Push Follow Along - Sweeps

[DoubleDouble]

HI Fixers,

I have been running a push campaign recently, wanted to share with you all the setup, how the results were as well as get any feedback from you in a "what would you do" scenario. (sorry if this isn't the best write up, I am no Shakespear!)

Offer: Iphone XS (I tested out 2 offers for this both offers were capped at 50 conversions per day)
Payout: 1.60
Landing Pages: 3 landers
Creative: 3 images 3 texts
Country: Tier2/3.
Spy tools used: SpyPush/Adplexity.
Tracker: Binom
CDN: Amazons3 + Cloudfront.
Traffic Source: Propeller ads

OK now we have the tech specs out the was time to go into the actual campaign details.

The creative that I started on the offer with was: I found this to be the longest running creative spying, the landing pages I had a bank of ones previously tested so I went with those.

Day 1.
Interestingly off the bat this campaign was profitable. With essentially every angle that I tested (aside from running direct to the offers) This hasn't happened to me so much before so needless to say I was very happy about the results. I started with a budget of $20 per day

Day 2.
Day 2, was silmilar to day 1, the offers was really smashing it. I increased the budget to $50 and the results stayed good. Note at this point I was cutting out placements that had 10 clicks or more with no conversions. Whilst this is a particularly strict rule I feel that on push if its going to convert then its going to do so pretty quickly By day 2, there were a couple of sources that were starting to shine. So I made a new campaign in propellor ads (a whitelist campaign) where I put a max bid and just ran on these sources.

Day 3,4&5.
Similar vein to the previous days, the offer remained converting well. I spoke to the affiliate network and managed to get the offer cap raised from 50 to 150. By this point one offer was outperforming the other as well as certain landers were performing much better. I had cut down to the better offer and the better lander to maximize profits.

By this point I wanted to try and scale this campaign, so i opened it on :
Megapush: big mistake, if you saw my other thread this burned through a budget and had an overspend of more than 150%
Zeropark: Didnt seem to convert so did not run much there.

What was very interesting was when they increased teh cap the conversion rate of the offer halved.

to this:

Not good! now here are teh questions for the forum:

a) is it possible for the offer to burnout like this, or was this the aff network shaving?
b) is it possible for the lander to burnout and should I have rotated the lander.
c) I did change the creative of the offer towards the last days to try and rejuvinate it but it didn't seem to make much difference.

what is everyones recommendations and how could this be improved?

Happy to share the information for everyone as more interested in the group learning and helping others.

 

[Graybeard]

Why would't someone that clicks a push alert reach the URL -> your landing page ?

 

[DoubleDouble]

Why would't someone that clicks a push alert reach the URL -> your landing page ?

Not sure what you mean, they will always reach the lp or offer

 

[Audiencia]

Thanks for sharing , trying to get understand push notification campaign optimization process , the question i got is since you're pushing on a segment of the same audience depending on the reach .

Can you exclude previously converted devices ID on propAds ?

 

[John aka]

Hello Doubledouble it good result hope you scale it up.
I just want ask you how much you bid when you start your campaign.
And which affiliate network ?

 

[Graybeard]

^^clicks to where?
Whose LP yours or the offer?

1006 |310 means you have a 69% drop-off on your LP CTR to the offer?

if you look at the server logs for Your landing page and compare the traffic metrics you may see the reason for the lower CR.
Lower quality traffic maybe or something to do with the day or time of day.

You are only going to be able to possibly deduce this info from your own server logs. If your landing page uses Google-Analytics that can help group results but no tool is going to detail your traffic metrics like the raw data from a server access log that you have set up for the metrics you want.

You could go as far as adding a custom script to your landing page to record the variables that you want.

 

[Adonno]

hello i have couple of questions in my mind
1. What type of landing page do you use (ex. wheel, mutliple questions, etc)
2. do you use cloaker?
3. is the location you running english speaking country?

thank you

 

[DoubleDouble]

Thanks for sharing , trying to get understand push notification campaign optimization process , the question i got is since you're pushing on a segment of the same audience depending on the reach .

Can you exclude previously converted devices ID on propAds ?

As far as I can research no, but you can set a frequency cap

 

[DoubleDouble]

hello i have couple of questions in my mind
1. What type of landing page do you use (ex. wheel, mutliple questions, etc)
2. do you use cloaker?
3. is the location you running english speaking country?

thank you

Mix, wheel or rrqrd type worked best
No cloaker
Could be

 

[DoubleDouble]

Hello Doubledouble it good result hope you scale it up.
I just want ask you how much you bid when you start your campaign.
And which affiliate network ?

Max bid, can't say the network as didn't find another with this offer, sorry

 

[DoubleDouble]

^^clicks to where?
Whose LP yours or the offer?

1006 |310 means you have a 69% drop-off on your LP CTR to the offer?

if you look at the server logs for Your landing page and compare the traffic metrics you may see the reason for the lower CR.
Lower quality traffic maybe or something to do with the day or time of day.

You are only going to be able to possibly deduce this info from your own server logs. If your landing page uses Google-Analytics that can help group results but no tool is going to detail your traffic metrics like the raw data from a server access log that you have set up for the metrics you want.

You could go as far as adding a custom script to your landing page to record the variables that you want.

Thanks Graybeard, my thoughts were that there is always going to be drop of and I viewed it as a bounce rate (correctly or not) 69% bounce whilst is high wouldn't be so crazy considering the nature of the landing page. If I look at the logs what would you look for to learn about the traffic? (thanks)

 

[Adonno]

Mix, wheel or rrqrd type worked best
No cloaker
Could be

what do you mean by rrqrd?
are you using CPM or CPC?

i tried CPM and got 1 convertion, iam trying CPCnow but traffic is slow nad still 0 conversion

 

[DoubleDouble]

what do you mean by rrqrd?
are you using CPM or CPC?

i tried CPM and got 1 convertion, iam trying CPCnow but traffic is slow nad still 0 conversion

sorry it was meant to say reward type. I was in the gym and fat fingered so my typing was not the best!
I used CPC, I did test out the campaign on CPM and got little/no results, it just did not seem to convert as well.

 

[Adonno]

Mix, wheel or rrqrd type worked best
No cloaker
Could be

sorry it was meant to say reward type. I was in the gym and fat fingered so my typing was not the best!
I used CPC, I did test out the campaign on CPM and got little/no results, it just did not seem to convert as well.

thank you i will test more heheh

 

[DoubleDouble]

If that 31% is your; CTR push=>Your (pre-lander) landing page => the offer's landing page (or sign up page) I think a 31 % pass through CTR is pretty good.

You could just use Google-Analytics on YOUR pages if SEO is not a concern on that domain.
To work custom analyze your logs you would want to use live commands over SSH or on a local terminal.

1. number of accesses by IP [by day and then by date range]
   
2. visitor GEO data
3. access time [if multiple pages; time between accesses]
4. visitor user-agent -- that can be broken down to o/s browser with version
5. I use a dummy javascript to attempt to detect who is a bot and what sort of bot they are [$ grep 'scriptame.js' access.log]
6. I can zgrep and zcat my compressed log archive and get the data from a time period lie 4 months ago or a time range. This depends in the size of the logs.

Then there is scripting;

1. screen size
2. cookie retention
3. many other factors

The point of all of this is to construct reasonable customer personas. Personas are aggregated profiles of customers that (might) convert. The problem is completing the loop with the actual data from the end seller i.e.; *the offer*
offer owners do not share this information so you cannot track the customer actions and the CR transaction on the offer owner's website. So with the affiliate model you can only do part of the job -- and shaving is always the 800 pound gorilla in the corner on the CR end.

That said, it is relatively easy to detect traffic padding or drop-off rates traffic source=> your pages using an access log. just use a server rewrite and take a count if you want to get the actual *hits [clicks]* delivered by a traffic source. page.html?s=5 will record on the log s[ource]=5[abc traffic network]

$ grep -c 's=5' access.log
1001
$ grep 's=5' access.log | do something with that list

$ grep 's=5' access.log | cut -d'"' -f1|sort -nr|uniq -c | less
Normally I would make a bash script .sh to do the following but this was just a simple test so I hacked it out

echo -n '"January,2019";' >>t.co-test && echo $(tac /home/user/x.___.com/logs/access.log |egrep   't.co/' |grep 'Jan/2018'|cut -d'-' -f1|uniq   |wc -l) >> t.co-test

Twitterbot growth pattern from a test feed ceased posting in September, 2018

"January,2019";179
"December,2018";200
"November,2018";213
"October,2018";185
"September,2018";170
"August,2018";135

unique v. raw hits over on month test
$ echo -n '"October,2018";' >>t.co-test && echo -n $(tac /home/user/x.___.com\
/logs/access.log |egrep   't.co/' |grep 'Oct/2018'|cut -d'-' -f1|uniq   |wc -l) \
>> t.co-test && echo  -n ';'$(tac /home/user/x.___.com/logs/access.log \
|egrep   't.co/' |grep 'Oct/2018'|cut -d'-' -f1   |wc -l) >> t.co-test

"October,2018";185;204

You could break this down by visitor IP and track IP changes of the same visitor by device possibly too
If someone arrives with a desktop and then a 4G you lose them cross-network usually but if the mobile is wireless the IP will be the same if they are using the same router public IP address routed by a NAT 192.168.1.100 as an example.

wow thanks for taking the time to write this out Graybeard! certainly a lot for me to work though here. The campaign stopped converting so I thought it better to take the profit and move on in this instance, I am setting up another campaign so will implement this and certainly follow up (shortly).

Re the CTR yes agreed, what was strange is that it just stopped converting when everything else remained constant. which led me to believe they were shaving. (unless offer lifetime really is that short on push). Either way some great learnings in this thread so far.

 

[Anthony.]

This happened to me to be honest.

Day 1 14 conversions
Day 2 12 Conversions
Day 3 total flop
(along those lines)

I wondered if it was offer burnout, lander saturation or creative burnout but I guess it's to do with the quality of traffic you first receive on propeller to assess your ad performance. (for propeller, the better the ctr - the more they make from you from sent out notifications)

Did you select "all" traffic to start?

 

[Graybeard]

If that 31% is your; CTR push=>Your (pre-lander) landing page => the offer's landing page (or sign up page) I think a 31 % pass through CTR is pretty good.

You could just use Google-Analytics on YOUR pages if SEO is not a concern on that domain.
To work custom analyze your logs you would want to use live commands over SSH or on a local terminal.

1. number of accesses by IP [by day and then by date range]
   
2. visitor GEO data
3. access time [if multiple pages; time between accesses]
4. visitor user-agent -- that can be broken down to o/s browser with version
5. I use a dummy javascript to attempt to detect who is a bot and what sort of bot they are [$ grep 'scriptame.js' access.log]
6. I can zgrep and zcat my compressed log archive and get the data from a time period lie 4 months ago or a time range. This depends in the size of the logs.

Then there is scripting;

1. screen size
2. cookie retention
3. many other factors

The point of all of this is to construct reasonable customer personas. Personas are aggregated profiles of customers that (might) convert. The problem is completing the loop with the actual data from the end seller i.e.; *the offer*
offer owners do not share this information so you cannot track the customer actions and the CR transaction on the offer owner's website. So with the affiliate model you can only do part of the job -- and shaving is always the 800 pound gorilla in the corner on the CR end.

That said, it is relatively easy to detect traffic padding or drop-off rates traffic source=> your pages using an access log. just use a server rewrite and take a count if you want to get the actual *hits [clicks]* delivered by a traffic source. page.html?s=5 will record on the log s[ource]=5[abc traffic network]

$ grep -c 's=5' access.log
1001
$ grep 's=5' access.log | do something with that list

$ grep 's=5' access.log | cut -d'"' -f1|sort -nr|uniq -c | less
Normally I would make a bash script .sh to do the following but this was just a simple test so I hacked it out

echo -n '"January,2019";' >>t.co-test && echo $(tac /home/user/x.___.com/logs/access.log |egrep   't.co/' |grep 'Jan/2018'|cut -d'-' -f1|uniq   |wc -l) >> t.co-test

Twitterbot growth pattern from a test feed ceased posting in September, 2018

"January,2019";179
"December,2018";200
"November,2018";213
"October,2018";185
"September,2018";170
"August,2018";135

unique v. raw hits over on month test
$ echo -n '"October,2018";' >>t.co-test && echo -n $(tac /home/user/x.___.com\
/logs/access.log |egrep   't.co/' |grep 'Oct/2018'|cut -d'-' -f1|uniq   |wc -l) \
>> t.co-test && echo  -n ';'$(tac /home/user/x.___.com/logs/access.log \
|egrep   't.co/' |grep 'Oct/2018'|cut -d'-' -f1   |wc -l) >> t.co-test

"October,2018";185;204

You could break this down by visitor IP and track IP changes of the same visitor by device possibly too
If someone arrives with a desktop and then a 4G you lose them cross-network usually but if the mobile is wireless the IP will be the same if they are using the same router public IP address routed by a NAT 192.168.1.100 as an example.

 

[DoubleDouble]

This happened to me to be honest.

Day 1 14 conversions
Day 2 12 Conversions
Day 3 total flop
(along those lines)

I wondered if it was offer burnout, lander saturation or creative burnout but I guess it's to do with the quality of traffic you first receive on propeller to assess your ad performance. (for propeller, the better the ctr - the more they make from you from sent out notifications)

Did you select "all" traffic to start?

yup all traffic and a high bid - I also tested the offer on CPM and there were no conversions at all, so not sure of the traffic quality in those instances, I would have the put some of Graybeards learnings to practice to see more.

 

[Audiencia]

As far as I can research no, but you can set a frequency cap

What about SUBID , even if the variable isn't listed near the target url params , it's working , i've seen a demo with someone using it on same traffic source for push notifs .

i'm setting same type of traffic there and now working on tracking mechanisms , i'll probably confirm that later if you don't before

 

[Graybeard]

the final arbitrator has to be the EPC in any affiliate campaign -truth

 

[Anthony.]

How's it performing now, any stats for yesterday?

 

[DoubleDouble]

How's it performing now, any stats for yesterday?

no i turned off the campaign as I wanted to protect the profit margin that I had. Would you recommend putting it back on!?

 

[ykshev]

I've noticed that on the first day on propellerads you get really high-quality traffic which you probably will not get on a second day and so on. Have you tried to play with high, medium and all traffic tabs on propellers? For example, when I start a campaign I use only high activity tab and you'll get not so much traffic, but it'll be really good quality. And overall, it's specific for web pushes to burn out really quickly. Have you tried other networks for this offer?

 

[ykshev]

And if you run it on a top bid it might be a case that you've got top quality traffic for a few days but then it ended on a network and they started to serve you all the traffic they have. (I'm not sure about propellers, but it definitely how it works on other networks)