How to optimize push notifications?

[Skymartins]

Hi guys,

I have recently started my first push notification campaign. AS it is normal the results are't really good, so I need to optimize it.

But the question is, how ? In push notifications there are no placements, no sites, etc... so what can you really optimize a part of the creatives?

Thanks

 

[Graybeard]

You have to have options in the advertiser's DSP. (demand-side platform)
If not you are just getting RON traffic and, and with no options, wasting money. (run-of-network)

When you do not have referrer transparency you will have problems optimizing.

 

[Skymartins]

Yeah, that's my point.

I'm using PropellerAds, and in Push Notification on cpc I don't see any option to customize it...

 

[marttoom]

I'm also running push on Propeller. The trick in my opinion is to choose wide appeal offers, where you don't need FB style targeting that much. Then I just let a few test campaigns run for a few days. I'm testing different creatives and images, this is pretty much all you can do there, but in most cases it's enough to get to profitability (if offer selection is good).

 

[Graybeard]

Look elsewhere -- if that is not there. Have you asked support there?
I am having to look hard for the right geofencing or even zip code (us postal) code optimization -- it's out there but not that common (all ad media).

 

[Japoman]

Hey
I optimize basically as if it were pop traffic in propeller.
the options are very few
you can not even separate mobile from tablet .....

 

[tyoussef]

Hi guys,

I have recently started my first push notification campaign. AS it is normal the results are't really good, so I need to optimize it.

But the question is, how ? In push notifications there are no placements, no sites, etc... so what can you really optimize a part of the creatives?

Thanks

not true , you can optimise based on placements and based on many other things.
just make sure you use a tracker so you can know what to blacklist .

 

[marttoom]

In push traffic case does the ZoneID correspond to the site where the user opted in to receive push notifications... or something else?

 

[Anthony.]

In push traffic case does the ZoneID correspond to the site where the user opted in to receive push notifications... or something else?

Yes, I think it corresponds to where they subscribed.
It's still good to block zones in your optimization because someone may of subscribed on a niche specific site, so when they see random notifications, they aren't interested.

A rough example I'd say is, if someone subscribed on a religious website, they don't wanna see dating/adult ads so that zone would be bad.

Depending what campaigns you run, not all zones will have good traffic for it in my opinion.

 

[marttoom]

That's a good question - I assume there might be some technical/privacy reasons, but at least partially it's definitely because they are aware how much crap sources there are.

 

[Skymartins]

not true , you can optimise based on placements and based on many other things.
just make sure you use a tracker so you can know what to blacklist .

In PropellerAds? You sure because I was not able to see the option...

 

[Graybeard]

I vote for transparency why are these networks hiding the real referrers?
I want to know what I am buying -- don't you?

That is just common sense ...

 

[Anthony.]

In PropellerAds? You sure because I was not able to see the option...

Yes you can it's in the campaign setup - change to advanced mode to see it all on 1 page.

I vote for transparency why are these networks hiding the real referrers?
I want to know what I am buying -- don't you?

That is just common sense ...

Push doesn't have referrers because it's direct from the users device where they click, but we do have zones where that user initially subscribed to which we can block if bad.

 

[awesomesauce]

Optimize by siteID, browser, OS, time of day.... same as pops.

 

[tyoussef]

In PropellerAds? You sure because I was not able to see the option...

i am sure, and @NeverGiveUp explained where to find that targeting options.

I vote for transparency why are these networks hiding the real referrers?
I want to know what I am buying -- don't you?
That is just common sense ...

i think if they do that, they will not be able to sell all of their traffic .
they want the buyer to get traffic from all the sites, then you can start blacklisting based on the results you get.
and if you know the referrers you will go the website owner and you will cut the middle man (Traffic source)

 

[Justyna Zeropark]

Here's a short Zeropark guide about push optimization. Most of the techniques (related to creatives, blacklisting, adjusting the bid) are pretty universal, so worth reading even if you're running with PropellerAds. Hope you can see some green once you play around the optimization.

 

[Anthony.]

Here's a short Zeropark guide about push optimization. Most of the techniques (related to creatives, blacklisting, adjusting the bid) are pretty universal, so worth reading even if you're running with PropellerAds. Hope you can see some green once you play around the optimization.

Great blog post, thanks!

 

[micoangelo]

Optimize by siteID, browser, OS, time of day.... same as pops.

Is it worth doing geolocation optimization, like by city, or is that getting too zoomed-in?

 

[RichAds]

At RichPush we have blacklists and whitelists including placements, and also we have Smart CPC which allows you to get more clicks for lower cost.

 

[Graybeard]

That's a good question - I assume there might be some technical/privacy reasons, but at least partially it's definitely because they are aware how much crap sources there are.

Close to the truth +1
Also what is called "geolocation" is often crap too. Cross-device location using GPS is becoming more commonplace for obvious reasons ... Why do you think all of these apps want your *location*?

 

[Mor]

I'm also running push on Propeller. The trick in my opinion is to choose wide appeal offers, where you don't need FB style targeting that much. Then I just let a few test campaigns run for a few days. I'm testing different creatives and images, this is pretty much all you can do there, but in most cases it's enough to get to profitability (if offer selection is good).

This is a really good approach. In the end the essence of push it's not targeted. You are directing ads towards people that subscribed, it's much more personal and you know nothing of the user. But what's awesome about the push that you don't need to, if you create a creative that is approaching and engaging you just don't need these details. What does it matter where he subscribed, once he subscribed and left the site he subscribed it has very little value.

 

[Mor]

Hi guys,

I have recently started my first push notification campaign. AS it is normal the results are't really good, so I need to optimize it.

But the question is, how? In push notifications there are no placements, no sites, etc... so what can you really optimize a part of the creatives?

Thanks

Well, it's not completely True, unlike pop, the site as really little to no meaning, since the user is subscribing from a site which then he might never be on again, so marking him as a streaming oriented user wouldn't be the exact targeting.

What you should do is work based on the results you are getting on your campaign, place a postback and once you have enough data start monetizing.

You do have sub id's in most networks i believe (we for instance work on sub id's so you could identify the source that got the job done)

From this initial optimization, you can change the bids, cut GEO's and pretty much do everything

Push is different in the regard that optimizing is quite new, because it's pretty much the only Ad format, aside from SEO and Adwords that doesn't have a real dependency on the domain

Hope you git green bro

 

[Tom - Version Two]

Talk to your AM at the Push Network, they should have some sort of breakdown for optimisations.

When I previously worked with them we could target by certain parameters such as vertical (casino, entertainment, finance, sports etc), as well as user sign up date to the push notifications. These combined can give you a small insight to the demographic that you may want to be targeting more specifically depending on your campaigns.

 

[Graybeard]

Is it worth doing geolocation optimization, like by city, or is that getting too zoomed-in?

There are many factors and myths in geotargeting;

Just what are you geotargeting -- are the city they are *locating* derived from the Maxmind GeoCity2 list (that list is primarily designed for 'Minifraud' detection) ? That city is the position of the person's IP and ISP Server (or what the ISP reports as routing in the Regional Internet registry Regional Internet registry - Wikipedia).

Modern cross device recognition, Mobile internet by GPS and by the location of the user's router by mobile GPS, is called geofencing.

The difference is maybe 72% accuracy by what you are getting with geotargeting cities or zip codes v. 92% accuracy with geofencing <<< those are the figures I am seeing bantered about.

This really matters when you are localizing traffic.

Location matters

mysql> SELECT CONCAT('Location ', zip, ' Income Index:') , `Income Rank` FROM zip_income_ranks GROUP BY `Income Rank` DESC Limit 50;
+---------------------------------------------+-------------+
| CONCAT('Location ', zip, ' Income Index:') | Income Rank |
+---------------------------------------------+-------------+
| Location 7311 Income Index: | 1053 |
| Location 32830 Income Index: | 1022 |
| Location 10282 Income Index: | 989 |
| Location 33109 Income Index: | 882 |
| Location 94304 Income Index: | 875 |
| Location 60654 Income Index: | 825 |
| Location 20004 Income Index: | 816 |
| Location 94123 Income Index: | 813 |
| Location 02210 Income Index: | 809 |
| Location 94105 Income Index: | 796 |
| Location 10280 Income Index: | 794 |
| Location 10162 Income Index: | 789 |
| Location 60661 Income Index: | 776 |
| Location 11109 Income Index: | 757 |
| Location 10069 Income Index: | 737 |
| Location 19436 Income Index: | 705 |
| Location 77010 Income Index: | 687 |
| Location 60606 Income Index: | 684 |
| Location 10005 Income Index: | 678 |
| Location 22213 Income Index: | 669 |
| Location 31411 Income Index: | 665 |
| Location 10007 Income Index: | 661 |
| Location 94158 Income Index: | 656 |
| Location 10170 Income Index: | 651 |
| Location 02109 Income Index: | 644 |
| Location 10022 Income Index: | 643 |
| Location 80238 Income Index: | 633 |
| Location 10597 Income Index: | 631 |
| Location 22301 Income Index: | 625 |
| Location 22308 Income Index: | 620 |
| Location 10075 Income Index: | 619 |
| Location 10021 Income Index: | 618 |
| Location 07030 Income Index: | 616 |
| Location 10006 Income Index: | 614 |
| Location 95134 Income Index: | 610 |
| Location 80547 Income Index: | 604 |
| Location 77441 Income Index: | 603 |
| Location 11765 Income Index: | 600 |
| Location 34228 Income Index: | 599 |
| Location 20148 Income Index: | 598 |
| Location 10004 Income Index: | 596 |
| Location 78739 Income Index: | 595 |
| Location 60601 Income Index: | 592 |
| Location 94028 Income Index: | 589 |
| Location 10065 Income Index: | 585 |
| Location 73151 Income Index: | 584 |
| Location 85263 Income Index: | 583 |
| Location 10028 Income Index: | 581 |
| Location 10514 Income Index: | 580 |
| Location 10010 Income Index: | 579 |
+---------------------------------------------+-------------+
50 rows in set, 1 warning (0.05 sec)

mysql> SELECT CONCAT('Location ', zip, ' Income Index:') , `Income Rank` FROM zip_income_ranks GROUP BY `Income Rank` ASC Limit 50;
+---------------------------------------------+-------------+
| CONCAT('Location ', zip, ' Income Index:') | Income Rank |
+---------------------------------------------+-------------+
| Location 28308 Income Index: | 53 |
| Location 22904 Income Index: | 58 |
| Location 80914 Income Index: | 59 |
| Location 23708 Income Index: | 60 |
| Location 96853 Income Index: | 62 |
| Location 82005 Income Index: | 63 |
| Location 19902 Income Index: | 64 |
| Location 31547 Income Index: | 66 |
| Location 23460 Income Index: | 68 |
| Location 96863 Income Index: | 71 |
| Location 92135 Income Index: | 72 |
| Location 28533 Income Index: | 73 |
| Location 23459 Income Index: | 74 |
| Location 98438 Income Index: | 75 |
| Location 41263 Income Index: | 76 |
| Location 37614 Income Index: | 77 |
| Location 32228 Income Index: | 78 |
| Location 92259 Income Index: | 79 |
| Location 13602 Income Index: | 81 |
| Location 23691 Income Index: | 83 |
| Location 22211 Income Index: | 84 |
| Location 96860 Income Index: | 85 |
| Location 59467 Income Index: | 86 |
| Location 31314 Income Index: | 87 |
| Location 20660 Income Index: | 88 |
| Location 31409 Income Index: | 89 |
| Location 72679 Income Index: | 90 |
| Location 62011 Income Index: | 92 |
| Location 13651 Income Index: | 93 |
| Location 32212 Income Index: | 94 |
| Location 38126 Income Index: | 95 |
| Location 30665 Income Index: | 96 |
| Location 58479 Income Index: | 98 |
| Location 23461 Income Index: | 99 |
| Location 31903 Income Index: | 100 |
| Location 12492 Income Index: | 101 |
| Location 14627 Income Index: | 102 |
| Location 21718 Income Index: | 103 |
| Location 36786 Income Index: | 104 |
| Location 36603 Income Index: | 105 |
| Location 38946 Income Index: | 106 |
| Location 24282 Income Index: | 107 |
| Location 38106 Income Index: | 108 |
| Location 39054 Income Index: | 109 |
| Location 29404 Income Index: | 110 |
| Location 14605 Income Index: | 111 |
| Location 33127 Income Index: | 112 |
| Location 32839 Income Index: | 113 |
| Location 25703 Income Index: | 114 |
| Location 30901 Income Index: | 115 |
+---------------------------------------------+-------------+
50 rows in set, 1 warning (0.04 sec)
*source: my algorithm applied to the IRS 2016 1040 AGI filings by zipcode. It is applied equally to each zipcode.

You are buying a bag of shit tied with a pink ribbon with *geolocation* as most of these ad networks are offering

 

[Mor]

Talk to your AM at the Push Network, they should have some sort of breakdown for optimisations.

When I previously worked with them we could target by certain parameters such as vertical (casino, entertainment, finance, sports etc), as well as user sign up date to the push notifications. These combined can give you a small insight to the demographic that you may want to be targeting more specifically depending on your campaigns.

In my opinion targeting by the site the user signed up on isn't much indication on push traffic, Since the user could have signed up at a file sharing site for instance and you got the user when he is on day 7 since he registred, do you remember on what site you were on a week ago? I think you should look at push on 2 params: Fresh and old and go from there.