.160.x, 43.165.x, 43.134.x and 43.157.x, plus a single Microsoft Azure address. We read those ASNs off the ranges in our own log, so treat them as a starting point and confirm them in your panel. The rest was residential broadband, first octets 177, 186, 181 and 190, so mostly Latin America, with some 14 and 45. By country it was Brazil 96, US 71, Vietnam 61, France 40 and Bangladesh 23. Our search traffic comes from the Philippines and Indonesia, so there was no overlap at all.

How we fixed it. We excluded the affiliate redirect paths from cache preloading, and we blocked the datacenter ASNs at the edge on those paths only.

Two warnings that come with that. The residential half cannot be blocked by ranges at all, because almost every request arrives from its own subnet, so it only shows up as a behaviour pattern. And I would not put a challenge on redirect path, because that breaks clicks for real players.

Worth pulling your own logs before you suspect the program. From their side our traffic looked like click fraud rather than a weak affiliate and that ended our conversation about a CPA deal