STM Mobile Tracker's Initial Test & Maybe More... (41)
Hey Stackers,
I'm gonna reveal one of the small campaigns that I've tested with STM Mobile Tracker here. I will try to list everything here and you guys can feel free to copy the entire campaign as you like 
Hope you guys would like it and Jordan, Besmir and Lorenzo ( sorry Lorenzo, I left you behind, my bad... Don't bite me ) should take the credit =)
*WARNING! :
Beware ADHD kids, this is gonna be a very long post. So be prepared 
Campaign Details:
Offer: Badoo Android App Install ES ( Neverblue )
Payout: $1.50
Traffic Source: Airpush
Server: Beyond Hosting VPS 1024 ( with Litespeed Installed )
Tracker: The Mighty STM Mobile Tracker
Promotional Method: 100% Direct Link
Total Spending: $753.50
Total Earnings: $463.50
ROI: -38.5%
Campaign Details on Airpush:
Airpush now has a few different more promotional methods, the one I used for this campaign was Push URL, the overall stats on Airpush for this particular campaign is as below:
And below is my targeting setting for this campaign:
The country is set for Spain, all states, all devices, all version, all carriers without WIFI traffic and was bidding at $0.010.
Now my creatives & stats:
As you can see above, I'm passing Airpush's dynamic variables back to my STM Mobile Tracker so later I can compare the carrier/device data with the STM tracker built-in carrier/device dectection data. Some may ask what dynamic variables I'm using, they are as below:
%carrier%
%Device%
%manufacturer%
%creativeid%
As for the Spanish ad copies, I simply go to Google translator and use those translated Spanish copies on google.es. And then I can grab some of the organic results or paid ads results that are written in proper grammar. This way I can test faster and don't have to wait for OneHourTranslation to get my Spanish copies done.
Here it comes the interesting part, STM Mobile Tracker's stats:
First of all, I would really like to compare the data that Airpush is passing me with the built-in detection feature, so the answer is as below:
Airpush's Data ( Carrier ):
STM Mobile Tracker's Built-in Detection Data ( Carrier ):
I'm not going to attach all of the data with pics here as we still have handsets, brand, OS, etc. But if you guys would like to get your hands on them, download them at the link below, I've labeled the excel files with the name accordingly. After you open them with excel just analyze them by using the pivot table so that you can compare them all if you want. If you don't know how to use pivot table, I urge you to watch the tutorial at http://ipyxel.com/fast-mobile-data-a...tables-primer/ ( Epic thanks to ffclogin )
Download the full data spreadsheet here
Ok now back to the comparison above, as you can see there is so much difference between the carrier data given by Airpush and STM tracker, so in this case if I were to optimize the campaign, I will be definitely using the data given by Airpush as you need to cut off the fat within the Airpush targeting, so this literally makes the data from STM tracker not reliable, maybe this is just for Airpush which I'm not sure. Perhaps Besmir can chime in and tell us how is the results for other ad networks.
Ok now I will share the most interesting part of this post which is comparing the "REAL" clicks that I really received in my CPA Network, please pay attention to the below image I took from my Neverblue account:
Please ignore the conversions and revenue as it's not consistent with the images above, this is due to I have some other clicks that were generated without using the STM tracker. So please only pay attention to the Clicks column.
Total Clicks received in Neverblue: 35000 ++
Clicks Showed in STM Tracker: 60000 ++
Clicks that I paid on Airpush: 78,000 ++
So if you guys compare the clicks between Neverblue and Airpush, there is some huge discrepancy there which is more than 50% of the clicks I paid for is totally gone, it really hurts as I almost pay for the extra 50% of clicks for nothing.
There is a few reasons behind this:
1. My server's problem ( This test is with BeyondHosting VPS 1024 )
2. Something to do with p202 itself while handling mobile traffic
3. Airpush's behavior
4. Too many returning visitors so Neverblue has already filtered those clicks ( but 50% of the clicks ! )
I would really like to see more experienced mobile affiliates chime into this thread and share their thoughts on this, this can definitely help all of us.
Besides, I would really like to see some other mobile veterans that are using Mobaff Tracker and iMobitrax can share their experience on the tools they use to track mobile campaigns too. That will be very much appreciated.
@Besmir
And one little suggestion or feature request for you if this could be done on this addon, is that possible to custom build a feature that can allow us to import the csv/excel file that we download from our CPA networks? So this way we can just direct all of the traffic to our CPA network links and just pass subid by using the old-school method, after that we just import the csv/excel downloaded from the networks to p202. This is absolutely useful for ad networks that can pass variables back to us. What do you think?
Ok, if you guys have more questions you would like to ask, feel free to shoot them in this thread. I will try my best to help.
Lastly, thank you so much for reading this post
UPDATE:
This time without using STM Mobile Tracker, I've sent another 7000 clicks to another offer using a simple PHP redirect script with my server just to find out something, this time my CPA network stat is showing only 2500+ clicks received, once I noticed this I directly got the campaign paused. I think the server plays an important role too when there is redirect script involved, so I'm sure a faster server will definitely help in this case.
2nd UPDATE:
Tyler is very helpful and got some test done on my server, we didn't find any problem with my server at all with the PHP redirect script that I'm using to pass that 7000 clicks mentioned above. So now the only reason left is to do with Airpush, I've found that different countries have different clicks discrepancy on Airpush. For instance, I've seen only 20% - 30% click discrepancy in France compared to Spain and Mexico that have almost up to 50% of the click loss.
This might be due to a few reasons as below:
1. Returning users clicking on the same ad, this leads to CPA networks filtering out duplicate IP/Clicks.
2. Publishers' performance in different countries vary
3. Limited demo size in that particular country so that a bunch of people are the same ones so it leads to problem number 1.
My apology to Tyler as I said it's definitely something to do with my server configuration which is not the case, kudos to Tyler as he's really awesome!
*I've heard that the new STM p202 Mod will be 4x faster for the next release!
DISCLAIMER:
This thread is not to prove how good or bad the STM tracker performs , it's really for all of us to share our mobile tracking experience and also discuss further upon what other factors could possibly affect our mobile tracking experience like I mentioned in the post. Besides, my another test ( updated above ) shows me that the server itself absolutely plays an important role while there is redirect script involved. After all, STM Mobile Tracker is free and Besmir they all will continue working on this tool to make it an even better one.
And I would just like to say STM forum is simply awesome!
Awesome post wyffgoal. I've been running a little traffic with Airpush across several campaigns and have noticed something similar in terms of click discrepancies. Direct linking etc. My CTR's from Airpush have been pretty good. 80-90%. A little too good to be true given I haven't seen one conversion yet. Although I haven't sent tens thousands of clicks like you have, I would have expected to have seen some conversions. I'd be interested to hear more about Airpush as a traffic source.
Just an idea, but looking at the click loss you're experiencing, why not just use NeverBlues in built tracking. You can add up to 5 sub ids and I think I'm right in saying (although I've never run any mobile traffic) that you can filter the reports based on carrier, handset etc.
Logic would dictate that 1 less link in the chain would speed things up and hopefully help on the conversions.
Discrepancy beween AP and NB is always huge. NB filters out duplicate clicks by IP, which is a ton in mobile. The numbers you are seeing are pretty normal as far as my experience. Don't read too much into the discrepancies, you've gotta just make it work.
@izzy
Every ad network behaves differently, there are so many variables will cause how an ad network behave such as demographics of the majority users on the network, quality of the publishers, the way they give you the traffic and so much more. So only test more will lead us to the answer.
@jonemd
I've been doing this already, it definitely helps in terms of the click loss. Unfortunately it's really pain in the ass while you analyze those data, you will only know once you try doing it yourself.
@goodrich
Yeah, this is what I've been thinking, but as jonemd says, cut out 1 more redirect link will definitely help which is our third-party tracking tool. I've seen much more improvement ROI wise when we just drive all the traffic directly to the CPA network's link.
The click discrepancy isn't a huge worry for me, as others have mentioned.
The huge carrier discrepancy is interesting though. Someone's database isn't doing a very good job and it's probably Maxmind. Would like to see what others think.
-Tom
PS, I feel pretty honored to be mentioned in the thread actually 
@ffclogin
Tom, you deserve it as I always learn some good stuff from your blog. Please do keep up the good work!
As for the click discrepancy, STM Mobile tracker is using Maxmind database, so I really have no idea what Airpush is using. Perhaps I could just send the data over to my rep and see how they react.
maxmind IS NOT a carrier db, there is no such thing....its just who the ISP is...
airpush (and many other traffic sources) prolly get the carrier info from the device itself (hard coded into the app) which obviously is more accurate....cuz in theroy if I am roaming it may show as a dif carrier
this post also made me check some stats....
with 202 seems like your loss is about 23% from AP to 202
in comparison I use has-offers....loss on avg I have seen is around 19% ap ---> has offers
this could be due to a global infrastructure they have .... either way 202 is free ho isn't so not sure if in the end it would even out or not....
anyways just wanted to give a comparison for those who are interested
Awesome post wyffgoal, i wanted to see users experiences with the tracker
First lets start off with the campaign itself and turning it into the GREEN. I've dibble dabbled with badoo, but i didn't love the cap limits, have they cap increased recently?
- Your ad title: "?Cansado de Facebook?"
- Translated: "Tired of Facebook?"
- Your ad body: "Badoo - unete a tus amigos ahora"
- Translated it means: "Badoo - Join your friends now"
I would recommend some spicier ads!
- Hey Babe, are you on BaDoo yet?
- Oh lala, are you on BaDoO??
*Add weird *&%$ in your text, grab the users attention!
For your body copy:
- Meet me on Badoo, download and reg now!
- Badoo is the coolest app out, download now.
Obviously translate all to spanish
Out of my tests the most interesting thing i tested was comparing a landing page vs direct linking. I got similar results, which tells us 1 thing. A better landing page will out convert direct linking. This page needs to look similar to the app, colors etc…
Onto the click differences, i completely forgot NB cancels duplicate clicks, which is absolutely major in the mobile world. So props to @goodrich.
Aside from that server setup/speeds is usually #1 reason for any click loss, i wouldn't completely ignore it, but i bet the actual % of loosed unique clicks is probably closer to 5%.
Try this campaign again with some new concepts and lets get it in the profit zone?
Hey man , first , great case study!
Jordan gave some great tips about the creative part so I am going to cover the "tech" side.
First thing , about the carrier difference. From my experience different ISP databases tend to name the same ISP with different names.
To give you an example , ATT is mostly known as "Service Provider Corporation" apart from other 5 different names.
The way I have done it is to spend like $10 buying only ATT traffic let's say and make the association with the name(s) shown in prosper.
So when I start the real campaign I already know that "Service Provider Corporation" is ATT.
Due to the big amount of carriers out there , it's pretty hard to hardcode this info on prosper for every country / carrier ( could be an idea for a collaborative database? )
Also , from what I know , the other tracking solutions out there , use Maxmind , so using a different tracker will yield the same results.
Another point I would like to explain is about the "click loss" issue.
First , in your specific case , I think the server specs play an important role.
Second and most important , by knowing how prosper internally works , NONE of the traffic is lost.
At worst it goes untracked , but is never lost. Meaning the user might have JS/images disabled so it won't get tracked , but that user can still click on the links on the page and convert.
Offtopic : the new release of the mobile tracker will have much faster redirects due to a new change how the detection of carriers is done
@stackman
Yeah, they have increased the cap quite a bit but you will still hit the cap in no time. So I just let go this campaign and focus on other offers that have unlimited caps and potential. Thanks for all of the inputs.
@bbrock32
Thanks for clarifying everything, really appreciate that!
Great reading yo!
UPDATE:
This time without using STM Mobile Tracker, I've sent another 7000 clicks to another offer using a simple PHP redirect script with my server just to find out something, this time my CPA network stat is showing only 2500+ clicks received, once I noticed this I directly got the campaign paused. I think the server plays an important role too when there is redirect script involved, so I'm sure a faster server will definitely help in this case.
I would have another test on some other ad networks and on a better server anytime soon.
Stay tuned!
^^ Haha funny that you "lost" more clicks going through a php redirect than through 202.
Yep a decent server is key, hit up tyler from beyond he's got the goods.
I just got that same VPS package from beyond Hosting and now this got me thinking.... Maybe I should upgrade to Their fastest option the R610 dedicated server?
I rather do that than lose money on potential winning campaigns....
You're already spending so much on this campaign. Wouldn't hurt to try out BeyondHosting's most expensive dedi.
@Mr Green
Yeah, that's weird and it's definitely something to do with my server. I've already contacted Tyler =)
@latinstack
You can try to consult Tyler from Beyond, I'm still waiting for his reply.
@ryanco1990
Yeah, I've already contacted Tyler regarding the server configuration.
Offtopic - Did some tests today and the next release of the STM tracker will have 4x faster redirects due to using memcached on everything
Hey guys, responding my slew of pms now and you will hear from me.
I want to work with you directly to see whats causing the click loss. Losing 4500 impressions on a php redirect concerns me..... I somehow think we will find out its not the server though.
2nd UPDATE:
Tyler is very helpful and got some test done on my server, we didn't find any problem with my server at all with the PHP redirect script that I'm using to pass that 7000 clicks mentioned above. So now the only reason left is to do with Airpush, I've found that different countries have different clicks discrepancy on Airpush. For instance, I've seen only 20% - 30% click discrepancy in France compared to Spain and Mexico that have almost up to 50% of the click loss.
This might be due to a few reasons as below:
1. Returning users clicking on the same ad, this leads to CPA networks filtering out duplicate IP/Clicks.
2. Publishers' performance in different countries vary
3. Limited demo size in that particular country so that a bunch of people are the same ones so it leads to problem number 1.
My apology to Tyler as I said it's definitely something to do with my server configuration which is not the case, kudos to Tyler as he's really awesome!
*I've heard that the new STM p202 Mod will be 4x faster for the next release!
Hi Wyatt,
As I thought it might be.. It's not your server at all. its the traffic source and network.
I ran a benchmark of 10,000 request to your server. (Roughly 4 seconds to make all the request)
[root@napalm ~]# ab -n 10000 -c 200 http://www.sensored.com/index.php
This is ApacheBench, Version 2.0.40-dev <$Revision: 1.146 $> apache-2.0
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Copyright 2006 The Apache Software Foundation, http://www.apache.org/
Benchmarking www.sensored.com (be patient)
Completed 1000 requests
Completed 2000 requests
Completed 3000 requests
Completed 4000 requests
Completed 5000 requests
Completed 6000 requests
Completed 7000 requests
Completed 8000 requests
Completed 9000 requests
Finished 10000 requests
Server Software: LiteSpeed
Server Hostname: www.sensored.com
Server Port: 80
Document Path: /index.php
Document Length: 0 bytes
Concurrency Level: 200
Time taken for tests: 2.572869 seconds
Complete requests: 10000
Failed requests: 0
Write errors: 0
Non-2xx responses: 10000
Total transferred: 2760000 bytes
HTML transferred: 0 bytes
Requests per second: 3886.71 [#/sec] (mean)
Time per request: 51.457 [ms] (mean)
Time per request: 0.257 [ms] (mean, across all concurrent requests)
Transfer rate: 1047.47 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 0 0 0.6 0 8
Processing: 32 49 25.3 44 1017
Waiting: 31 49 25.3 44 1016
Total: 32 49 25.8 44 1017
Percentage of the requests served within a certain time (ms)
50% 44
66% 47
75% 51
80% 54
90% 59
95% 65
98% 99
99% 174
100% 1017 (longest request)
Not a single dropped page or a redirect longer than 1 second, infact 99% of the redirects were 174 MS!
As I thought it might be.. It's not your server at all. its the traffic source and network.
I ran a benchmark of 10,000 request to your server. (Roughly 4 seconds to make all the request)
[root@napalm ~]# ab -n 10000 -c 200 http://www.sensored.com/index.php
This is ApacheBench, Version 2.0.40-dev <$Revision: 1.146 $> apache-2.0
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Copyright 2006 The Apache Software Foundation, http://www.apache.org/
Benchmarking www.sensored.com (be patient)
Completed 1000 requests
Completed 2000 requests
Completed 3000 requests
Completed 4000 requests
Completed 5000 requests
Completed 6000 requests
Completed 7000 requests
Completed 8000 requests
Completed 9000 requests
Finished 10000 requests
Server Software: LiteSpeed
Server Hostname: www.sensored.com
Server Port: 80
Document Path: /index.php
Document Length: 0 bytes
Concurrency Level: 200
Time taken for tests: 2.572869 seconds
Complete requests: 10000
Failed requests: 0
Write errors: 0
Non-2xx responses: 10000
Total transferred: 2760000 bytes
HTML transferred: 0 bytes
Requests per second: 3886.71 [#/sec] (mean)
Time per request: 51.457 [ms] (mean)
Time per request: 0.257 [ms] (mean, across all concurrent requests)
Transfer rate: 1047.47 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 0 0 0.6 0 8
Processing: 32 49 25.3 44 1017
Waiting: 31 49 25.3 44 1016
Total: 32 49 25.8 44 1017
Percentage of the requests served within a certain time (ms)
50% 44
66% 47
75% 51
80% 54
90% 59
95% 65
98% 99
99% 174
100% 1017 (longest request)
Not a single dropped page or a redirect longer than 1 second, infact 99% of the redirects were 174 MS!
Bawse status server support right there.
Originally Posted by wyffgoal
Each phone does not have a unique IP address. Carriers/operators have a pool of IPs that are shared by all of the phones on that network. So you could get hundreds of different phones coming through the same IP.
Using your example, the carriers in France that you are targeting might have a large pool of IPs, so you are seeing less duplicate IPs and "click loss" from them. Whereas in Mexico/Spain the carriers are routing their traffic through a smaller pool of IPs, leading to a high number of duplicates and "click loss".
I did similar tests several days ago. I tried STM tracker, Mobaff tracker, iMobitrax tracker, name.com 301 redirect, .htaccess 301 redirect, php 301 redirect and direct link. The traffic is cheap Indian clicks from Aripush, and the offer is from mobpartner. My server is BeyondHosting VPS2048.
Here is the data:
My conclusion:
1. only ~55% Indian clicks from Airpush can be tracked, no matter what tracking system you use (or direct link).
2. with any redirection, you will loss ~1/4 clicks further, no matter what redirection method you use.
Personally I think it's a problem of Push Notification. Normally I see ~25% missing clicks in my Airpush campaigns, and also ~20% missing clicks in Leadbolt. Maybe many users just close the browser as soon as possible after clicking the notification.
@augustye
Awesome test you have done there, thanks for posting the stats!
@augustye
Yeah good to see how we stack up, great test.
Like I've said before unless you can prove outright click fraud its better to just treat it as a "tax" on your clicks and a cost of doing business, there will always be a drop off from people who click and hit the back button before a proper load has occurred.
@augustye
Great test man!
I have the same results with push notifications , nothing you can do about it.
As for the comparison between trackers , you are just comparing imobitrax vs stm tracker.
The mobaff tracker is hosted on very fast servers ( cloud ) so can't really run a split test with hosting the stm tracker on a VPS
Great posts by wyffgoal and augustye!
@augustye
It's surprising to me that both STM and MobAff perform better than a simple PHP redirect. Or is that PHP script doing more than just that?
These tests are really awesome. So thank you guys for putting this together.
It sucks that no matter how fast the redirect, you lose 25% additional clicks. That is a significant number. I wonder what is causing that because a lot of these redirects can probably achieve around 100ms (MobAff especially). Perhaps 25% of the devices/connections are still behind and simply can't achieve that level of performance.
I think here the biggest problem is using push traffic.
A lot of the clicks are accidental or when users see it's not a real phone notification they close it right away.
So I wouldn't consider it click loss , just another type of traffic.
I think with banners from WAP loss should be smaller.
Awesome case study!
At the end of the day, I'll have to agree with Jordan that your roi is the ultimate thing that you have to look at.
It's sad to know that we pay at least 25% of our traffic for nothing, but we just have to deal with it, take it as a cost and work towards campaigns that can cover the costs. I believe there are plenty of them out there for sure.
I am glad that my data is useful for you guys
@cygnusx: php_301 is a simple php redirection, nothing more. In fact I will ignore the 2% difference. In this kind of tests, if the difference is less than 10%, I don't think you can tell anything for sure.
Well, the issue of 25% additional click loss when you put a tracker in between (or, rather, any type of additional redirect) is an issue. That seems to be real loss, and you have to decide if it's worth it to use a tracker vs. DL.
Originally Posted by ffclogin
Originally Posted by augustye
Here is the data:
My conclusion:
1. only ~55% Indian clicks from Airpush can be tracked, no matter what tracking system you use (or direct link).
2. with any redirection, you will loss ~1/4 clicks further, no matter what redirection method you use.
Personally I think it's a problem of Push Notification. Normally I see ~25% missing clicks in my Airpush campaigns, and also ~20% missing clicks in Leadbolt. Maybe many users just close the browser as soon as possible after clicking the notification.
Assuming this is a relatively controlled experiment, the test suggests that adding any kind of additional redirect on top of DL causes an additional 25% of loss (i.e. in addition to what you're already losing from DL).
But you're right, a test based on profitability would be more accurate.
Keep in mind one issue with click loss is due to slow speeds and users hitting the back button. Typically "impatient" users aren't the ones that were going to sign up in the first place so I completely agree with Tom looking at how it affects performance in terms of revenue generated would be a great test!
On a separate note from server tests, @Wyff Any new stats on the campaign, give it a good ol 2nd try?
@stackman
I will put something up for sure to split test the profitability with the tracker + without the tracker as mentioned in this thread. I'm moving my server to a well-optimized dedicated server with memcached and some other things configured for mobile. Need to move everything there then I will throw in some testing budget again by using your creative suggestions.
Stay tuned for the update.
All that volume for 1 cent clicks international?
I'm so there.
Home > Mobile >