Home > Mobile > Follow-along Campaigns

---

Something Different: a Beginners Pop Campaign (14)

---

Hi everyone.

I got my feet wet with the Mobile Cookbook Appetizer.....and after all signs pointing to pops; here I am.

I'm following @MrBrauns awesome pop guide/process which you can find here: http://stmforum.com/forum/showthread...ic-(Algorithm)

Aim: Split test offers until I find something that shows potential, then run that offer with different angles and landers to try and get into the green.


Campaign 1

• GEO: IQ
• Vertical: Phone Cleaner / Antivirus
• Traffic sources: Zeropark and Popads.net
• Offers: 3
• Landing Pages: 5 (ripped from Adplexity)
• Payout: $0.90 -$0.96
• Budget: $60 ($30 spent on each traffic source)

Campaign setup:

This is how I set up my campaign in Voluum. Im split testing the 3 offers against 5 landers I ripped from Adplexity:



Bidding/Traffic source business

• CPM - I went for $1 - in both traffic sources. I'm still unsure of how I set my initial bid
• Frequency- I did 1 view per 24 hours. Again - I just guessed

Results:

• Conversions from popads: 5
• Conversions from zeropark: 1

Interesting. It seems as if one offer is definitely performing better - no surprise since my affiliate manager has recommended this offer in the past.

One traffic source (Popads) also seems to be performing better than the other (Zeropark)......hmmmmmm

Here is a screenshot from Voluum from Popads. No point in showing Zeropark since it got only one conversion:





In regards to landers I dont think there is enough data to start speculating yet as there is not much data:





What Im doing next:

1) Since one of the offers is getting conversions I am going to run another campaign and spend $25 with Popads to just the one offer but rotate the 5 landing pages to see if any of the landing pages show potential. If so, I will create variations of that landing page and see where that gets me.

2) I'm finding it hard to find more offers in this GEO and vertical to test, so I'll go to a nearby country (e.g. Jordan) and run a similar test. I could also sign up for more affiliate networks to find more offers.

Update coming soon

---

VERY nice start!

And mrbraun is a fantastic teacher!

I'm sure he'll chime in with some suggestions later. In the meantime, here are some comments from me:

-Doing initial testing on both sources is a great idea! However, now that you've seen better results on popads, it won't be necessary to leave the other campaign running - you only need to spend money to cut offers and landers once, not on both sources.

-I almost always start with 1/24 frequency as well! As you increase the frequency, your ROI will probably go down. But the extra amount of traffic you'll get can possibly result in higher profits overall. But I wouldn't recommend testing this until your camp is green with high enough ROI to sustain the drop.

-A bid of $1 is fine for most tier 3 geos on most pop sources. If the traffic source shows an average bid, it would be good to bid slightly above average, just to make sure you're getting good quality traffic to do your offers justice.

-Do you know how to check offers for statistical significance? Here's the thread:

http://stmforum.com/forum/showthread...211#post289211

I would have checked that for you - but couldn't do so without knowing the payouts of all your offers. Unless the offers that haven't converted, had payouts that are a lot higher than the one that converted, the winner will be the one that got 5 conversions. But it would still be good to check and make sure.

-Have you drilled down to OS and carrier?

Sounds like a plan!

Just please be careful with one thing: When testing landers, be sure to finish one round of split-testing before starting another round, i.e. do not add new landers before you find a winner.

Once you add a split-test candidate, a new round of testing will start, and you won't be able to compare current/future data with past data.

Looking forward to seeing some results!



Amy

---

Hi Vortex,

Thanks for your advice.....its all making more sense now.

I've had some interesting results so here is the latest update:

Campaign Results:

• Cost: $25
• Revenue: $13.44
• ROI: -46%

Yes its negative but i'm assuming thats not too bad considering I've done no optimisation yet

Here's what I noticed from the data once I drilled down a little:

Landers - There seems to be a few getting good conversions



Carriers - Just 2 of the carriers are getting conversions



Operating system - again, just 2 of them are getting conversions. No prizes for guessing where they are coming from!



-----

Next steps:

I'm a little unsure of how to best optimize (i.e. what order to start in) however I found this post quite helpful (thanks again Vortex!):

http://stmforum.com/forum/showthread...Considerations

Here are my thoughts on where I think I should go from here.

Launch 3 more campaigns based on the data from this one:

• Campaign 1 - Cut the poor performing landers (+ added 4 new landers*) AND cut poor performing carriers
• Campaign 2 - Cut the poor performing landers AND cut poor performing carriers AND cut the poor performing OS's
• Campaign 3 - Same as campaign 2 but try a slightly higher bid (average bid + 20% as suggested by Mrbraun)

*Note: I got impatient and created lander variations based on one of the winning landers BEFORE the test was over (just like Vortex instructed NOT to do). This lander was winning at the time, but a few others caught up in the end. I made variations and tried changing one element per page e.g. CTA color)

I've got a few questions too which I can't seem to stop dwelling on:


1) Ive been spending $5 per lander for these tests and I pulled that number out of thin air because its nice and round. Is there a formula I should be following for how much to spend per test based on the number of landers being tested?

2) I've been running these test over 3-5 days but now I realised that in terms of day parting, i'm not seeing the full picture. Is it safe to say that its best to split the total campaign budget over 7 days so you can see which days perform better? I think this mistake has come from my lack of patience.

3) Since i'm a newb, im happy being transparent since I know the learning part of this is more important than the profits. Just a thought though: If I'm TOO transparent, could the campaign be negatively effected by people ripping and stealing the exact details?

---

yeah, but note that your main 2 carriers are the major carriers....the rest are not taking up much spend

Check out your OS though. One is at -17, and the other is at -67.

Problem is your LPs are nowhere near statistical significance to determine best yet.

I would kill the shitty OS, and just target the -17% ROI OS, and continue sending traffic, until a LP gets more significant statistics.

What affiliate network are you working with? Have you got their ok for the LPs? Some like to fuck people for noncompliance.

---

Usually there are some rules of thumb like spend 3x payout for the combinations of landers and offers.

You need definitely more data if you want to day part, but most of the time day parting is not really needed. You can check on that date once you got at least those 7 days, but it will be still not enough, you would need weeks of data to have it significant. Or it should really stand out.

Don't worry about this. They are ripping each other anyway same as you ripped the landers from the spy tool ;-) Focus on the process of learning and understanding the basics at this stage ;-)

---

Are these stats your overall stats? Or just for the winning offer?

Now that you have a winning offer, it wouldn't be a bad idea to cut landers based on stats for JUST that offer.

The "carrier" stats you attached - are they for "ISP/Carrier" or "Mobile Carrier"? If the former, then it would be nice to see the latter too.

The difference in CR for Android vs. IOS has potential for one of them to be cut to increase your overall ROI by quite a bit. However - it goes back to my question on whether the stats you've shown were overall stats or just for the winning offer. The inferior OS may be in great loss overall, but may still be converting OK for the winning offer.

You can choose to cut the inferior OS at this point in order to test more cheaply. However, keep in mind that by doing so, 1)when you mass-test offers later you won't be able to see which offers will convert well for the eliminated OS, and 2)your testing will go slower after you cut that OS. The choice is yours.


I'm confused - are you saying you want to start these 3 camps at the same time?

If so, why would you have 2 camps cutting the same landers? That would be wasting twice the money on cutting the same things.

Thedav has pointed out that the "non-converting carriers" have not received enough traffic to indicate whether they'll end up converting better or worse than the first 2 (that have converted). So it probably wouldn't be a good idea to cut them right now.

I'm assuming you've already cut down to the best offer (for now). The next step would be to cut down to a winning lander.

If those 3 landers keep going head-to-head with around the same number of conversions, just pick one to be the winner, and rotate another batch of landers in to start a new split-test.

Or, instead of testing more lander, mass-test offers!

If you want to test bids - you can do that after one of the above split-tests is over, and before you start another split-test.


LOL! Nothing like getting a slap on the wrist by your actual stats! Yup that's what every affiliate marketer is guilty of doing one time or another, myself included although I certainly should know better.


Basically I would suggest to keep the camp running, and to cut landers as they reach statistical significance, for as long as the camp is still showing some promise (i.e. losses aren't very great).

I personally almost never set campaign budgets for pop camps, but I know a lot of people do and that it helps them. Caurmen is planning on writing a post on how to set budgets for campaigns - so stay tuned!


I wouldn't worry about that for now. Right now you're cutting stuff all over the place, so your hour-to-hour and day-to-day stats will not exhibit the normal trends anyways.


I'm with Erik on this. Also, you don't necessarily have to reveal all the details. e.g. We don't need to know which specific offers you're running. As for landers, all of them are available on Adplexity for ripping anyways. And stuff like carriers and OSs are not worth keeping secret because their performance will depend on the offer and maybe even the lander, so it won't be like people would all start targeting a specific carrier because it intrinsically converts better.



Amy

---

Thanks for your advice Thedav!

Ok that makes sense but what’s the best approach here:

(a) Cut bad OS from original campaign and let it keep running? Or;

(b) Leave original campaign to gather more data, make a clone of original campaign and cut bad OS from clone.

I'm leaning towards (b) because I have an assumption (and I know assumptions are bad in AM) that editing a campaign that is running may fuck with results....am I totally off here?

Initially, no but i've now just asked my AM for approval. Thanks for the advice!

---

Hi Erik,

Thanks for your answers. That makes total sense – focus on learning now as mastering the process is more important than anything.

---

@Vortex - I'm so appreciative you took the time to write such a detailed response - thank you!

I'll be more specific because It looks like I've left out lots of critical info...

Sorry I should have been more specific. These stats are for the winning offer. I cut the shitty offers and run the winning offer to 5 landers.

---

That was ISP/carrier. Here are the mobile carrier results. Doesn’t reveal much to my newbie eyes:

---

Yep – As I mentioned above this is just for the winning offer. The ROI shows a big difference between the two OS’s – but i'm not sure if its statistically significant or not

---

Yes I can see why that's a mistake now. My reasoning behind the 2 camps cutting the same landers was to split test a higher bid vs the normal bid. But I can do that later as you have mentioned. I just went a little gung-ho on all the possible tests. So much to test. One step at a time.

---

Ok – this makes sense about cutting the bad landers from the original campaign while it is still running. One thing I was super confused about here however, was if I should be cutting other things from the original campaign such as bad OS’s, ISP and mobile carriers. I thought by doing so that I would skew the data, and this is the reason I suggested launching all those new campaigns – which were clones of the original but more targeted. It didn’t matter anyway because when I launched those campaigns they were getting almost no traffic at all – I guess the targeting was too narrow?

---

Issues with Targeting on Popads


I've been pulling my hair out over this one. When I go to narrow down a campaign in the Popads.net targeting tool, I'm finding its 'traffic estimator tool' often tells me I'll get less than 1000 impressions per day, even with the number one bidding spot.

I believe I figured out the reason for this - connection targeting. When I whitelist ALL connection types and speeds (see screenshot below) I get traffic, but when I choose cellular and cellular/carrier respectively, I get very little if any traffic. The offer has no restriction on the carriers it accepts, it just says 'NO incent. No co-reg. No adult.'

I'm thinking this is just a case of running offers in a small-ish GEO, right? Is it a case of adding more traffic sources?

---

Next Steps

I'm planning to run the following campaign moving forward. If you think i'm on the right track please let me know!

• Winning offer
• Cut bad landers
• Broad targeting

Once that's reach significance on the landers i.e. we (hopefully) have a winning lander, I'll keep the campaign running and start optimizing in the following order, which comes from a post by MrBraun:

1 - Optimize lander elements
2 - Cut bad devices
3 - Cut bad OS's and versions
4 - Cut bad browsers
5 - Cut bad carriers
6 - Black or white list the favourable or unfavourable sites sending traffic

---

You're right - changing campaign conditions while you're in the middle of a split-test (e.g. lander test or offer test) can skew your results. It depends on how accurate you want your results to be. Goes back to the accuracy vs. efficiency concept I find myself bringing up all the time - basically these are the two ends of a spectrum. You can choose to be more accurate but less efficient (i.e. having to spend more money and time to collect more data), or you can be more efficient but less accurate (i.e. save money and time but sacrifice on accuracy). And we're constantly having to find a balance between the two when optimizing a campaign.

I wouldn't clone the orig camp and cut IOS, because I really don't see any benefits from doing that (please let me know if you do!) You're testing landers right now, right? Here are some of your options:

1)Just cut IOS now, and still keep cutting landers based on overall lander stats, i.e. across all OSs.

2)Cut IOS now, and cut landers based on lander stats for JUST android alone. (So drill down to OS > Landers and expand stats for android.) If you've already started cutting landers before this point based on overall lander stats, switching to this now can skew results.

3)Leave IOS in and continue cutting landers based on overall lander stats.

Another consideration is that if you cut out IOS now, then your winning lander will be the winner for android and not necessarily IOS. Whereas if you include IOS in your lander testing you'd get a lander that will work for both OSs. It really depends on how granular you want to make your testing. Accuracy is great, but remember you'd be paying for it by sacrificing efficiency - which is something you may or may not want.

Also remember that testing landers is just the first step. Once you have a winning lander, you'll want to mass-test offers with it - at which time you may very well find an offer that will convert well for IOS. So in that sense, you may want to leave IOS running so that your winning lander will perform well overall for both OSs.

Or, if you want to go granular, you could cut landers for EACH major OS separately, so you'd have a winning lander for IOS and one for Android (which may or may not end up being the same lander). Then, when you get to mass-testing offers, you can set up a camp for Android and a camp for IOS, and use the corresponding winning lander for each.

There are so many ways to optimize a camp, and I'm not wanting to confuse - just listing a few options.



One important note here: Many people are confused on how to cut traffic segments - carriers, OSs, browser etc. And some will attempt to use the peakconversion split-test calculator to check for statistical significance - which is the wrong method.

Stuff like offers and landers need to be split-tested, because we can only show ONE lander and ONE offer to EACH visitor. So, we'd want to know which would be the best.

For traffic segments - because we can target ALL of them if we wish, we don't need to split-test to find out which one is best - we simply want to target ALL segments that are PROFITABLE.

Therefore, the best approach would be to wait until you've found the best offer+lander, and maybe even cut a few of the worst placements, and THEN take a look at traffic segments and cut them then, and only if they're still not profitable.

Even when you're still testing offers and landers, you could still cut traffic segments to save on test budget. You can retest those segments later, after you've found a good offer+lander. Drawback: Your best offer+lander may not be the best for the traffic segments you've left out of the split-testing.

Again - efficiency vs. accuracy, and the choice is yours to make. There's no right or wrong here.


Absolutely correct! The small volumes are due to your having specified cellular and cellular/carrier. This will target only carrier/3g/4g traffic, which is when visitors are connecting via their data plan and not wifi.

There's no reason for you to do this for your particular offer, because judging from your screenshot, all of your conversions thus far are from wifi. BTW that's what the "Other" category is when you drill down into "Mobile Carrier".


Sounds great!


I would actually suggest that you take your winning lander and mass-test offers. It's usually during that stage that you can get the biggest boost in ROI. And it's so much more satisfying to optimize and scale on a good offer instead of just an OK one.

But that would depend on what you want. Things could go either way - by testing more offers there's also the possibility of not finding a better offer than your current one.

Either way you choose, I look forward to seeing more stats soon!



Amy

---

Thanks Vortex!

I’ll admit I’m confused and overwhelmed but it IS making sense – especially the part about the trade off between efficiency and accuracy. I’ve decided that rather than launching new campaigns I’m going to keep a diary of what I have done and when I have done it. E.g. If I add more funds to a campaign and cut and operating system, I’ll note down the day I did it so when viewing the Voluum stats I can filter for those days and see the results…if that makes sense....


What we did with the Popads campaign:


Ran it broad and cut the poor performing landers.


What happened:

Not much, really. We didn’t get much traffic or conversions at all. It seems like its hard to get traffic with this traffic source right now where it was a little easier a week ago.




New Traffic Source

We launched the campaign on a new traffic source and got tonnes more traffic and conversions. Looks like one lander is getting most of the conversions again so the plan is to make variations of this lander and keep running traffic to see if we can get better results.



Lander stats for new traffic source:




Questions

1) Based on my lander stats do you think its worth creating small variation changes on it and split testing those elements (e.g. CTA colors etc)....or.....make large drastic changes e.g. test new/different angles.

---

2) Ok - this makes total sense but looking at the above stats would you call that a winner and go and start testing offers? Or would you do as I have suggested in the question above and try to improve upon it?

Thanks!

---

OK so here’s an update. I made a few changes:

1. Cut the landers from the last round that were not doing so well and left just one
2. Cut the poor performing OS out (note: there were only 2 main OS's sending traffic)
3. Blacklisted one of the referrers that was sucking budget and giving no conversions
4. Spent $20 more dollars.


Results: The ROI has improved to negative -47% but It looks like I need some large ‘wins’ to get it into the green.



Based on the traffic source (targeting options are relatively limited) ….and….to my (limited) knowledge, these are my options to improve the performance of this campaign:

1. Come up with new angles and test more landers. *
2. Ask AM for a Pay bump from AM **
3. Cut a few more placements ***
4. Try staggering my bids
5. Test other traffic sources (I’ve only tested 3 pop traffic sources so far)
6. Find the same offer on other Affiliate Networks and see how it converts


* I need some big improvements to get the campaign positive. Therefore I believe that number 1, and maybe numbers 5 and 6, are what I should focus on – the 80/20.

** In regards to asking my AM for a pay bump – yes this will help but what type of a % increase could I realistically expect? I’m assuming something like a 50%+ pay bump would be quite rare

*** In regards to blacklisting more placements, I initially thought about cutting the following placements. I chose them because they are all getting visits but not many clicks and no conversions:



Then I realised that my criteria for cutting those placements was pulled from my arse. My criteria: anything with more than 100 visits with no conversions. Looking back on this I think it’s a mistake. And it won’t deliver any drastic changes anyway. I can always re-visit this later.


Next steps

Test more landers.

I’m going to brainstorm some new angles and rip a few more landers from other GEO’s in this vertical.

My current CR is 1.07% and that’s with a ripped lander – no changes. I don’t know what the typical CR is for this this vertical with pops but If I can get to 2% I’ll break even and if I can get to 3% I’ll be at 50% + ROI

Thanks for following. Suggestions and criticism welcome

---

Hey there! Sorry for having been MIA - just got back from AWA and recovered after sleeping for almost 2 whole days.

As you've so correctly observed, you'll need for something to improve in a drastic way in order for your camp to get to green. I would suggest testing every offer you can find that's relevant for your lander, for the same geo.

A few more comments:

-Regarding the lander stats in your screenshot: Do you know the correct process for cutting landers? Here it is for your reference:

http://stmforum.com/forum/showthread...Banners-Part-1

Basically you would have needed to run landers 3 and 4 some more until you get a winner. The rest were ready to be cut.

-Testing lander variations would be another good thing to do, but I would definitely mass-test offers first, as that tends to have better potential of increasing ROI by larger degrees.

-Testing bids is important also, but try not to do that while you're using the camp to split-test offers/landers to avoid introducing too many moving parts. Maybe wait until you've finished a round of split-testing (of offers or landers).

-You're right about cutting placements not being too much of a value-added task in this case. Given your payout is $0.96, the only placement I would cut is the one that has spent $2.10 - I like to cut placements that are in loss by 2 payouts or more. This is not always a good approach though, for example if you're running in a geo where there's a ton of traffic, in which case you may want to cut more aggressively.

-Also - have you drilled down to major traffic segments yet? Stuff like OS and mobile carrier?


Again, I would focus on testing more offers first, and everything else second.



Amy

---

Thanks Amy,

This GEO is a little small so finding more offers to test may be tough so I think I'm going to have to apply to many more affiliate networks. I could always take this lander and go into larger GEO's too.

I'll report back when I get a little closer!

---

Home > Mobile > Follow-along Campaigns