The Most Active and Friendliest
Affiliate Marketing Community Online!

“Adavice”/  “RollerAds”/

How do you optimise your campaigns?

dragonfly666

New Member
affiliate
One thing that I'm struggling to get some info on is (after gaining some statistical data over a good period of time) what should I focus on viewing and blacklisting or whitelisting first?

Is there any particular list you go through in a certain order to make decisions on what to observe and cut first? OS, carrier, site ID?

I hope this makes sense!
 
IF clicks > x AND LP clicks < 1 THEN blacklist placement

IF spend > $x AND conversions < 1 THEN blacklist placement

IF clicks > x AND LP CTR = 100% THEN blacklist placement

IF spend > $x AND CPA < $x THEN increase bid

IF spend > $x AND CPA > $x THEN decrease Bid

There's a few examples for what I do for optimisation. It applies to any tracking parameter you're looking at. Only do these when you've let said placement spend enough money. Most say 5-10x of payout but it honestly depends on the offer payout. I do some of those before it's even spent 1x of payout, especially the bot check ones like LP clicks less than 0 or LP CTR at 100% as you can see those pretty quickly.
 
IF clicks > x AND LP clicks < 1 THEN blacklist placement

IF spend > $x AND conversions < 1 THEN blacklist placement

IF clicks > x AND LP CTR = 100% THEN blacklist placement

IF spend > $x AND CPA < $x THEN increase bid

IF spend > $x AND CPA > $x THEN decrease Bid

There's a few examples for what I do for optimisation. It applies to any tracking parameter you're looking at. Only do these when you've let said placement spend enough money. Most say 5-10x of payout but it honestly depends on the offer payout. I do some of those before it's even spent 1x of payout, especially the bot check ones like LP clicks less than 0 or LP CTR at 100% as you can see those pretty quickly.

perfect, that gives me a good outline on what to do now, thank you!
 
perfect, that gives me a good outline on what to do now, thank you!

No worries mate. Good luck out there! Also if you're working with push traffic, be aware that I have noticed a recent trend of a few push traffic sources sending pop/redirect traffic disguised as push traffic. It can be hard to tell the difference besides possibly LP CTR which is why I've been trying to implement a new parameter which tracks how long they stay on the landing page. Generally a push notification clicker is at least somewhat interested in the ad and will generally look at the landing page for a longer time than someone who just got a random pop-up who will most likely close it straight away. That is my biggest annoyance currently along with raped subscriber bases who have been spammed 10s or 100s of notifications a day and are now blind to everything. Been building my own push DB because I'm sick of all this nonsense.
 
how can you identify a certain IP as a bot IP
Take the IP list and count the dupe IPs


IP-to-ASN - Team Cymru
Code:
$./asinfo.sh
###write this script to the server chmod 755###
#!/bin/bash
#asinfo.sh

echo "Pls enter your ip:"
read ip
whois -h whois.cymru.com "$ip"
 #curl "https://ipinfo.io/$ip"

exit

for bulk (I keep this to a few hundred at a time)
install netcat on the server
Code:
netcat whois.cymru.com 43 < infile | sort -n > outfile

People pay services with blocklists because this is complex. find an ad network that uses foresniq Forensiq Ad Fraud Detection - Impact - Impact
 
Last edited:
use logging with a refreshing AJAX Javascript.

Yeah I was looking at something along those lines. There's a library here: jasonzissman/TimeMe.js

What I'm planning to do is implement that and then when the user leaves the page I'll execute a postback to my tracking server with the time spent value and click_id. Then for those that leave the page open indefinitely I'll set it to postback if the value reaches >= 2-3minutes - something like that.
 
One thing that I'm struggling to get some info on is (after gaining some statistical data over a good period of time) what should I focus on viewing and blacklisting or whitelisting first?

Is there any particular list you go through in a certain order to make decisions on what to observe and cut first? OS, carrier, site ID?

I hope this makes sense!


Hi, you can use a tracking tool. In general they have a special kit for detecting bot and suspicious traffic. I use AdsBridge (aff link) with their anti-fraud and bot-filter. As for me it is quite accurate.
 
Last edited by a moderator:
thanks, guys for all the help, I'm using BeMob, but they don't have bot detection? I mean I would assume that they do, but I can't find anything anywhere about it.
 
One thing that I'm struggling to get some info on is (after gaining some statistical data over a good period of time) what should I focus on viewing and blacklisting or whitelisting first?

Is there any particular list you go through in a certain order to make decisions on what to observe and cut first? OS, carrier, site ID?

I hope this makes sense!

1. Start with a whitelist next to it so you can run faster profitable
2. Blacklist OS and carrier really depends on the offer (for mobile content you will chose faster to blacklist OS or carrier) but i prefer most of the time blacklisting site_id's first.

For excluding bot traffic we can help you with that!
We have an advanced fraud filter system and are really transparant in what we call "suspicious traffic". Most of all push traffic we automatically refund suspicious traffic. Add me on skype for more info: Jaerts92
 
See what the sources are for the C blocks 12.234.59/24 (.1-.254)
Code:
~/$ cut -d'.' -f1-3 list.file|sort|uniq -c|sort -nr
that can be very revealing but you need to check the allocations.

list.file ip 1 per line
 
I just wish all that networks give us more correct traffic in all their terms anyway. Who can pay that much for traffic when sometimes all your potential ROI from mediabuys is around 20% and you totally can see it clearly here and there all together. Let's hope it will be no more.
 
I highly recommend this API for you:

1) ip-api.com, they can detect Non-residential IPS, proxies and VPNs...

Just find a developer to integrate this solution for you...

2) Check ISP and Operator...

3) Additionally to that create an interactive Landing page, and based on the user interaction you could prove he is human..., then Ping an Ajax Script...
Then when the user reaches the CTA, your server-side has all the necessary information about the user...
It is human ==> Go landing page...
else if not human go ==> fake landing page or revenue share...

4) Check user Agent of browsers and compare it to browser size...

There are multiple tactics to prove that the user is Human or not...
 
Maybe someone who is more in coding can give here step by step solution
Code:
#!/bin/bash
#-vx
#rand_network.sh
COUNT=1
start=$(date)

dest=sample.csv #outfile
for word in `cat ipstat-sample`#infile
do
echo "$start - $(date)"

                 

echo -n  $word,>> sample.csv

echo $word

netdata=$(host $word)
printf "%s" "$netdata">> "$dest"

echo $COUNT

echo >> "$dest"
COUNT=$((COUNT+1))


done
#infile is a list of IPs one per line.

Sample output #outfile:
Code:
87.91.9.26,26.9.91.87.in-addr.arpa domain name pointer 87-91-9-26.abo.bbox.fr.
37.166.58.154,154.58.166.37.in-addr.arpa domain name pointer 37-166-58-154.coucou-networks.fr.
176.181.236.70,70.236.181.176.in-addr.arpa domain name pointer i16-les02-ntr-176-181-236-70.sfr.lns.abo.bbox.fr.
37.173.255.237,Host 237.255.173.37.in-addr.arpa. not found: 3(NXDOMAIN)
31.38.43.49,49.43.38.31.in-addr.arpa domain name pointer pla13-h02-31-38-43-49.dsl.sta.abo.bbox.fr.

not found: 3 (NXDOMAIN) is 99% of the time a server with no assigned hostname

chmod 755 on a server to use this script (or on your LINUX install)
execute command:
$ ./rand_network.sh
in the directory it is in
COPYLEFT: charity-ware free to use with no warranty express or implied and offered AS-IS.
 
Last edited:
People don't use datacenter IPs (normally)
Hackers, bots, proxies and *guys like me do :D"
Who is the traffic and from where matters.
Browser signatures are so easy to forge.
Server farms' locations skew the GEO location of traffic networks.

How much garbage are you paying for in your traffic matters.
This garbage traffic negatively skews your conversion metrics when optimizing.

Garbage is garbage --you can't optimize it --that traffic will always be garbage --you can blacklist those ips.

I looked at a 600 CPC sample today of push traffic sampled from 9452 CPC billings and 17.83% was NXDOMAIN bot traffic. FRAUD traffic.
It was *just traffic* in this guy's tracker.
 
MI
Back