Home > Mobile > Follow-along Campaigns

---

Mobile Pops (21)

---

Hi,

I've been doing affiliate marketing for a while, mainly SEO, but I'm now looking to get into pops.

What I've done so far:

• Bought Voluum
• Bought Adplexity
• Set up amazon S3 + Cloudflare DNS and Cloudfront
• Deposited 1250$ in Propeller Ads for testing
• Registered with a handful networks
• Bought domains for tracking, for landers etc.

I'm looking to run sweeps and AV in tier3 geos. ATM I'm more concerned with building a system, which is why I chose this vs. some more expensive stuff, to dial in the tracking and getting the system up.

My plan is atm to rip 5 different adplexity landers, clean them and add any missing scripts, then run 3 bids. Rotate 2-3 similar offers. Then use that as a basis for testing additional offers.

Not sure about how much to set the bids yet, 2x recommended, recommended and one as low as possible? Run some spend and see ROI, then optimize from whichever bid has the best.

Qs:

• Should I daypart straight away, based, I've seen some of you guys do that, to quickly get into green
• Bidding, what do you recommend? Is there any "strategy" or rule of thumb for initial bids.
• I'm looking at running sweeps in either VN, TH or possibly KE/GA. The benefit of the latter is that they're English speaking, at least as an official language (not sure about GA, but KE for sure), so I could use landers from ZA or other english speaking Geos. What do you guys recommend for a starter geo?

Making this journey for accountability, and hopefully in the future it can be a resource for beginners. Long term plans are to use the knowledge gained here to scale into t2 countries, and even other sources (mobile display/Facebook).


Anyone based in Barcelona? Anyone looking to bounce some ideas on pops/mobile?

---

Hi Jesse and thanks for starting this follow-along! Sweeps in tier 3 geos sounds like a plan - sweeps are still hot and you'll find no shortage of offers for sure.

Picking bids is kind of like playing roulettes - it's hard to tell which bids will give you the highest ROI. One thing I'm sure about though, is that testing 3 bids will give you a better chance than testing a single bid. If you don't have your own bid-picking strategy yet, I would suggest to check propeller's estimates graph, and test either ends of the range plus a middle bid - let's use this as an example:



I would test any 3 of the 4 values marked with arrows, or even test all 4 values. If you feel that the max value is so high that it scares you, test the other 3. Basically I like to test bids right after big jumps in traffic volume. Please remember to set low budgets (daily and total) to curb potential losses (especially for the higher bid camps).


If you're getting enough traffic to test things, then dayparting from the start would be a good idea, as it will likely allow you to test for cheaper. Typically the best-converting hours will be between after-work and bedtime - so around 5pm-12am. If you need more traffic then try 9am-12am (start of workday to bedtime).


As long as they're low-competition geos you should be fine. Whichever geo you choose now, you'll probably go on to test multiple other geos anyways.

For traffic volume VN and TH are both great. For less competition KE sounds good. What's GA? Gabon? I've never run there - with only 1.672 million population there's not much traffic there, but if you can find a good offer then go for it! These little geos can make money silently for you for a long time by flying under the radar due to having practically no competition.

If you want to find other geos that speak English - this will help:

http://www.infoplease.com/ipa/A0855611.html

Looking forward to seeing stats from your first campaign!



Amy

---

Thanks a lot for the feedback -- your replies in other posts have taught me a lot of the basics on pops and mobile!

What's the source on that bid map? Is it from Propeller?

GA was actually Ghana, but that's probably GH? But I might be looking at some smaller geos in Africa too, especially with the mobile vs. dekstop users there I think it's really interesting!

I'm waiting on my AMs to get back to me with some offers. I hope to start running tomorrow or the day after (I also work on Aff SEO fulltime).

---

Thanks for the nice comment! Glad my posts have helped.

And yes that screenshot was from propeller (that's where you're running right?)

Ah! Ghana - yes the 2-letter code is GH. I run in so many geos I get confused all the time as well. This here is one of my best friends:

http://www.worldatlas.com/aatlas/ctycodes.htm

African geos have a LOT of potential. Many visitors are a lot less tech-savvy than those from tier 1 geos, and competition is a lot lower.

As for cloaking - that's really outside my area of expertise. Perhaps start a new thread on the topic and see if you can get some feedback from cloaking veterans?



Amy

---

Sounds like a good plan! sweeps are still doing awesome numbers!
A wise decision to start in those GEOs.
Good luck!

---

Grabbed 5 landers from adplexity, recoded, optimized (added images as base64/sprites, inlined css, removed redundant stuff, removed leech scripts and backdoors), minified html and JS. Submitted for approval.

Set up a s3 bucket to test the landers on a few different mobiles.

Going a bit slower than expected.

Edit: ok so 5 landers clean and set up on cloudfront behind s3. Load time about .3 seconds to the cloudfront distro, give or take depending on location.

---

Following your follow along jessejames!
Im from Barcelona too

---

Awesome! I'll have some updates here soon!

---

EDIT: I made a mistake about the order in which Voluum evaluates rules. I've set up testing to confirm that Voluum will actually evaluate the TOP rule FIRST, and work its way down to the last rule. Any traffic that doesn't match any of the rules will go to the default path. Apologies about the mistake!

>default
>android
>mobifone/vinaphone all
>mobifone/vinaphone minus iphone

Would have been nice to see the actual rules. But based on what I see, this is what should be happening:

1)Traffic will get evaluated with the bottom rule, and all non-iphone traffic from mobifone and vinaphone will get redirected, leaving the rest to go onto the next rule.

2)At the rule "mobifone/vinaphone all", iphone traffic from mobifone and vinaphone will get redirected.

3)At the rule "android", android traffic from all carriers other than mobifone and vinaphone will get redirected. (Note here though that android traffic from mobifone and vinaphone will have gotten redirected at the bottom rule before reaching this point. Is this what you want?)

4)Lastly, the remaining traffic - which will be traffic that is NOT android and NOT from mobifone or vinaphone - will end up going to the default path.


Could you double-check the offers you've specified for each rule to see whether they're getting the right traffic as I've laid out above? If you're not sure, I would gladly design some rules for you - if you'd list the OSs and carriers each offer will accept.



Amy

---

Thank you, that should be how it's set up -- but I pm'd you with the overview of the offers etc.

---

That's normal. What's the percentage?


Need more data. Run at LEAST $10 more to the bid2 camp, then check stats again.

Would also be great to see OS stats.


Ideally yes, but basically when you're targeting both wifi and carrier, most of the traffic you're getting is for wifi anyways so you can treat this as a wifi camp. Carrier traffic is expensive, so unless you're running offers that are DESIGNED to convert well for carrier traffic (such as 1-click carrier-billing offers), it may not be worth bidding really high to get carrier traffic. I would suggest to just keep targeting wifi and carrier like you're doing, for now. If you see that an offer converts well for carrier, you can always start a separate camp to target carrier traffic and bid higher to get more volume.


WAY too early to be starting a whitelist - you don't have enough conversions.

I would highly suggest to not focus on cutting placements right now. Focus on finding the best offer + lander + targeting first. You're not in great loss, so cutting placements can be done later.



Good plan!

When drilling down into stats, see if you can spot trends as to whether a particular offer is converting the best for a specific OS, or if a lander is converting especially well for an offer or vice versa.

We can keep things really simple by just cutting offers and landers separately, i.e. cut offers based on offer stats alone, and cut landers based on lander stats alone. This would be the more efficient way, but less accurate, because by doing this we'd be assuming that the best lander for Offer1 would also be the best lander for all other offers.

If you want to be more accurate at the price of being less efficient, you can collect enough data so that there's an obvious winning offer, and then drill down to lander stats for JUST that offer and cut landers based on those. You would then also identify the profitable/promising traffic segments for that winning offer.

Could you please post these stats and we can try to look for some trends:

->Offers
->Landers
->Offers -> Landers
->Offers -> OS
->Offers -> Landers -> OS
->Offers -> Mobile Carriers

Let's start with those. I'll go answer your PM shortly...



Amy

---

Thanks for including all those screenshots!

Since you're only running wifi and android, "carrier" and "os" stats aren't as important.

One thing to note about Offer -> Landers stats:

Based on your screenshots - looks like you're cutting things correctly!

Just to verify - you did use this tool to cut offers right?

https://win-vector.shinyapps.io/CampaignPlanner_v3/

I was just chatting with matuloo the other day and agree with him that it may be better to run offer and lander stats BEYOND statistical significance is reached - just to make double-sure we're identifying the best offer and lander. Just a "for future reference" kind of suggestion.

Cutting placements now isn't a bad thing to do, as long as you keep in mind that there's a chance they may be profitable when you're running just your most-profitable lander+offer. So when your camp is green you may want to retest some of the placements you've cut in the past. A placement can only be as good as the lander+offer+targeting you're running.


That's definitely on the high side. Are you hosting your landers on a CDN? Other than making landers load faster, I don't know of any other way to curb clickloss. Unfortunately that's a cost we just have to absorb.


Testing more offers is ALWAYS good! The more the merrier.

Why would you want to make a separate campaign for the lander with the backbutton? That's not necessary at all. If you're split-testing landers, they need to be rotated in the same camp such as they're subject to the same set of conditions.

Regarding testing and cutting by offer+lander permutations: Yes you can do that. This would give more accurate results, as it would take into account of how different landers have different effects on different offers. However, that would also take a bigger test budget, which is the main reason why I usually don't suggest this to new people that are usually working with small starting budgets.

A cheaper but less accurate way, would be to identify the most profitable offer, and then just cut landers based on lander stats for JUST that offer.

An even cheaper but even less accurate way, would be to just cut landers and offers separately (i.e. cut landers based on lander stats alone, and cut offers based on offer stats alone). This is the most straight-forward way and requires the least test budget of the 3 approaches. And, if you only test offers that are very similar to one another, this approach should yield accurate-enough results.

When you have a proven lander, then you can just mass-test offers. To me a lander isn't proven until I've tested all the different landers from adplexity and identified it as the winner.

Having a testing and optimization and scaling strategy/process is very beneficial, and almost necessary when it comes to succeeding with pop. This is because pop camps will typically not last very long, so long-term profits will hinge on your ability to continually find new profitable camps to replace dying camps. Therefore the speed of execution is crucial, and having a system will help immensely.




Amy

---

Thanks for the in-depth reply. I use an excel sheet for cutting (bayesian), but it's the same concept (because then I can copy paste from Voluum and retain the data). For the placements, I tried to gauge relative CTR on top vs. low performing placements, and spend vs income per placement. Some of them had a high spend but no conversions. Bot traffic seems low, around 10% on those placements that contribute significant traffic, so no issues there.

The idea of lander/offer combinations is interesting, I'm going to keep that in mind. I think I have a good setup for that by making a macro in Excel that can do all the permutations between two sets (i.e. landers and offers), and then pull the data from Voluum and add in.

For now all my offers are pretty similar in the sense that they're all AV pin submits -- but I'm definitely going to look into it.

I host my landers on Cloudfront and use fast DNS, the main hog seems to be Voluum. Voluum use AWS I think, so the closest data center would be in Singapore (or Korea? Not sure which one), but when looking at Pingdom/GTmetrix the main contributor for load times is Voluum. (When I get to proper scale I'm going to run php+server logs only like someone posted in the scripts section .. but that's a down the line thing)

But this actually made me thinking: traffic that quickly bounces, do we really want it? Could be a lot of bots, people with slow connections etc. So even if I lose 30%, am I losing 30% of the lower end traffic on average? in the same sense that cheap traffic converts worse. I mean, so a 30% loss in traffic may not result in a 30% loss in potential income, but rather a much lower cost, comparable to other low marginal "costs of doing business" like CDN, tracker credits etc. Just a thought I had -- could be way off, of course.

I'm hoping to launch two new Geos today, I just want to get a good process for launching campaigns quickly -- which I believe is mission critical.

---

Your excel with macros that can pull all the necessary data and analyze it - sounds badass! That kind of automation can save quite a bit of time.

Speculation or not - it's a very positive way of looking at things! Since it's something you probably can't do much about, coming to terms with it is probably the best thing you can do.



The way you launched your current camp sounded great actually! Are there parts of the process you still have doubts about?

The initial testing is mainly to use proven offers (i.e. AM-recommended offers that are proven to convert by multiple affiliates) to test landers, to find one that's at least decent.

The real fun begins when you have a winning lander, which you can then use to mass-test offers.

That's usually when you can uncover a real gem (or 2 or more if you're lucky).

And when you test several geos in this fashion, that will dramatically increase your chances of finding a profitable offer+lander+targeting combo.



Eager to see how your new geos do! Also, for your original camp (VN AV), once you're down to a winning offer and lander, you may want to either test more landers (since you only tested 5 the first time), or start mass-testing every VN AV offer you can find on all the aff networks you're a member of.



Amy

---

As I've told the OP in a PM: I made a mistake regarding how Voluum evaluates rules - please see the "EDIT" in the beginning of this post:

http://stmforum.com/forum/showthread...l=1#post285487

All the logic is still correct - except the ORDER of the rules should be REVERSED. I've set up an actual test to confirm this.

I think I've been lucky that I never had rules set up where the order mattered (i.e. when rules criteria are mutually exclusive / don't overlap). Well now I know!

Many apologies for the wrong information!



Amy

---

New campaign started

6 offers, 6 landers
3 bids to test, 7$ budget each, 10$ budget for the highest
CPM 2, 2.5 and 4$

Will update with some stats later. I'll let all teh bid campaigns run out their budget, and stick with the best ROI one.

---

Which traffic source is this again? Just wondering if you're bidding appropriately.


Amy

---

Propeller, here's the bidding landscape (not that different from when I made my bids). https://www.dropbox.com/s/66iun5s7ru...50.01.png?dl=0

It does show "Warning: Your bid is too low, you may not receive any traffic at all. Please increase the bid", but I'm considering the payout of the offer vs. the bids too.

I have to open accounts with PopAds and PopCash too, but I deposited on propeller because they had a promotion in Aug.

---

I would also include a bid around $6 or $8 - but if you're not comfortable bidding so high then let's just see how your bids will do.

I don't feel anything when I see $10+ CPM bids on propeller, because I'm so used to running carrier traffic which typically costs quite a bit higher. (Think $30+ or even $50+ CPM...)

That also means I've lost my "sense" of what would be appropriate to bid for wifi. So your guess would be as good as mine. Let's see how your bids do.



Amy

---

First, I understand, you've got a much stronger intuitive feeling for testing camps than a beginner such as myself, but I wonder how this may fit into the context of initial-test camps. If I am trying mass-test say 10 ~$1 payout content offers against each other (seems like others are direct linking these based on my spying...for whatever thats worth), and i see something like this -- maybe a little less dramatic.

I'm looking at something now that would suggest the arrows would be at $1.5 cpm (900 impressions/day), $4.75 cpm (2400 imp/day), then a big jump to $17 cpm (5000 impressions) -- maxing out at $35 (7k imp). I guess this means that someone is making some good money on a few big placements. My "rules of thumb" tell me ~$1*10*(3->5) --> $30-$50 to see if these offers convert.

If I tried $17 bid, is it even possible to get enough volume to make statistically significant data (thats like 200 impressions an offer for the whole initial test -- not even taking click-loss into consideration). Or perhaps, $4.75 would be the sweet spot to get some data (albeit slowly), and hopefully one will get some excess traffic drippings from those real 'juicy' placements??

Basically it seems impossible to run the initial test split across those bids and stay in budget? Perhaps I need to find a converting offer first, before it make sense to test that high bid given the low volume available?

---

To address your concern regarding traffic volume, I completely agree with you that from a profits potential perspective, there's not enough traffic in this particular case to justify running a campaign. That graph was just a random example I grabbed from propeller to illustrate the approach of picking bidding points to be a bit higher than each bend to get good value (i.e. you get to pay just enough to jump to the next level of traffic volume).

Something worthy of pointing out: You'd need to judge a geo+targeting based on traffic volumes across at least all the main traffic networks you have plans to scale the camp to. If there's tons more traffic (for your targeting options) on other networks, then you can justify testing on a smaller source if the quality there is better (popads would be the perfect example). But even in that case though, 2k/day would still be too little traffic - testing would slow to a crawl like you said.

Experience can be a double-edged sword actually. People with less experience can see things from a fresher perspective. Appreciate your feedback.



Amy

---

Home > Mobile > Follow-along Campaigns