Home > Mobile > Follow-along Campaigns

---

Two Brothers From Amsterdam Leave Everything Behind And Try Affiliate Marketing (34)

---

Hi guys!

My business partner/brother and I started our mobile pop journey two weeks ago after being inspired by Charles Ngo's ultimate guide to affiliate marketing.

I figured I would start this Follow Along thread to get some guidance from the more experienced affiliates on the forum when we run into problems, have questions and to help/motivate other new affiliates to get started as well (it's a lot of fun guys!).

A little background info on us:
We are two brothers from Amsterdam, The Netherlands and we have been involved in online marketing as entrepreneurs for the past two years (stuff like email marketing, SEO, niche websites). We recently decided to switch to affiliate marketing full-time. That means lurking this forum, finding offers, crunching Voluum data and everything else affiliate marketing is now what we do full-time

Our budget for ads for the coming 6-months is at least $1000 per month.

We currently use the following tools:
- Tracker: Voluum
- Hosting: VPS (liquid web)
- CDN: CloudFlare (free plan)
- Spying: AdPlexity
- VPN: Buffered

Almost instantly after joining STM, we came across Vortex's newbie guide on how to set up our first mobile pop campaign. We've been executing her guide step-by-step the last few days. Here's a summary of what we have done and how things have gone in the last couple of days:

1) We signed up for Mobidea and started looking for offers with the following criteria:
- Vertical: gaming or video (because we are going to direct link)
- Payout = less than $1.50
- GEOs = Egypt, Indonesia, India, Malaysia, Thailand, Turkey, or Brazil.

We applied for about 40 offers in total and were approved super fast after speaking with both Jose and Eva from Mobidea support (thanks guys!).

Note: we also applied for one offer with a payout of $2.64, we wanted to experience why it was a bad idea to choose an offer with a payout higher than $1.50.

2) Next, we signed up for PropellerAds and added $100 to our account. As we didn't have any experience yet with testing multiple offers in a single campaign, we created a new campaign for each offer (both in Voluum and in PropellerAds). We launched around 15 campaigns during the remainder of the day with the following set-up:
- Direct linking to each offer (no landing page)
- CPM = $2.33
- Daily budget = $10
- Total campaign budget = $10
- Targeting for each campaign according to the specifics on Mobidea's platform for that specific offer.

3) At the end of our first day of testing offers, we suddenly saw our first ever conversion appear in Voluum. We celebrated with beer, thanking that random person in Kohima, India who we will probably never meet and has no idea how special he/she is to us, hehe

Later that night multiple offers started to convert. We were happy and slept very well that night.

4) The next morning we analyzed each campaign separately. Here's what we did:

First, we updated the cost of each campaign manually in Voluum according to the 'Cost' column in PropellerAds.

- If a campaign had spent 10x the payout of the offer without converting at least once, we stopped the entire campaign.
- If a campaign had spent 10x the payout of the offer and converted at least once, we followed Amy's advice and started excluding ZoneID's in PropellerAds that had spent x2 the payout of the offer.

Note: Amy's advice was to cut an offer after spending 5x the payout with any conversions, but we figured we'd let things run to 10x the payout, because we wanted to have data so badly, and nothing had converted after spending 5x the payout. Currently, we are cutting offers after spending 5x the payout without any conversions.

After looking at each campaign, we were left with 5 campaigns (meaning 10 didn't convert at all, and 5 converted at least once).

(I'd love to add screenshots of our data in Voluum, but we already archived a lot of campaigns that didn't work out. I will try to make more screenshots for you guys now that I know I've started this thread).

5) After optimizing, we added another $100 to PropellerAds and continued the campaigns we decided to keep.

6) Next, we decided to simply launch more campaigns as we thought it would increase our chances of finding a winning offer faster. We launched another 8 direct-linking campaigns to gaming/video offers using the exact steps from above.

During the rest of the day, we also monitored all our campaigns and updated costs in Voluum every few hours to see if we could exclude more ZoneID's from PropellerAds' targeting (which we could from time to time).

7) At the end of the day, we saw that a specific offer in Brazil that we started running had converted twice within half an hour. When inspecting zoneID's, we saw that both conversions came from the same ZoneID. We figured it would be worth a try to set up a new campaign that only targeted this specific zoneID, just to see what would happen. We knew it could simply be a coincidence that both conversions came from the same ZoneID, but because we're learning and exploring the world of affiliate marketing, we gave it a try anyway.

8) The next day (Friday the 14th) we updated the costs of all our campaigns that ran through the night. Voluum showed that the campaign we started, targeting the specific ZoneID, was profitable.... it made $0.20! We decided to duplicate that campaign and test different CPM amounts, but when logging into PropellerAds we quickly found out that the maximum amount of impressions available for that single ZoneID was only 130 per day... We are currently still running that campaign as it is technically making money, but we have no idea how to scale it within the same traffic source. We know that we could try other traffic sources like PopAds and PopCash, but we're not sure if this is smart. What do you guys think?

We also tried to figure out if any other affiliates are using a landing page for this specific offer, but when ran into the issue that Mobidea isn't listed as an affiliate network within AdPlexity. Is there another way to spy on other affiliates running offers from Mobidea?

About the other campaigns we're running, we kept cutting placements until we had spent 40x the payout of the offer. Then we checked if the ROI of -50% or higher since the start of the campaign. Sadly, none of our campaigns are. I guess we should abandon them, right?

I guess our next step is to just test more offers until we find one that shows more promise.

Also, we are going to start testing multiple offers within one campaign as Amy teaches us in lesson 8 of her guide.

Some extra things we learned so far:
Never assume anything about an offer, just test it and see what happens. The offer that ended up having the most conversions was the one that we hesitated to test because we simply didn't understand why people ever wanted to sign up for such as service.

Thanks for reading and we'll make sure to keep you guys posted!

Adster

PS: If any affiliates in Amsterdam are reading this and want to meet up, shoot us a message!

---

Hey Adster, GL on your first campaigns!

I'm in Amsterdam for business From this Wedensday to Monday.
Looking to connect with affiliates and SEO's for some tight talks, LMK your skype so we can meetup.

Following

---

Hi Guys!

Good to see there is some fresh Dutch blood in town!! I'm also based in the Netherlands not directly from Amsterdam, but we can always set up a meeting in A'dam. I am sure we can set up some nice business together!

Cheers!

---

Sorry I missed this thread - and glad I came across it eventually! GREAT start guys - good job!

The $1.50 was an arbitrary number - basically the higher the payout, the more we need to spend on testing and optimization. As long as you'd be willing to spend more on testing, then payout wouldn't be an issue.


You can do it this way, but it's a lot more work to set up 15 camps instead of rotating all offers in 1 camp. I go into detail on how to do this in Lessons 8 and 9:

https://stmforum.com/forum/showthrea...oluum-Version)


LOL! Yup that first conversion will always be special. You guys will go on to make hundreds, thousands, perhaps even millions, but that first conversion will always hold a special place in your hearts.


Fair enough! It's always nice to see people experiment with different approaches. "My" approach was just to get people to start taking action with minimal risk. But eventually you'll all need to start thinking for yourselves and making your own decisions, and the sooner you guys do that, the better.

As long as something makes sense at the time, just do it. You can always tweak your approach if it doesn't work out. Don't be so afraid to make "mistakes" that you never experiment. And I put the word in quotes because there aren't that many REAL mistakes you can make - there are a million ways of doing things and everyone does it differently anyways. You guys too will develop your own style over time.


Congrats on the initial green - however small the amount, it's SOMETHING!

With pop traffic, whitelisting placements will often not work very well. Often, once you whitelist placements, the traffic would dry up. Try to go the blacklisting approach instead, i.e. keep cutting unprofitable placements.

One thing to look for: See how many of your highest-volume placements are profitable. If most of the placements are in loss, then don't even bother to continue cutting - unless there are other bad traffic segments you can cut (carriers, OSs, browsers, devices, etc. etc.)

Scaling any camp that looks like it has potential, to other sources, would be a good idea! Especially PopAds has good quality traffic - worth a shot for sure.


It would often be too restrictive to try to look for landers for a particular offer.

Rather, try to look for landers for a particular TYPE of offers.

It would be quite difficult to find/create a lander for 1/2-click gaming offers - because these games really don't need a lot of pre-selling. A lander is a double-edged sword. It will decrease the number of visitors that will click through to see your offer page, but the idea is that it will pre-sell the visitors so that those that DO click through to the offer, would be MORE likely to convert. This means that in order to justify using a lander, the lander will need to pre-sell well enough such that the positive will outweigh the damage it does in decreasing the number of visitors that will reach the offer page. And because these gaming offers require little to no pre-selling, it would be difficult to create a lander that will have an overall positive effect on conversion rate.

However, don't let me stop you guys from experimenting!


Let's do a post-mortem on them to see what we can learn - after all, you guys have already spent the money!

Let's start with screenshots of the following for each campaign:

-Placement stats - sorted by decreasing conversions, and also by decreasing impressions.

-OS stats, browser stats, device stats.

The basic thing to consider when trying to decide whether to ditch a camp or not: Do you have traffic segments that is/are making enough profits to justify your time in optimizing/managing the camp? Traffic segments can be placements, OSs, browsers, carriers, whatever.

If you have enough segments making enough profits, then simply cut unprofitable segments. If not, then don't bother cutting further.

This is assuming you're sticking with the same funnel (i.e. lander+offer, well just offer since you're direct-linking). If the camp isn't making enough of the traffic profitable, then you'll need to test more offers and/or landers until you do.


With the direct-linking approach, yes you'll need to test a LOT of offers in order to find something that has promise.

Once you guys are comfortable with setting up camps and analyzing stats, I would HIGHLY recommend moving onto landers ASAP.

Then once you guys are comfortable with optimizing camps, I would recommend expanding outside of pop, into other traffic types. Pop traffic is not the most promising traffic type - I went into this in one of the lessons.


Very good!


Please do Adster! Looking forward!



Amy

---

Good luck guys! We have an office in Amsterdam, don't hesitate to visit and have a beer with us

---

Greetings from Rotterdam (also in the Netherlands)

---

Did you guys create your own banners or just use the banners supplied by Mobidea? Also, what did you use your CDN and VPS for?

---

Hi gmfdirector,

We were direct linking, so we didn't use any banners. We also didn't use our VPS + CDN for this campaign, because we didn't have to host any landing pages, I just included it in the post to give everyone an insight into what tools we use from the start of this thread, so following along is easier.

Update: We switched from our VPS + Cloudflare to Amazon S3 + CloudFront. After testing these two set-ups against each other we found that Amazon S3 + CloudFront is as fast or sometimes even faster on 3G connections (and also a LOT cheaper). Here are the test results from GTMetrix if you're interested:

Sao Paulo (Unthrottled connection)
——
VPN + CloudFlare = 1.2s
First paint = 0.7s


S3 + CloudFront = 0.393s
First paint = 0.2s


Sydney (Unthrottled)
——
VPN + Cloudflare = 3.3s
First paint = 0.3s


S3 + CloudFront = 1.1s
First paint = 0.7s


Mumbai (Unthrottled)
——
VPN + Cloudflare = 0.264s
First paint = 0.2s


S3 + CloudFront = 1.5s
First paint = 0.7s


Hong Kong (Unthrottled)
——
VPN + Cloudflare = 0.358s
First paint = 0.2s


S3 + CloudFront = 0.378s
First paint = 0.3s


Dallas (Unthrottled)
——
VPN + Cloudflare = 0.7s
First paint = 0.4s


S3 + CloudFront = 0.6s
First paint = 0.1s


Vancouver (Unthrottled)
——
VPN + Cloudflare = 0.6s
First paint = 0.3s


S3 + CloudFront = 0.5s
First paint = 0.4s


—-–––—————————————————————————


Sao Paulo (3G)
——
VPN + Cloudflare = 2.1s
First paint = 1.1s


S3 + CloudFront = 1.6s
First paint = 0.9s


Sydney (3G)
——
VPN + Cloudflare = 1.6s
First paint = 0.9s


S3 + CloudFront = 1.6s
First paint = 0.9s


Mumbai (3G)
——
VPN + Cloudflare = 1.5s
First paint = 0.8s


S3 + CloudFront = 1.5s
First paint = 0.7s


Hong Kong (3G)
——
VPN + Cloudflare = 1.5s
First paint = 0.8s


S3 + CloudFront = 1.8s
First paint = 0.6s


Dallas (3G)
——
VPN + Cloudflare = 1.8s
First paint = 0.9s


S3 + CloudFront = 1.9s
First paint = 0.9s


Vancouver (3G)
——
VPN + Cloudflare = 2.0s
First paint = 1.0s


S3 + CloudFront = 1.6s
First paint = 0.8s

---

Thanks Adster for those stats!

Cloudfront is what I've been using also - without issues.

How are you guys doing? Any updates? Been wondering how you've been getting on...



Amy

---

Hi everyone!

Thanks for all the replies.

It's been a while since I've posted. We've been on holiday in Germany for two weeks and just returned this week.

Germany was awesome, and we have trained our beer-muscles quite well there.

In other news: we have started running a new campaign!

Vertical: Sweepstakes
Niche: win an iPhone
GEO: Netherlands
Traffic source: PropellerAds

Here are the steps we took and the data that has come out:

1) Talked to my AM about which offers are currently converted well. He gave me two iPhone7 offers and three iPhone7 RED offers (high-five Rick!).

2) Looked around AdPlexity for well-performing landing pages using these filters:

Ad Type: Popup
Country: Netherlands

Keyword (on the landing page): iPhone (sort results by received most traffic).

I found that there were 5 different types of landing pages people are using. I opened all of them in a new tab. Then I sorted the box in the top right called 'Landing Pages' by 'Days Running' in each individual tab and downloaded the one that has been running the longest.



Question 1: Is it smart to only download the version that has been running the longest?
Question 2: Are there other stats on AdPlexity is should be inspecting?

3) Next, I opened the HTML of each landing page in sublime-text 2 and removed all the Google, Facebook and Apple logo's as the offers I want to run didn't allow the use of any brands of logos. After that, I checked the grammar of each landing page and found they all had a massive amount of spelling mistakes. I corrected all grammar mistakes. Lastly, I changed all the URLs to my own Voluum 'Click Url'.

I ended up with 3 landing pages, as I couldn't seem to edit the 'Click Url' in one of them (it's a piece of Javascript that I haven't figured out yet) and the other one had so many Google logo's on it, after I stripped them all the page just looked terrible, hehe.

4) I uploaded all the pages to my S3 server using Cyberduck (it's an awesome tool that lets you manage your S3 server like it's an FTP server. Overwriting, deleting and managing files on your S3 is super easy with Cyberduck.)

I use this URL structure for storing my landing pages, but I'm not sure if it's the best option:

http://mydomain.com/lp/NL/iphone7/lander-1/index.html
http://mydomain.com/lp/NL/iphone7/lander-2/index.html
http://mydomain.com/lp/NL/iphone7/lander-3/index.html

Question 3: Is this a smart way to do it? Or will things get messy as soon as I'm a few months in using this structure?

5) I added the two iPhone7 offers and my three to landing pages to Voluum. Next, I set up and campaign dividing the traffic equally between the 2 offers (50/50) and the three landing pages (33/33/33)

6) I set up a new campaign in PropellerAds using these settings:

Pricing Model: CPM:
Frequency: 1
Capping: 24
Countries: Netherlands
CPM: $2,33

Daily campaign budget: $20
Total campaign budget: $20

Question 4: The payout of both offers is $0.88. I figured I'd let both offers run around 10x the payout (which would be $8.80) and then rounded that number up to $10 per offer, making the total budget $20. Am I spending too much money on testing these offers?

Ad delivery method: Distributed
Targeting: All (3G and WiFi)
OS's: All
OS Versions: All
Devices Types: Phone, Tablet
Devices: All
Mobile ISP: All
Proxy: Not Proxy
Browsers: All
Browser language: All

And then I hit that magic blue button..... 'Save and Start Campaign!'

7) Now for some Voluum data from the overal campaign, after spending $20.11 in Propeller Ads:



Visits: 7,772
Clicks: 153
Conversions: 3
Revenue: $2,25
Cost: $20,11
Profit: -$17.86
ROI: -88,81%

Here's the data from the two different offers:



Here's the data from the three different landers:



Here's the data from the ZoneIDs sorted by visits:



Here's the data from the ZoneIDs sorted by clicks:



Here's the data from the ZoneIDs sorted by cost:



Here's the data from the ZoneIDs sorted by conversions:



Here's the OS data:



Here's the Browser data:



Here's the Device data:



If you need any more data, please let me know and I'll provide it!

8) What I've done so far is:
- Cut lander 3
- Cut the ZoneIDs that have spent x2 the payout (which is only 1 so far)

Question 5: How should I continue optimizing? Or is it even worth optimizing?

My ideas: Test more landers, cut iOS traffic, cut tablet traffic, keep running until I've spent 50x the payout and then see if the overal ROI of the campaign is at least -50% or better. What do you guys think?

PS: Amy -> I guess I posted this follow-along in the wrong sub-forum. If you want to move it, feel free to do so

---

Hey Amy,

We've been on Holiday, I just posted another update, would love your opinion and advice!

Adster

---

Good luck guys, also based in Amsterdam as an aff, feel free to hit me on skype, drop me a PM

---

We've held the last 2 Affiliate World Conferences (Europe) in Berlin - absolutely share your sentiments about Germany being AWESOME! Glad to hear you guys had a wonderful vacation!


What you do is exactly all that I do to rip landers.


May be better to have ripped a few more - but then you could always test more later on.


Your structure is as good as any other - but I personally find it more convenient to do vertical / offer type first, then geo. So in your case, [path]/iphone7/NL/1/index.html etc. This is because when you're testing a vertical you won't normally just test one geo - I just find that having multiple geos under one vertical folder minimizes the clicking around.


I wouldn't say you're spending too much - I think that sounds about right - basically you'd be spending over 3x payout testing each lander+offer, which is OK for a start. I think Mr Payne and MrBraun have some formulas for calculating test budget. I personally would just run some traffic and look for trends, and then decide on the next step.

Actually, I prefer keeping the campaign running and just cutting offers/landers as they reach statistical significance - as long as the campaign still looks like it has a good chance of ending up profitable. Because you can't tell ahead of time how much spend it would take for landers/offers to reach statistical significance, it would be difficult to set a budget. Also keep in mind that even if your current 3 lander + 2 offers don't end up profitable, once you've cut down to 1 lander + 1 offer, you can test more landers, and then more offers - after a couple more rounds of split-testing, results can look VERY different.

So I would suggest to just run the $20 as you've planned, then look at the stats and decide on the next step.



Wow - it's very impressive that you would provide so many screenshots! I've reviewed them carefully. There's promise there certainly!

What you didn't show, are offer > lander stats, but fortunately in this case because all 3 conversions were made by a single offer, it's not hard to figure out that the best-performing offer+lander combo so far, is doing around -60% ROI ($1.50 revenue, ~$4 cost).

I would have done exactly what you have, i.e.:
Except - I think you meant cut OFFER 3 and not lander 3?


YES this is a campaign that's DEFINITELY worth keeping running! Remember you've only tested 2 offers with 3 landers so far!

Why would you want to cut IOS/Tablet traffic? There are only 3 conversions so far which tells you nothing about trends in OS/devices/browsers. If you were losing a lot of money, I would suggest to just target the best-converting traffic segments and doing further testing there (although doing so could skew your test results). But this isn't the case, so I would recommend to keep all traffic segments running in order to get your next split-tests done faster.

Don't focus on the cutting for now - instead, focus on improving your funnel, which means testing offers and landers!

Either do a round of lander-testing first, then do a round of offer-testing, or the other way around - and we'll see what happens!

Once you have a funnel that's good enough to make a good chunk of the total traffic profitable, you can then focus on cutting the unprofitable parts.


Done!


Thanks for a most-detailed post! Looking forward to further test results!




Amy

---

Thanks for the super clear advice Amy!

I'm almost done making my changes to this campaign, and I will post another big update soon. In the meantime, I had a couple of questions:

1) You said: "Either do a round of lander-testing first, then do a round of offer-testing, or the other way around - and we'll see what happens!"
I'm currently adding more offers and landers in 1 round, is that okay to do as well?

2) I'm currently adding new iPhone7 offers to this campaign (from others networks and other advertisers), should I give the ZoneID's that I have blacklisted another chance? Or may I assume the blacklisted ZoneID are better to stay blacklisted because the offers are about the same topic (win an iPhone7)?

3) Is there a rough conversion rate percentage which indicates when a landing page is performing adequately with pop traffic? For example: in email marketing, a lot of people say an open rate of 30% is an average you should aim for.

4) To which traffic source/method should I switch once I get the hang of mobile pop traffic? I hear mobile pop is great to get started, but many have advised me to switch once you feel more comfortable with affiliate marketing. Which options should I consider once that time comes?

Thanks!

Adster

---

I suppose that could work as well, but doing so MAY take longer for results to reach statistical significance. I don't know (and probably can't know without working out the statistical math involved) whether testing 6 landers would take the same amount of time as testing 3 landers and 3 offers.

Again - I guess we'll see what happens then - when we review your stats.


Well it's not just the topic - if you have an offer that does higher eCPM than your previous offers, then there is a chance that some of the previously-blacklisted placements could be profitable now.

I would suggest to leave them blacklisted until you can make the campaign green first. Then you can choose to retest some of the bigger ones that don't look hopeless (i.e. that aren't major budget-drainers that have performed badly).


Such a conversion rate doesn't exist I'm afraid, and would not be useful at any rate. Looking at CR alone would be out of context, and won't tell you what ROI / profits to expect, without also considering what the offer payout is and what the traffic costs are.


Mobile display, native, search (adwords), FB. Especially the latter 2 because they're the biggest and have great profits potential.



Amy

---

Hi everyone,

@Amy, thank you once again for the quick response. You've been really helpful and your answers and advice are a great remedy against information overload!

Here's another big update of the iPhone 7 campaign:

First some bad news: the offer that was doing well got paused, and was replaced with another offer from a different advertiser.

This week (among other things) I've looked into using Voluum tokens on my landing pages.

I thought it would be a nice test to duplicate lander-1 and lander-2 and insert some variables in there to see if it would improve the conversion rate. I've inserted tokens like {brand}, {model}, {region} and {browser}. The duplicated landers with the tokens are lander-4 and lander-5 in the Voluum results below.

Besides that, I downloaded 3 more landers from AdPlexity and added them to the campaign. Lander-6, lander-7 and lander-9 in the results below.

That means that I now have 7 different landers.

Lastly, I looked around for some more iphone7 offers on another affiliate network, and found 3 offers that I could run. That means I'm currently testing 5 different offers.

I've spent another ~$60, making my total investment on this campaign $80.28 so far.

Now for some Voluum data:

Landers:



(4 conversions in lander-1 are from the paused offer and 2 conversions from lander-2 are also from the paused offer)

Offers:



Landers + Offers (1):



Landers + Offers (2):



Question 1: Something I'm currently not sure about is when to cut an offer and/or landing page.

1) Should I look at the offer overview, and cut an offer when it has spent 5x-10x its payout without converting?
2) Should I look at the lander + offer overview and cut an offer as soon as it has spent 5x-10x the payout in each landing page + offer combination?

Question 2: When should I cut a lander? Lander-4 hasn't had any conversions yet, but has the highest CTR of all landers. I'm not sure what to conclude from that. Is the lander good and are the offers the problem, or is the lander bad even though it has a high CTR because it hasn't had any conversions yet? The same goes for lander-9, which has the second highest CTR but also no conversions yet.

My idea would be to first cut the offers that haven't converted, and then send traffic to the offers that have converted still using all landing page variations to see how all the landing pages perform with the best offers. After that, I can start cutting landers. Is that a good idea?

ZoneID's sorted by clicks:



ZoneID's sorted by conversions:



ZoneID's sorted by profit:



Question 3:
When a ZoneID is in loss but is showing high CTR when compared to the other ZoneIDs, should you keep it around? Or should you just cut it no matter how high the CTR is? Here's an example of one:



What I've done so far:
- Cut the ZoneID's that have spent 2x the average payout of all the 5 offers combined ($2.64)

My ideas on how to continue:
- Cut the offers that haven't had any conversions yet, and keep running traffic using all landing pages. Then cut the landing pages that are performing badly.

In other news:
1) I visited Advidi's office this week and had lunch with my AM, Rick. It was great! Planning on visiting more often and hanging out together.

2) I've decided to start a mastermind group in Amsterdam. We're having our first meetup this Tuesday. So far, 4 people will be there. If anyone else is in the neighbourhood and wants to join, send me a PM and we can talk. The goal of the mastermind is to have fun, learn from each other's mistakes and successes, and to motivate each other when times are tough and you could really use a hug.

That's all for now folks!

Adster

---

Hey Adster!

Interesting follow along you have here! Your campaigns and approach seem pretty structured, and you are thinking of details trying to plan future steps - good job!
I'll follow up, and try to help as much as possible too.

Regarding your question when to stop the offer: definitely it's better to stop the offers that has 0 conversions, since the amount of traffic you've brought seems pretty decent. And if there is not even 1 conversion at this point, being me, I would not optimize more but stop the offer and look for different opportunities to push.

You could choose the offer that seems more promising in terms of stats you've gathered and test all landing page for that one - this idea of yours is really nice. Pay attention to those landers that already have a low CTR, and cut them first if the situation doesn't improve.

Talking about ZoneID's - it's better if you focus on eCPA/eCPM of the zone. Remember that CTR is not what defines your profitability. It can be that high CTR is coming from the users accidentally clicking on your banner or even bot traffic.

Best of luck! And I hope it helps

---

Thanks a lot for the advice Mobidea! I really appreciate it. I'll make sure to keep posting updates regularly here in this thread.

Adster

---

Hi everyone,

Just wanted to share my game plan for the coming weeks. I’m wondering what you guys think of this approach, and whether it’s a good one:

I’m currently running my iPhone 7 sweeps campaign in NL. I’m learning a lot about how to run campaigns that contain multiple offers and landers. I’m also learning more and more about which landers convert well, and which don’t. Something that really helps as well is that I can actually understand the text on my LPs, because they’re in Dutch.

I’ll spent the coming time learning how to optimise campaigns like this one. I feel like I’m getting the ins and outs of how to analyse my data and when to cut offers and landers in these kind of campaigns a little more each day (thanks to everyone’s help and advice).

When I feel like I’m ready for the next step, I will have my best lander (or best 2, depending on how things go) translated to Polish, (Brazilian) Portuguese, Thai and Indonesian.


After that I will set-up campaigns in these GEOs, running iPhone sweeps offers.

I chose Poland, Brazil, Thailand and Indonesia as I’ve heard they’re good GEOs to target as a beginner, and because they’re big countries. I have friends who are native Polish and Brazilian Portugese speakers, so getting my LPs translated will be free for those GEOs.

Next I will start running sweeps campaigns on a bigger scale than I’m doing right now, testing all offers that are available for each GEO.

I got this idea from reading MrBraun’s follow-along last night:
https://stmforum.com/forum/showthrea...l-20-years-old

Would love everyone’s honest feedback on this strategy!

Have a good weekend everyone,

Adster

---

First of all - it's terrific that you're testing more landers and offers! And adding voluum tokens will take your landers to the next level - definitely worth testing.

However, a word of warning:

Every time you add new candidates to a split-test, it starts another round of split-testing, which means it would be inaccurate for you to count any data from before you added the new candidates.

What does this mean?

Say you had been testing landers 1-3 since last week, and then yesterday, with the 3 landers still running (with no statistical significance reached yet), you added landers 4-5 to the test.

When comparing lander stats for statistical significance, you can't set the time range to be last week and this week - you can only set the time range to yesterday+today, i.e. only look at the timeframe during which ALL landers were being run simultaneously.

Why can't you include stats for landers 1-3 before yesterday? Because campaign performance can vary from day to day, and even hour to hour.

If the old batch of landers happened to have been running on some bad-performing days of the week, and your new landers were added on a good day of the week, then, comparing ALL available data since the start of the campaign, would be unfair to your old batch of landers.

You may already know this, but this concept is important, so I'd like to clarify just in case there's any confusion.


When landers are in the picture, you can't really use an x-times payout rule of thumb to judge an offer.

Coincidentally, I've just written about that concept in detail in another thread earlier today (I love these coincidences!):

https://stmforum.com/forum/showthrea...l=1#post322886

This is what you should be aiming to do to optimize your funnel:

1)Use a couple of offers recommended by AMs, that you know has converted semi-decent for multiple affiliates, to test 5-10+ landers. Use this method to cut landers down to a winner:

https://stmforum.com/forum/showthrea...Banners-Part-1

2)Take the winning lander, and mass-test offers (all similar offers you can find on your aff networks and perhaps beyond). Use this method to cut offers down to a winner:

https://stmforum.com/forum/showthrea...211#post289211

3)Hopefully, the best offer+lander (last offer+lander left standing after all the cutting) will be able to make enough of the traffic profitable (e.g. at least a few big placements jointly providing enough profits/day for the camp to worth your time in continuing). If this is the case, proceed to cut everything that isn't profitable (placements and/or other traffic segments, such as OSs/browsers/carriers....)

The above is a general approach only. The idea is to keep eliminating the worst offers and landers until you're left with a winning lander and a winning offer. If this offer+lander is still not able to make enough of the traffic profitable, you'll need to test more landers or offers or both until you find a lander+offer combo that CAN make enough of the traffic profitable.

And if you think you've tested all the popular landers plus a lot of offers, and still can't make enough of the traffic profitable, consider giving up on the campaign.


And it's not a good habit to judge a lander by its CTR - except in cases where it's very low. Quoting what I said in another thread for your reference:

You've pointed out a catch-22 situation that's so commonly seen in initial stages of a campaign: When you throw some landers and offers into a camp and don't get conversions, you don't know if it's because the landers suck, or if it's because the offers are duds, or both.

The testing process I've proposed in the last section above, was designed to minimize exactly this problem (i.e. using a couple of AM-recommended to cut landers down to a winner, then taking this winning lander to mass-test offers).

When you rip at least 5-10+ of the most popular landers you see, chances are good that at least one of them is a good converter. And when you start a campaign using AM-recommend offers that have converted reasonably well for multiple other affiliates, chances are good that you'll get at least some conversions.

So, the idea is to lock down the 2 main variables one at a time: First use proven offers to cut landers down to a winner (thus locking down a good lander), then use this proven lander to test all the offers you can find (to lock down a good/best offer).

You said: "My idea would be to first cut the offers that haven't converted, and then send traffic to the offers that have converted still using all landing page variations to see how all the landing pages perform with the best offers. After that, I can start cutting landers. Is that a good idea?" <--- Good idea - our approaches are similar. You can run all offers and all landers together first, see which offer APPEARS to be converting well, turn off all other offers temporarily, and cut landers according to statistics until you're down to the last winning lander. BUT THEN, you may want to use this lander to retest the offers you've paused before. (Because, remember that the offer you were using only appeared to be the best, not actually THE best, because statistical significance hadn't been reached yet.)


Again, I wouldn't normally consider CTR - not for landers OR placements.

You may have suspected this to be the case already - basically if a placement is losing money, then it doesn't matter how high its CTR is.

(In fact, if most placements have low CTRs, and you see some placements have abnormally high CTRs - it may not be good. The high CTR may be because bots are clicking through the lander automatically, or because the placement website owner has otherwise rigged their setup to artificially increase the CTR.)

I want to share a couple of thoughts that went through my mind while looking through your placement data:

1)Some of the placements in excessive loss should be blacklisted, although you're not yet running your best lander+offer combination. You should blacklisted them now, and maybe choose to retest some of them after you've identified your best lander+offer.

2)Most of the placements that HAVE conversions, are in profit - this is a good sign, especially when you haven't even cut down to the best offer+lander yet.


Splendid idea!


Solid plan.

I know your overall campaign ROI seems really bad, but you're testing lots of landers and offers all at once. There's definitely promise at this point in time. Eager to see more stats before we decide on the next step.


Sounds like you're doing some major networking!

That's one of the best things you can do, both to grow as a person and as a business.

Plus it's fun to talk to people (if you're an extrovert - which you sure sound like one!)


Did you rip your current landers from Adplexity?

If so, then instead of translating, why not just rip similar landers in the other languages from Adplexity also?

Also keep in mind that there are no promises that every geo will respond the same to different landers. When you test the same landers A and B in geos X and Y for example, you could see lander A doing the best in geo X, but lander B doing the best in geo Y.

And then, there may be a lander C that may do very well in both geos X and Y (and maybe even Z).

Of course, you could take your chances, and stick to the plan you've laid out - that would be the less accurate but more efficient way. I'm not going to say "don't do it this way" because it does have its merits: You could potentially conserve a lot of test budget doing things this way.

But I WILL suggest an additional option: Rip and test the most popular 5-10+ landers for the first few geos as you normally would, see if you can identify 1-2 landers that do very well consistently across all geos (of at least most), THEN scale to all other geos using these 1-2 landers.

And here's a THIRD option: Group geos that have similar cultures - for example West Europeans, East Europeans, Africans, South Americans, Arabic - test the 5-10+ landers as normal for ONE geo in each group, then scale the 1-2 winners to the other geos in the group.

Scaling will be fun! I'm very happy for you that you're thinking big. But I would suggest for us to run and optimize a couple of camps together first. You haven't got all of the basics down yet. You will though, soon enough.



Amy

---

Hey Amy,

Thanks for the response, learned A LOT of things from your last post!

I have been implementing your advice over the last couple of days and once again have some results to share!

I picked 2 offers from the batch of offers I was testing that had converted and also talked to my AM over at Gotzha for the currently best performing iPhone7 sweeps offer in NL.

That left me with 3 offers that I'm going to use to test my landers.

I’m currently trying to cut down my landers to a winner using the system you explained inside this post: https://stmforum.com/forum/showthrea...Banners-Part-1

Here’s some data, after running for two days.

Landers:

Voluum-data-landers.png  Views: 56  Size: 117.8 KB  ID: 16532" class="thumbnail" style="float:CONFIG" />


Offers:

Voluum-data-offers.png  Views: 42  Size: 93.8 KB  ID: 16533" class="thumbnail" style="float:CONFIG" />

Offers -> Landers data:



And my spreadsheet with the data from the Peak Conversion A/B testing tool (https://www.peakconversion.com/2012/...-calculator/):



As you can see, not one lander has an 'Apprx probability of being best' that is <10%, so I haven't cut any landers yet. I guess I will just have to keep running more traffic until landers start getting below 10%.

Question 1:
I'm currently not sure what to do, I feel like I have 3 options:

Option A: Keep running the campaign like it is (with all 3 offers and all 7 landers) and start cutting landers as soon as they get below a probability of being best that is <10%.
Option B: Cut the Advidi offer from the campaign and start a new round of testing all landers with just the YepAds and Gotzha offers. Then start cutting landers as soon as they get below a probability of being best that is <10%.
Option C: Cut both the YepAds + Advidi offer from the campaign and start a new round of testing all landers with just the Gotzha offer (because it's the best converting offer). Then start cutting landers as soon as they get below a probability of being best that is <10%.

I'm not sure whether it's a good idea to start cutting offers that haven't converted during my landers testing round, but I feel like I'm losing some unnecessary money sending traffic to the Advidi offer.

Other things I have been doing:
- I'm still cutting ZoneID's that have spent x2 the payout. I've currently cut 17 ZoneID's.

Answer to your question:
Q: "Did you rip your current landers from Adplexity?"

A: I did rip them indeed. But I've found that the grammar and copy used is so extremely bad (for the NL landers I downloaded), that I don't really feel confident just running landers from AdPlexity without having them checked first. I feel like having them properly translated gives me a bigger chance to beat my competition.

That's all for now!

Adster

---

I just hate it when all the landers are performing around the same. I would suggest to run them a bit longer to cut at least some of the landers. But if, in the end, you end up having several that keep going head-to-head converting the same, then heed caurmen's advice:

https://stmforum.com/forum/showthrea...ugh-Clicks-Yet

"3. If you've had a few hundred clicks worth of data for each landing page, all of them have probabilities of being best between 25% and 75%, and you're seeing high, overlapping graphs like the ones below, the remaining landers all perform about the same. Pick one to continue running, pause the others, and start a new split-test (see "What To Test Next")."


Option D: Cut landers and offers simultaneously as they reach stat sig.

Here's a tool for cutting offers:

https://stmforum.com/forum/showthrea...211#post289211

Sure enough, by cutting landers and offers separately like this, we're ignoring the possibility that certain offers may convert better with certain landers - but this would be the more efficient approach.


Good! Once you have a winning lander+offer you may want to retest some of those - in case they'd be profitable then.


Fair enough.

Google translate nowadays (unlike before) does a pretty good job translating simple sentences. So what you can do to avoid spending the time and money upfront to get proper translations, would be to break sentences down into short ones and use google translate - just another option.


While you're waiting for stat sig for the current batch of landers and offers, this would be a good time to seek out more offers and lander styles to test - for your next split-tests.

Things are progressing well - looking forward to seeing how this camp will turn out!



Amy

---

Thank you Amy!

I was under the impression I first had to only cut landers in phase 1, and then cut offers in phase 2.

But if I understand correctly, you’re saying I can start cutting offers right now (using this tool: https://win-vector.shinyapps.io/CampaignPlanner_v3/) and also keep cutting the landing pages using the peakconversion tool, at the same time.

I've used the campaign planner tool, here are there results:

Chance that Gothza offer is better than YepAds -> 97%
Chance that Gotzha offer is better than Advidi -> 100%

Based on these results I've stopped the YepAds and Advidi offers and am now only running the Gotzha offer.


One question I have about that:
Now that I've changed the set-up of the test (removed 2 offers), should I only look at the data that has come in after I removed those two offers when I’m deciding which landers to cut, or may I take all data in consideration when using the PeakConversion tool (i.e. also the data that was collected when I was still running all 3 offers)?

Other things I've done:
- Cut 6 more ZoneIDs this morning.
- Signed up with ClickDealer and set-up 3 more iPhone7 offers which are ready to be tested when the time comes.
- Had breakfast

Thanks!

Adster

---

Hi everyone,

I'm currently setting up PopAds.net as a traffic source in my Voluum dashboard and I'm wondering if I've set it up correctly. I can't find any guides or tutorials on how to set up PopAds.net with Voluum. Have I done it correctly like this?:



Thanks in advance,

Adster

PS: The reason I'm setting up PopAds, is because I've heard that they allow more aggressive marketing. I've got a couple of landers that get rejected on PropellerAds because of the use of copyrighted logo's.

---

Correct!

I can understand the confusion - it's because I'm always suggesting to focus on cutting landers in phase 1 to lock down a good lander first, and then use it to mass-test offers in phase 2.

However, if you're using several offers to generate the conversions required to lock down a lander, then cutting offers at the same time would make your testing cheaper. And the last offer standing will then act as the test control when you add new offers in for the subsequent mass-testing of offers later.


Fabulous!


All data from all offers.

We're not taking into consideration the possibility that each offer may convert better on different landers. In other words, we're assuming all offers will convert the best on the same lander. This may not be the most accurate assumption, but the alternative would be to consider EACH offer+lander combination as one test candidate - which would require a LOT more test budget.

To remedy this inaccuracy: Once you've found a winning lander, used it to mass-test offers, and found a winning offer that converts well, you can always test all kinds of landers for this winning offer - to find a lander that converts especially well for just this particular offer.


Very nice LOL!

The last one is especially important IMO. We need to always take care of our stomachs first. Food is crucial.


It's so simple you don't need a tutorial. You've set it up just fine - you could fill up all 10 spaces if you like. Nothing wrong with collecting more stats.

Or, you could just use the template from Voluum:






Amy

---

Thanks Amy!

Here's a quick update:

I've finally been able to pick a winning lander. After days of running traffic, lander-4 and lander-9 ended up going head-to-head. After running more traffic during the weekend, lander-9 had a probability of 80% of being best. I decided to continue with lander 9, made 3 variations of it and started a new round of split-testing.

Running traffic now and will post another update with stats soon!

Adster

---

Hi everyone,

Here's another update of the iPhone7 campaign.

As I already quickly said in my previous update, lander-9 and lander-4 ended up going head-to-head for a long time. After running for a couple of days, this were the results:



A = lander-4
B = lander-9

I decided to continue with lander-9. I immediately started a new split-test and created 3 variations of lander-9. After running for a day, this is what happened:



A = lander-9 (original)
B = lander-9 (variation-1)
C = lander-9 (variation-2)
D = lander-9 (variation-3)

Lander-9 (variation-2) had converted 4 times and was actually in profit after running for 24 hours!

I decided to continue with lander-9 variation 2, and again made 2 variations based on that winning lander.

Here's what's happening now:
I'm running traffic to lander-9 variation 2 and its 2 new variations since yesterday afternoon. Sadly none of them have converted yet, so I guess I will have to wait a bit longer until I can pick a winner out of that bunch. After that, I will start testing offers. I currently have 13 offers from 4 different networks waiting to be tested.

Question: I'm kind of surprised by the big difference in ROI per day. Lander-9 variation-2 converted four times within 24 hours during the first split-test. Now (a day later) it hasn't converted even once after running the same amount of traffic during the second split-test. Is that normal?

Other things I have been doing:
- I have cut around 39 ZoneIDs so far and have about 68K impressions per day still available.

That's it for now.

Hope to give you guys another update soon!

Adster

---

Hey there!

Is this still that one offer for Netherlands? And what are you referring to when you talk about "trials". Are those clicks or visits?

From the last screenshots with the 4 variations of a lander-9, it feels like it's too soon to optimize. Just one day passed, and being you I would give it another try.

Regarding your question, if it's normal - I believe it's not normal but it can happen. This is the stats you've got only from one day of running, and being you I'd give it another try.

Let us know how it goes after all!

---

Hey Mobidea,

Yes, this is still the iPhone7 campaign running in The Netherlands.

Trials are visits (impressions) of the landing page. The screenshot is of this tool: http://peakconversion.com/2012/02/ab...al-calculator/

It's the tool that is recommended in this article on how to cut landing pages from a split-test.

I see what you're saying, that it might be too early to declare a winner after 1 day of running.

I'm curious to find out what Amy thinks about this, and whether I should go back to the first split-test of testing lander-9 and its 3 variations.

Adster

---

Woot woot! *applause*

This is the power of split-testing! Nicely done!!


Sounds perfect! Was just about to suggest not to test too many rounds of lander variations. The thing about optimization is that sooner or later you'll optimize past the point of diminishing returns. A better approach would be to split-test one variable, then another variable - for example a round of landers, then a round a offers etc. That approach will likely result in faster increase in ROI.


This often happens when you resume a camp after pausing it for some time - you'll get more conversions in the initial hours/day. Same thing for new camps - sometimes they'll do better initially. Also, a camp can encounter good days and bad. Give it another day.


Very nice! Having enough volume is important - as you've illustrated, the more volume you can get, the more you can afford to cut and still have enough left over.


This goes back to the accuracy vs. efficiency argument I'm constantly bringing up.

You can be more efficient and cut landers based on one day's worth of data, or you can be more accurate and cut landers based on multiple days' worth of data.

I'm of the opinion that, because all the landers are being run simultaneously and are therefore subject to the same set of conditions, the day-to-day volatility should have similar impact on all landers. Therefore I don't usually drag out my split-tests. But that's just me. I never did bother to confirm this - it would take quite a few split-tests to gauge whether the extra accuracy would be worth the extra time.

You can of course re-run your most-recent lander split-test and throttle it over several days (for example). But I'd give the current winner a chance first - if after a couple of days the performance holds, that would be a good sign. At that point you'd be ready to scale the camp to other traffic sources, then test bids and cut placements on each source.

Things are heating up! Looking forward to seeing how big you can scale this!



Amy

---

SmartCPM is a good way to get a sample of traffic from many placements, which is a good thing when you're testing offers - this way you can be more confident of traffic quality, i.e. that you're not just getting crappy traffic from bidding too low.

However, as you've found out, it would make it more difficult to blacklist placements.

Here's a tool that can help you keep track of costs and revenue for placements:

https://stmforum.com/forum/showthrea...kly-amp-Easily

I don't know of a way to update cost per zoneID in Voluum. AFAIK, it's not possible at the present time. You can verify this with Voluum support.

As for increasing bid on a placement level - AFAIK PropellerAds' system does not allow this, unlike Zeropark. The best way to handle this would be to set up staggered bid campaigns, and cut placements for each campaign. It won't give you as much freedom as being able to tweak bids for individual placements, but at least you'd have several "bid pockets" to target placements with.

So, one approach would be to use smartcpm for initial testing of offers, but then once you find a promising offer, to switch to the original CPM bid setting.

Sounds like you're very confident in the potential of your offers! Best of luck with that - I look forward to seeing some green!



Amy

---

Hi all!

Wow, it's been a while since I posted.

During my last few campaigns I've been experimenting with PropellerAds' new model, smartCPM, which lets you define separate CPM bids per placements (instead of paying one specific amount for all placements). Propeller is now also able to send the cost of each placements to Voluum, which makes optimizing a lot faster and easier.

Now then, some problems I've been running in to:

Basically my AM at Propeller told me that smartCPM is perfect to find profitable placements, and CPM is perfect to buy as much traffic as possible from a placement (she even told me CPM campaigns have priority over smartCPM campaigns traffic wise). So she suggested that as soon as I found profitable placement that I wanted to get as much traffic as possible from, I should:

1) Set up a new whitelist campaign targeting only that placement.
2) Blacklist the profitable placement from my smartCPM campaign.

I told her I've been advised to not run whitelist campaigns on Propeller (which surprised her), but I decided to give it a go anyway to see what would happen.

So I had a couple sweeps campaigns running in NL and DE, some of them had just reached profitability, and some of them were almost break-even. I tested her system and set up new CPM campaigns each targeting a profitable single placement from my smartCPM campaigns to get as much traffic from that placement as possible.

Sadly, different types of issues came to light:
- The traffic estimator would indicate a certain amount of traffic for my bid, but actual traffic was way lower. I consulted my AM at Propeller about this, and she told me to bid higher. I was kind of disappointed with the fact that traffic estimator was displaying incorrect data and the solution was to constantly speak to her about what to do. I made me feel not in control of my campaigns and my strategy.
- ROI would drop into red, simply by targeting it in a specific campaign instead of my smartCPM campaign. For example, I had one placement doing about +50% ROI after spending $40 on traffic from that placement. After running it as a whitelist campaign, ROI dropped to -40%.

Being confused about all this, I decided to pauze my campaigns right after Christmas and come back to this problem later on. Meanwhile, I spoke to a lot of people and asked them for advise on how they would proceed if they were me. Some of them advised me to switch to another traffic source (like Facebook), others told me to start running things more black hat on pop and simply increase my profits that way. Not really sure on what do, I decided to give it some time and think things over.

While looking around on STM last week, I came across Mr Payne's follow along from back in 2016 which made me realize I have never actually looked into using Zeropark as a traffic source. I also read he ran whitelist campaigns there. This made me think about whether this might be the missing link in my scaling strategy (for now). If I'm able to identify profitable placements and simply target those placements in a whitelist campaign, I think I will be able to create some profitable campaigns with all the knowledge that I have now.

I'm wondering what your experience is on running whitelist campaigns on Zeropark? Does it work better than for example at PropellerAds?

Another question: Chrome's Built in ad blocker will be released in about 4 days: How will this affect pop traffic (as we don't use clickable ads to get people to our landing pages)?

Thanks for reading everyone!

Adster

---

Hello,

did you actually try to return back to smart cpm and try a whitelist that way? Reps of ANY traffic source want to make as much $$$ as possible and it's simply more profitable for them to have people using higher fixed CPM bids than utilize the Smart Cpm option ... networks also prefer bigger campaigns targeting tons of placements, because again, it makes them more money than small campaigns that use tiny whitelists. Can't confirm this with any hard evidence, but times and times again, I'd swear that many of my whitelisted campaigns simply started to drop off after a while. Whenever possible, I try to just blacklist the worst placements and keep the rest running. And I use Smart CPM on all networks that have it

As for the ad blocker ... all of us are waiting for the effect this is gonna have ... I'm pretty sure many networks will find a solution sooner or later, maybe some tweaked popunder code will "fix" it ... the problem is the possible action google might take against sites that still serve pops. We will have to wait and see what happens. A decline in volumes is to be expected, at least a temporary one.

---

Wow thanks for checking in! It's been a while!

Matuloo has answered your questions. One thing I want to suggest that you try: Instead of starting a whitelist campaign, just blacklist the bigger placements that are NOT on your list of good-performing placements you wanted to whitelist. This way you'll be mostly targeting what you're wanting to target, but it won't be a whitelist campaign because you'd also be targeting a bunch of smaller placements.

What I've found regarding the way the algorithm on many pop sources works: The more broadly you target (i.e. the more targeting options you include in your campaign), the more traffic you'll get, and I don't mean JUST the sum of the parts! It's like the algorithm will favor your campaign when it's targeting lots of traffic segments.

By doing what I'm suggesting above, sometimes you can "trick" the algorithm into thinking you're wanting to buy more traffic than you actually are - which would get them to assign more traffic to you.




Amy

---

Home > Mobile > Follow-along Campaigns