Home > Mobile > Follow-along Campaigns

---

Engineering --> AM (52)

---

Figured I should make one of these since I already keep a personal journal. Although this will probably be more helpful since I can get feedback from others!

Quick background, I graduated with an engineering degree this year but I'm not interested in doing engineering my whole life (or at all to be honest, but somethings gotta pay the bills for now)...

Focusing on Pops right now since that seems to be the consensus for a beginner. Launched a few campaigns so far to get all the technical stuff working and even made a few conversions so that's a plus! But I haven't had a real strategy. That's why I'm making this.


Today
Found 3 offers: iPhone 7, iPhone 8, iPhone X sweeps in a non-tier 1 GEO. Browsed adplexity and downloaded like 10 landers that I saw were recently used and getting good views. Only testing 3 of them right now for my 3 offers. Took me forever to actually rip those 3 landers because I was having trouble getting the Voluum parameters to work (works now). Also had to edit some images, remove some logos, etc. A $500 campaign isn't worth getting sued by Facebook/Google for pretending to be them/using their copyrighted material IMO.

Here's how I'm structuring this test:
Campaign 1 - iPhone 7 GEO A
Lander 1 (iPhone 7)
Lander 2 (iPhone 7)
Lander 3 (iPhone 7)

Campaign 1 - iPhone 8 GEO A
Lander 1 (iPhone 8)
Lander 2 (iPhone 8)
Lander 3 (iPhone 8)

Campaign 3 - iPhone x GEO A
Lander 1 (iPhone x)
Lander 2 (iPhone x)
Lander 3 (iPhone x)

The landers are all the same for the most part. Only difference is I'm changing the text for each and the main photos. E.g. don't want iPhone 7 photos on the iPhone X landers. But the angle is the same.

I'm spending 10x payout on each campaign and splitting that 50/50 between wifi/3g.

Buying traffic from PopAds for all three. Nothing fancy here, multiplying average bid by 125-150% and splitting up the campaigns between WiFi and 3G.

iPhone 8 campaign is ready, just need to tweak the two others and launch them on PopAds (hopefully by the end of tonight, tomorrow afternoon the latest)

---

What type of engineering did you do? Computer engineering would definitely be a big help ... chemical or civil engineering, probably less so.

---

Mechanical. So not relevant at all... lol

---

1st test done.. and not promising. Spent 10x payout on each offer (although one is still running. looks like there's not too much traffic since it only used 60% of the budget. nonetheless, i'll let it run and see where it goes.)

Here are the results for the three campaigns tested above:


Not much to do here. Onto the next one!

Found 9 similar sweeps offers in a single GEO today. Spent the day working on those. I lumped them up into 3 separate campaigns because one group only allows one carrier, one group allows two carriers, and one group is in a different language.

Something like this:

Campaign 4:
4 offers - 1 carrier

Campaign 5:
3 offers - 2 carriers

Campaign 6:
2 offers - 2 carriers - different language

Dropped the budget for these to 5x payout since the offers are nearly identical. As for the LPs, I'm using 3 different ones in total. No extra customization needed here since the product is the same, but I did translate them for the campaign with a different language.

I'm trying something new with PopAds this time, and not sure if it'll help at all. Instead of bidding 1.25-1.5x the average bid, I'm trying to find the lowest bid that'll get me to the top 5 bidding queue. Makes more sense to get the traffic first, and not fight for the left overs with the other low bidders, no?

Also, can anyone confirm if I'm setting PopAds up correctly with connection targeting? If I want only WiFi, under Connection Speeds I should select everything except cellular/carrier, correct? On the other hand, if I want to get only 3G, I should select Cellular/Carrier AND select the ISP(s) below, right? What if I selected just the ISP but didn't specify the connection speed? E.g. I select Verizon but no connection speed, will that give me a mix of wifi and 3g, or will it give me only that 3g provider?

---

Just here to let you know you're not alone, I dropped out of a civil engineering degree after studying for 3 years to pursue AM. Best decision I ever made
You can do it!

---

Was pretty happy waking up to 16 conversions this morning, until I realized they were all smart link offers lol. Knew something was up when the 9 offers I was testing pay out $0.8 but I only made $0.8 from 16 conversions o_O

For campaign 5, I set it up like this: PopAds with connection type set to Cellular, Connection Speeds set to Cellular/Carrier, and only selected the two carriers in ISPs. The offers are in PH and the two ISPs are Smart and Globe, but in Voluum, the conversions are all coming from Innove Communications, which is actually a subsidiary of Globe. No clue how to filter this since there is no Innove Communications in PopAds. However, 70% of the traffic is still going to Smart/Globe (just no conversions there).

Not sure if I should keep running these campaigns or not. Seems like a waste if I can't target the correct people. I'm definitely thinking of cutting campaign 4 real early because it looks like there is no traffic with those settings. I put a $20 budget for it and it only spent $0.15 today. The 4 offers in that campaign are 3g only, one ISP, smartphone and desktop, but from what I've gathered, SMS billing on desktop doesn't convert too well so I just targeted smartphones.

---

Tested 3 Download campaigns this weekend. Revenue: $0 Spent ~$60. First time doing a download offer, but definitely won't be the last. Like how many different ways there are to promote them. I took 3 popular and fresh landers from adplexity (virus, update browser, and mobile cleaner angles) and tested them. Surprised I didn't even get a conversion.

Spent 10x the payout on each offer (these were a little higher, around $2 each) and split that 50/50 with WiFi and 3G. Traffic came very quick, but thats because my bidding was extremely high. Not doing that again, at least not with an initial offer test. Resorting to the previous 1.25-1.5x the average bid strategy for now.

Found 4 offers this week that I'll be testing. Some have a higher payout ($3-5) but the AM recommended them, so I'll give them a shot, but probably not spend 10x testing since that'll drain my budget quick.

---

Ha! Another mechanical engineer turned affiliate marketer! (I am one also!) Welcome to STM!

Thanks for providing the stats! By the way - to display images, it's a lot better to upload them to imgur.com, grabbing the bbcode links, and pasting them into the post. Forum attachments are so unreliable - don't show up half the time and when they do, appear small.

Hmmm! Lander CTRs are low. I need to know which geo this is. For geos like AU and UK for example, such CTRs are common. For many developing countries though, these are dreadfully low.

I know - I'm the one who's always saying not to focus on CTR. But when all CTRs are low, it tells a consistent story.

Also - 3 landers are hardly enough. I would suggest 5-10. And have you optimized the landers for speed? <2s is the bare minimum, better to aim for <1s.


Actually - it would be better to target one carrier per campaign, because prices of each carrier will vary, so that by targeting them separately, you can avoid overbidding for one carrier while underbidding for the other.

It's nice to see the kind of thinking you put into your planning though!


Right now your priority is to test a bunch of landers and identify the best of the batch, so that you can then use this best lander to mass-test offers to find gems.

While testing landers though, you need conversions to reach statistical significance so you can cut landers down to a winner. Therefore you need an offer that converts not-too-badly in order to avoid racking up too much cost while cutting landers.

So what I'd do is: Ask your AMs on various networks for 3-5 offers for the same geo+carrier+whatever else, that have converted reasonably well for at least a few other affiliates. Throw 5-10 landers into the same camp with these 3-5 offers, run until one of the offers starts converting - say until one of the offer reaches 2 conversions. Then just pause all other offers, and use JUST that offer to cut landers down to a winner.


Can be a good idea yes! Depends on what you're trying to achieve.

Testing higher bids though is DEFINITELY recommended after you've identified a promising offer+lander (after extensive testing). Different bids will trigger traffic from different placements. Because after a while, your competition will all know which placements convert better than others, so will all bid higher for more traffic from those placements, if you don't bid higher, you'd never even get access to traffic from those placements.

One approach would be to cut placements in your original camp first, then clone the camp, set a higher bid, run and cut more placements, then clone the camp again and set a higher bid and cut more placements....and so on.

Try testing bids in $0.50-1 increments, for 3-10 bids. Don't be afraid to continue testing to high bids - as long as you set a low budget to avoid accidents, you'd be fine. Sometimes you'll find gem placements only when you bid really high, but the high CR would be worth the high bid.


Right in general. But when it comes to the way they handle carrier targeting, popads is a bit screwy, so you can never be sure.


Again - there's no telling for sure with PopAds. Either scenario would be possible. The best way is to run a bit of traffic to do a test:

https://stmforum.com/forum/showthrea...esting!-Part-1


Do let me know how your camps do!



Amy

---

Nice. More engineers!

Haven't updated this thread in a few days, been launching new campaigns everyday, just nothing is sticking. I'll add more landers to the next campaigns and see how that does. Will be back with numbers!

I've also noticed that my campaigns tend to generally run for just a few hours. Seems like a bad way of testing an offer, no? Although, I've tried using legacy bid and that changed almost nothing, and PopAds has a minimum for throttling a budget, which makes this impossible to do for low payout offers.

---

CPA offers are a dime a dozen. You're right that we'd get more accurate results by running continuously for a few days or even a day instead of a few hours, but the trade-off is more time and/or more money. IMO, unless you have it on very good authority that you have a golden offer on your hands, spending that extra time/money may not be worth it. But of course that would be up to you to decide - everybody runs with a different style.

When you say nothing is sticking, what do you mean exactly? Don't hesitate to share what you feel are "failure" camps - you've spent the money, so let's see if we could learn something from them!

Also - don't be discouraged when the camp isn't making money during the lander-testing stage - it's usually when you've cut down to the last lander and start testing more offers with it, that you'll see the real potential. Not to mention that are still other optimization tasks you can do to increase ROI: Cutting placements etc., testing bids....

Feel free to post some info on those camps so we could do a post-mortem. And you never know - there may still be a faint heartbeat going in some of these that we may be able to revive!




Amy

---

Good point. I agree, moving quick is definitely important here.


What I meant by that is that I'm not getting any conversions. Not expecting to be green with an initial test, but I am hoping to get at least 1 conversion so I can test it further. Otherwise, what's there to optimize?

Working on a campaign right now where I'm testing 4 iPhone 7 offers in the same country with 6 landers. Made a separate campaign for each carrier on PopAds, but I fear that some of those carriers will not have enough traffic. I'll report back when it's done!




Here's a campaign I ran over a week ago, but dropped it (maybe too early?)
It's a South Africa iPhone X sweeps offer that the AM said was doing very well.

Campaign results:


Lander results:



CTR seems very low. I host my landers on the Amazon CDN so load times are pretty quick, don't think that could be an issue. I always test the offer links and what not to make sure they work before I launch the campaign too. Looking back at this one, I'd test more landers and let it run for longer. Can't really remember why I cut it so early, but that's something I won't be doing again.

---

Not only that, but mainly because there are so many such offers, and not many are good-converters, that if you were to spend too much on testing each offer, you won't be able to make back your losses with the winners you do find.


PopAds has good quality traffic but lacking in traffic volume for many carriers. This makes it a good source to do testing in, but you should make sure that there's enough traffic volume on other networks that you can scale to. If you can't confirm sufficient volume, don't even start testing - trying to work with little bits of traffic would be far from satisfying.


Eek! ZA! Not the ideal geo for newbies.

It's one of those geos that has a lot of potential to be very lucrative, but traffic is expensive and you'll also need to be prepared to cut a ton of placements before you can make the geo profitable.

I don't know which traffic source you had the camp set up in, but next time you attempt this geo, start on Propeller - use Zeropark for scaling - don't start at ZP as it will take a lot of cutting, so you'll want to only put your best offer+lander there to minimize your investment.

If you're not prepared to invest a lot of money into identifying bad placements, stick with developing countries for now (i.e. tier 3/4s).

Your ZA camp actually did not look hopeless - I would have run further for sure. But like I said, it may be best to leave this geo for later, when you have more experience on optimizing camps.

Thanks for posting stats!




Amy

---

Hmm, I'll check out PropellerAds in that case. Really not a fan of the ISP targeting options on PopAds anyway.

I ran that ZA offer since I read Matuloo's blog post on starting with tier 2/3 geos, and in his list ZA was a tier 2. But I see why others would say it's a tier 1 geo and why it's best to avoid it right now.

Side note, applied to Clickdealer since I see a lot of their offers on adplexity, but they rejected me Can't blame 'em though. Need to make more money first.

And the most important part, here's an update on that campaign from yesterday.

Tested 4 offers with 7 LPs. Same geo, carriers, etc for all offers. I split the campaigns up into 3 separate ones in PopAds (carrier 1, carrier 2, and WiFi).

Offers


Landing Pages



As I suspected, carrier 1 and 2 barely got any traffic. Both conversions came from WiFi. Since I actually got some conversions, I'm gonna throw more money at this tomorrow, but just at the WiFi campaign since the others have no traffic.

---

Sounds like a good decision! But what are your bids like? Would like to remind you that your bid for carrier traffic shouldn't be too low - unlike for wifi camps, if you bid too low you won't get much traffic at all.

And stats are still looking very promising at this point! Looking forward to seeing you cut down to a winning lander - and then using that to test more offers.



Amy

---

I was bidding 1.5x the average bid, but even with the bid set to $1000cpm, Popads said the daily impressions would be under 1k. Interestingly enough, I got 1 conversion from one of those carriers before I shut down those campaigns. I only spent like $1-2 on it too. Should I try them out again, maybe with a lower budget? I'm getting better results with the WiFi campaign so not sure if it's even worth testing the carrier.


Here's the results for the past 2 days:

Offers



Landers



Here's the plan for the next campaign...

The offer with 0 conversions will be cut, since it has a 6% chance of beating the current best offer. The other 3 will stay.

No point in cutting landers right now since there isn't enough data. Same goes for everything else.

So I'll spend another $20 on it tomorrow and see what happens. Hopefully I can narrow it down to 1/2 offers at least.

---

I wouldn't even bother setting up a campaign for that kind of traffic volume.

For wifi, be sure to use caurmen's newest bot detection method to weed out the worst placements before doing other optimization.


Using conversions resulting from different offers to cut landers, is not recommended.

When we use the stats calculator at peakconversion (or any other calculator that can be used for this purpose, I'd imagine), it is assumed that every conversion is the same in nature. So if the conversions are from different offers, then the verdict you get would be inaccurate.

What I would do instead, is run a few offers at the same time (which you're doing), wait for one of the offers to reach 2 conversions, then pause the rest of the offers and just run that offer - and use the conversions from just that one offer to cut landers.

Basically you already have 2 offers that fit that bill - so just pick one of them to use, to keep running and cutting landers.

The idea is to use an offer that converts well enough to give you the conversions you need to cut landers, without costing a fortune in the cutting process.

Once you have a winning lander, you can retest the other offers, and/or new offers.




Amy

---

I jumped ahead of myself and forgot why I was even doing this test in the first place lol. Yeah, no cutting landers at this point. Woke up and one of the offers had 3 conversions while the other 2 had 2 and 1, so I went with that offer to test the landers.

Also, I read about the bot test. Here are the results from today: only $6 was spent (took a bit to get approval on popads, by that time the GEO was sleeping :/



Top one is the bot offer (Parses Javascript) and bottom is the real offer.
Looks like most traffic is bot traffic (if this is setup correctly). Does this look right to you guys? High CTR makes sense for bot traffic, and the CTR is about the same with the actual offer as it was in the initial test.

Edit: Guess I had it the other way around. 56.8% are real people, and ~40% is bot traffic. Think I'll blacklist any placements where there is more than 20% bot traffic

Edit 2: maybe that's a bad idea since most of these placements have >20% bot traffic

---

Updates for this week:

Ended up not cutting any of those bot test placements since most of the highest traffic ones were getting between 30-60% bot clicks. Not sure when a placement should be cut, considering you could be profitable even with a high bot %, no?

Threw more money into the 1 offer that had the most conversions. Now it's not converting. Here are the results for the past 3 days:




There was actually only one conversion for this offer. The other was for a smartlink offer since the payout was different. Not sure how that happened--I checked where the conversion came from, and it's within the restrictions set by the offer.

Not sure why it's not converting now. $70 is a lot for one conversion, considering I spent 1/4 of that in the initial test and got 3 conversions. I double checked to make sure my LP and settings were correct, and they were. Went through the tracking link and through each LP and they all went to the offer.

I did notice that 4/6 of the converting placements from the initial test were not in this new campaign. 2 were, but 1 of them only received 1 view and the other didn't get much traffic either(only $0.50's worth).


Decided today to move on from it, test the 2nd best offer from the initial batch. Just launched that campaign so we'll see where it goes.


Questions
I had some suspicions for the cause of this poor performance, but not sure if they have any backing.

Does making a new campaign on Voluum and PopAds affect conversions? Sort of like how editing an existing adset on FB can skew your results, I was wondering if something like that happens on PopAds? Or if creating a new campaign on Voluum and using that new link can do the same? I'm 99% sure it doesn't, but it's something that popped up in my mind.

Similar to that, I wonder if it's bad to change a bid that is already converting? In the initial test my bid was $2.3 and I got 6 conversions, but for the single offer test, I bumped it up to $4.3 and got 0 conversions. Did that to get data quicker since it was running a little slow.

Lastly, Primespot or no? I think Finch recommended always using primespot. Makes sense to me.


Current Tasks:
- Test the 2nd best offer, and see if that works better than the first.
- Found 2 other offers I want to test. They're DL offers. Got 5 landers and just need to translate them now.

---

I think a good start would be to blacklist placements that have 70-80% bot traffic. Then, run some traffic and see. If you're getting a ton of traffic and in really negative ROI, and want to cut fast, then also cut the 50-60% bot traffic placements. Otherwise, keep running those to see what the overall trend is. If after a while you're finding that most of the 50-60% bot placements are red, you can blacklist all of those then. I've found that some of the best-converting placements could contain up to 50-60% bot traffic, so you never know until you test.



Amy

---

This is the first thing I check if CR takes a plunge. Try increasing the bid to see if you'll get some of that traffic back.

Yeah it's brutal how much you spent and not get many conversions...once you're down to one lander, testing more offers may uncover some gems.


Well - when you set up campaigns on different days, even if the targeting is exactly the same, the competition levels will be different, so it would be normal to get different results.

Other than that, I'm not sure if that's the cause of the dive in performance. It could be a number of things, and not just the traffic source (i.e. the offer may be the cause too).

And regarding the changing of bids - by all means test the $2.3 bid again. Sometimes bidding more doesn't mean you'll get more traffic, I'm not sure why either.

When a bid is converting, it would be best to just continue testing to finish your current split-test (landers/offers). Then once you have a winning lander+offer+targeting, you can clone the camp a few times to test different bids.


He did?

Prime spot traffic is more expensive though.

Non prime spot traffic may not convert as well, but it's also cheaper. Often, conversion rates are what drives the prices, so you'd get what you pay for.





Amy

---

Back with some updates on the past week. I stopped that campaign that I've been working on in the past few posts. Some of the offers got paused and the others that I tested tanked.

Here's the summary for the whole campaign



So as you know I was split testing 4 iPhone 7 offers there. In the initial test 1 of them got 3 conversions so I decided to keep testing that to find a winning lander. No clue what happened, but it didn't perform well at all when put into its own campaign. After a $70 spend I decided to try the 2nd best offer from the initial test, which had 2 conversions. After $85 spend, I decided to stop it as well because I just wasn't getting conversions. Funny how the initial test yielded 6 conversions at 1/5th the cost of the two other campaigns lol. Is what it is.

Analysis
After $150+ spend, I didn't get enough useful information. I dropped the campaign and have moved onto a new one. Don't think anything technical was the cause of the bad performance. Clicks discrepancy between Mobidea and Voluum were what you'd expect it to be. I know I need to optimize LPs more and reduce the speed. I'm hosting them on Amazon's CDN, but some LPs still take 2-3 seconds to load. While that will help get a higher CTR, it's most likely not the reason for low conversions here.

Landing pages were ripped from AdPlexity. I looked for the highest hits in the last 7 days and used 7 of 'em.

Offers were not proven to be winners for others (not enough data according to the AM).

So it's either the offers and/or the landing pages. I'm thinking it's both. So I am going to get a bit more creative with the landers I rip. Not going to make them from scratch though (yet).

Objectives
1. Spend a little more time optimizing landers to reduce speed (initially. When I find one that is the best, then I'll perfect it)
2. Get more creative with landers. Try new angles. Combine different landers.
3. Keep testing offers


Side note: This marks a full month that I've been doing this. Results? Something like a -95% ROI. No biggie. I was expecting it to be like this. Lets see how the next month goes

Few notes-to-self:
1. Stop wasting more money than necessary. Offer not looking promising at all after 10x payout? Cut it and move on.
2. Differentiate. Copy and pasting landing pages won't work. Use the copywriting skills you've picked up in the past few years.
3. Work with a sense of urgency. Get more done in less time.
4. Focus on the big picture stuff. Stuff that drives the bottom line.

---

11/9 - 11/11 Test
The AM at Mobidea said there was an iPhone X offer that was doing very well in Singapore, so I tested it out. Funny enough, I tried this offer a few weeks ago (had no idea it was doing well, was maybe my 5th offer tested at that point) and I had something like a -80% ROI direct linking it. Obvious mistake made there.

Campaign Structure
Ripped 5 landers from AdPlexity and changed images/angles on all of them. Main thing I focused on was making them appear more believable. Tried doing that by localizing them--finding stuff that is big only in Singapore and revolving the lander theme around that. The landers don't look too great, nor are they perfectly coded--will devote more time to that when I know they are converting.

Only targeted WiFi since there isn't much traffic for the carriers.

Tried bidding 2x the avg bid. Avg was $0.001.

Results (Landers)




Looks like bidding high gets you a ton of traffic, but you'll need to get rid of the bad placements. The highest cost placements seem to have the fewest clicks.


On the flip side, the placements that have received the most clicks didn't spend too much. I looked back at the bot test I ran last time and the placements you see in the next pic had 40+% CTR in the bot test. Most were above 60, but the lowest was 40%. So they're doing ok in terms of bot clicks.



Most likely the offer/landers were the cause of bad results, but I've never bid 2x the avg before, so I thought the result for the placements was interesting. Interestingly enough, in the last test (iPhone 7 offers just a few posts above) I was bidding at minimum $2.3, which was only 1.5x the average, and I was getting less traffic.



Tasks
Found 3 new iPhone X offers in a different GEO. That's currently the focus. Will use these same landers, but with some tweaks.

---

The only advice I can offer, being a beginner myself, is don't change the landers unless you are more experienced. I mean assuming you are ripping tried and tested landers which are running for multiple days and doing volume, there is no reason to try modify them. Or you can modify them and split test with original but probably better to stick with what experienced affiliates are using.

Also try cut camps sooner. I've done similar mistakes with letting an offer get to $70 spend and only making $2 from it but I don't think it serves any purpose. Either limit your offers or the landers you are testing to bring down the required budget for a camp.

---

Hmm. I guess I'm skipping a step here. I just figured the top landers would be beaten to death and by the time someone like me gets around to using them, most people on the placements would have already seen those landers, which means I'd be fighting for the small amount of people that haven't. Though I'm sure there's more to it than this.

Agreed. Just annoying when an offer is getting ok results one day so you throw more money at it thinking it'll continue like that, or get better, but ends up tanking lol.

---

What did you mean when you said "it didn't perform well at all when put into its own campaign"? You were running all landers and offers in the same campaign, correct? If not, then you should have.

As for campaigns performing better in the beginning - yes that can sometimes happen. If you're finding that conversions just stopped coming in, check to see if you're still getting the same traffic volume from the placements that were giving you conversions before - if so, increase your bid to get some of that traffic back. Also, sometimes just running for another day will fix it, and the conversions would come back. Testing more offers would be another potential solution. I know it's frustrating that this happens, and this phenomenon seems to be happening more and more now. The only thing we can do is keep trying until something "sticks'. As of now, I don't have a definite solution to combating this type of volatility. Analyzing stats and doing what makes sense is only part of the equation - and then there's always an element of luck involved.

Next time, don't spend so much money on one offer. Basically the aim is to find an offer that converts decent enough to help you cut landers down to a winner, without breaking the bank in the process. I would say that if an offer isn't converting at least once per 10x payout, then you should pause it, because it wouldn't be good to use for cutting landers. Consider this: If you're testing 4 landers (I'm using your example), then you'll need at least 6-10 (I'm estimating) conversions in order to cut down to a winning lander, depending on the distribution of conversions across landers. So if you're using an offer that converts once per 10x payout, you'd need to spend 60-100x payout just on identifying a "winning" lander. And the "winner" is just the best out of the 4, and doesn't necessarily mean it will convert very well (unlike if you're including 5-10 landers - that would give you a better chance of finding a good lander). So the 10x payout rule would be appropriate - my personal opinion.


DO spend some time to bring the loading speed down to <2s, preferrably <1s.

And do use a fast DNS service for your landing page domain. That will shave some time off the loading time as well.


You are absolutely correct!

There are 3 main components of a campaign: Traffic, offers, and landers.

Traffic: It would be appropriate to bid at least average to ensure you're not getting junk traffic.

Offers: It would be better to start with "proven" offers, e.g. offers that AMs are telling you multiple other affiliates have run with decent conversion rates, or offers doing high volume according to Adplexity.

Landers: Try to include 5-10 landers for the best chance of having at least one good one in the batch. Make sure the loading speed is reasonable. Manually click through each lander to make sure all the functionality is working. Use browserstack or similar to make sure landers appear fine on all mobile devices.

Do all of this to give your campaign the best chance of succeeding, then hope for the best. As you were implying, if both your offers and landers are unproven, then when you don't see conversions, you really wouldn't know which one is the weak link (it could very well be both!)


Sounds like a good plan!

It's always rough in the beginning, when you don't yet have a good testing approach, and don't know when to stop testing something or stop running a campaign altogether. This is where experience comes in - it will get better.


Is it just an eerie coincidence that we both mentioned 10x payout as a rule-of-thumb? Very nice - I love coincidences!

Actually - copying landing pages as-is DOES work, and it would be a perfectly good approach to test ripped landers first to find a winner, and then create variations of the winner and test those. Then of course you can make completely original landers (it would be tricky keeping under the radar though, with all these spy tools around...)

Other than wanting to point that out - you've hit the nail on the head with everything you said. The fact that you're able to make these observations and come to certain conclusions/decision, means all that money wasn't spent in vain. And you WILL learn to spend money more efficiently over time (like you've resolved to do!)




Amy

---

Changing landers is a very good thing to do - but it may be better to start by testing the originally ripped landers as they are.

Remember our discussion in my last post, on minimizing the variables to give the campaign the best chances of success? Well, using tried and tested landers will be a part of that effort. If you change all landers from the start, there would be no guarantees that your new angles will work. So we're back where we started: If you don't get good performance, you won't know whether it's the offers or the landers.

So it would be better to test originally-ripped landers in the first round of lander testing, and then for the second round of testing, test the winner against landers with your new angles.

You're definitely on the right track with localizing! I've had great results with this approach. Check out google trends and/or similar for the target country to see what's hot at the moment.


Yup - wifi = expect some bot traffic.

If you're not sure how to set a cut-off percentage for bot traffic (e.g. cut all placements with xx%+ bot traffic), try to err on the low side if you're getting a lot of traffic (i.e. set the cut-off to a lower percentage), or if you're not getting a lot of traffic, try to err on the high side (i.e. set the cut-off to a higher percentage). You can always decide to either cut more or retest some placements later.




Amy

---

The approach I've been taking goes like this: I setup one Voluum campaign with several offers and all my landers. I let it run until an offer hits 2 conversions. Then, I make a new Voluum campaign with just that offer and all of my landers. Are you suggesting to stick with one Voluum campaign even after the initial test, and remove the offers/landers that I'm not testing?

Hmm, so I noticed that when I do a speed test for a LP, the first test will usually be ~3s, but then I hit "re-run" (webpagetest.org) and the speed gets knocked down to usually under 1s, which would make more sense for a CDN. Any ideas for why this happens? Which one is the true speed?

Yup, that's a good point. I'll stick with the stuff that works at first. Keep overthinking stuff and jumping ahead of myself!


And thanks a lot for the help so far! It's really speeding up the learning curve.

---

This is likely related to caching. The first test will be the most accurate and the one you want to optimise for. Webpagetest have a bit on caching here.

---

Yes! There's no benefit to starting a separate campaign.

Just remember that when examining lander stats to decide which ones to cut, you'll need to drill down to offer stats, and base decision on the conversions for JUST that one offer.


The first test is what we should go by.

The way I understand it is that some of the page contents will get cached (or something...) which will make it faster to load when the visitor accesses it a second time. Since pop ads are what we're dealing with and not a "real" website with repeat visitors, we need to make the first test time fast.




Amy

EDIT:

LOL perfect thanks Roland! I had this post opened since yesterday - forgot to refresh before I replied, or would have seen your reply before replying with the same answer haha.

---

Ah, that's what I had in mind. Makes sense now. Really need to work on some of these landing pages in that case.

Ah I see. I'll start doing this for future tests. Starting to think that not only is there no real benefit to it, but it's actually harmful lol! (Ok maybe not literally.. but maybe it's bad luck for me) I am currently testing 3 offers right now, and the best offer ended up at -30% ROI in the initial test. So I made a new test with its own campaign (before I read your post) and it got 0 conversions! Stop teasing me PopAds. Anyway, I'll be back in the next post with some numbers here.

---

So I was reading some threads on here on how to lower LP speed, and I came across this:

https://stmforum.com/forum/showthrea...uent-Load-Test

Now I'm confused lol.

---

Nice! Another engineer
One of my professors said that you're not going to use 90-95% of what you learn in school anyways, school is more about training your mind to work and solve problems a certain way. I did Industrial Engineering but always felt like I didn't want to work in that specific field. If you are happier in this industry then by all means stick with it. Your education and experience should help.

Best advice that I could give is - don't look at the ROI of campaigns. Look at efficiency instead.
If you spend $100 and get $70 in conversions, then your campaign is running at 70% efficiency. Keep testing and optimizing to reach over 100% efficiency, and then you scale.
Basically the point is to restrict ad spend when you are running at less then 100% efficiency. During this time test as fast as you can. When you reach over 100% you can scale volume while also testing ways to keep improving efficiency.

Looking at the ROI # of campaigns can be disheartening and generally not as useful. A big part of this game for many people is mental fortitude, so you want to use as many of these little tricks as you can.

---

I was told something similar haha. I barely remember what I learned just this last semester!

I like your approach--interesting way of looking at it.

---

I just had a brief discussion with Caurmen in attempt to clear this up once and for all - and this is his wisdom:



So - for a new lander, run the speed test a few times to make sure that the CDN has already stored copies of your landers on servers that are at/near the test locations. After that, the "initial load" should be close to what your visitors experience.

And taking the average of 3 loads as suggested by Caurmen sounds very sensible as well.

Any way you look at it though - the faster the load speed, the better. So I would suggest to aim for the initial load time to be under 1 second if possible, instead of using the subsequent load time as an excuse to avoid optimizing for better lander speed.

Thanks for your question! I've learned from our exchange.




Amy

---

One more small note on this: make sure if you're using Pingdom that you use the same location to test from each time.

Otherwise your results will be invalid.

---

Haha, sorry about that! Actually just saw that thread and was just about to update you in here but glad to see it's already been done.

I figured it was browser caching for the visitor, so their subsequent visits would be faster, I've also learnt from this. caurmen will definitely know this better than me.

---

Thanks for clarifying above.


Quick update: Found an offer that's converting decently so far. I started with 7 landers and now I'm down to 3 (cutting via peak conversion of course). These last 3 are performing very similar. Not sure how much I'll need to spend to determine which one is the best, but at this rate, it's not looking like it'll be cheap lol


Edit: Here's how they look so far.


Just read Caurmens post about this. The weakest lander isn't quite as good as the top 2, so I'm going to spend another ~$20 and see if I can cut it in this next test. Otherwise, it looks like it makes the most sense to pick one of the top 2 landers and split test that one against variations of itself. Will be back with some numbers

---

Yes! If multiple landers keep going head-to-head, then it wouldn't add too much value to keep running them just to get a winner.

You can pick one and start the next round of split-tests (of either landers or offers).

Or, if the results are good enough, and you've decided to stop split-testing for now, then it wouldn't hurt to keep the multiple landers in rotation, just to vary things up a bit to delay burnout (although, with landers, the effects of burn out is normally not apparent over short periods).

Looking forward to those numbers!



Amy

---

Back!

Here's how this offer performed, since day 1, including all the lander testing.



First row is the original bid (1.3x avg), second is a high bid (2.5x avg), and 3rd is a low bid (0.5x avg). High/low bids got no conversions, even after ~10x spend. I only started using tiered bids after I cut down to the best lander.

Best lander result:


Interestingly enough, it started off at like -30% ROI for like 2 days, then dropped a lot. The next two best landers performed similarly.

Once I settled down on the that lander, I did a drilldown report into it and looked at how it performed since day 1. Here's how it went:

Devices


I cut Samsung models because that accounted for 1/4 spend with 0 revenue. Kept Apple, even though it was unprofitable, since that's where the majority of traffic came from.

ISP


Didn't do anything with these. I drilled down further into the -38% ISP and found that I had a +40% ROI with Android, but the traffic was very low, <1000 impressions per day on PopAds, so it didn't make sense. Maybe I should've cut the 3rd and 6th row? Although that wouldn't really bring the -70% campaign ROI up much. I didn't cut the top 2 rows since that's where the bulk of traffic is coming from.

OS Version

iOS 11.0/Android 7.0 don't look too good. I cut both of them, but realized that cutting Android 7.0 was a mistake. Looked further into it, and about 90% of Android 7.0 traffic was from Samsung models, which I know was doing -100% ROI. The other Android 7.0 models/brands hadn't spent enough to be determined as losers yet, so I kept Android 7.0 but still blacklisted Samsung models. iOS 11.0 is a clear loser though. ~10% spend, 0 revenue.

Placements




Not much to do here. Not enough spend for each. I did the bot test and got rid of the ones with >80% bot traffic.

Analysis
I'm going to stop testing this offer. Even though it was the best one in the initial test, I don't know if a -70% ROI (or 30% efficiency) is worth the time/money. I've got several offers that I'm going to test with the best lander so I'm going to get started on those. Launched one just this morning, so we'll so how that goes.

Several things I noticed about this campaign:
- I lost track of spending. It's the first offer that actually converted for me, so I just kept throwing money at the test. It kept converting at least once per 10x payout, so I didn't want to cut it. But, several days I went way over my daily budget since I wanted data quick. I now know to keep an eye on this in the future.

- If the best lander is -70% ROI, is there even a point in optimizing the campaign? I couldn't really find any pockets of profitability with this one.

- Some of the cuts I made were too aggressive as they reduced traffic significantly. I'm guessing this comes from removing iOS 11.0 and Samsung phones. Although, I looked at PopAds to see what the estimated traffic would be, and it was only a few thousand less with these cuts, but then I checked a bit later and the traffic was down to like 3k impressions/day with the new targeting. Just something to take note of for future references.

In short, not a winner, but at least I finally broke through the plateau of testing offers with 0 conversions lol.

---

Yup that would be the best next step!

You needed a converting offer to help you cut down to the best lander. But once you have that, start split-testing new offers with that control offer right away.


Yes you did overspend, but that was a good lesson.

It's OK to lose money when cutting down to the best lander, but once you have a decent lander, there's no reason to not start mass-testing offers. It's so much easier to use statistical tools to cut inferior offers, than to have to make a call on when to stop running a single offer. This is why it's better to test batches of offers rather than a single one.


Depends. If you're operating in a big geo with lots and lots of placements, such that you can cut a ton of placements and STILL have lots of profits left over to recoup your losses with, then maybe.

BUT, there would be no reason to have to resort to that BEFORE mass-testing offers first to find the best one. It's so much better to invest your money into testing offers than into identifying placements to cut - would be less expensive plus you'd have more profits in the end (because finding a better offer will allow you to make more of the traffic profitable, whereas cutting placements will just increasingly restrict your traffic volume and therefore profits potential).


Again, cutting your way to profits should only be used as a last resort IF even that - and only after you've tested a good number of landers and offers.

It's infinitely more satisfying to improve your funnel so you can make more profits out of the existing traffic, compared to having to cut out big chunks just to have a minute green camp on your hands that isn't worth the 5 minutes/day you'll need to spend checking stats everyday for it.


This is part of the learning process. The more "mistakes" you make, the more lessons are learned. And the more you test, the faster you "fail", and the faster you'll gain experience.

You're doing exceedingly well here.



Amy

---

Still here! Just got a little lazy with this.

I've been testing a lot of iPhone offers with the lander from the last post. The first few campaigns were doing around -50 to -70% ROI, so I cut them because if they're performing that poorly with the best lander, it doesn't seem like it'd be worth the time cutting small stuff.

The current campaign is a bit different. I started off with 4 offers from 2 different networks. I cut down to the best offer using stat sig, and for the past 2 days I've been testing that one offer.
Here are the results:

Yesterday:


The whole campaign was at -20% ROI, but I noticed that all the conversions came from Android, so I thought that this looked promising! I cut iOS and even with that, which didn't affect the traffic volume too much (this is a larger GEO. I know it's been suggested to avoid these, but I've been running out of offers to test on the smaller GEOs, and some of the medium/larger ones have allowed me to do testing quicker).

So I removed iOS on PopAds and increased the budget from $10/day to $30/day, and started the next round of testing.

Today:


And bam! The usual. Offer tanks. I double checked to see if I accidentally changed something in Voluum and PopAds, but I confirmed that I did not. The only difference is that I increased the budget, and I blacklisted iOS. Still plenty of traffic volume, and I am getting roughly the same amount of traffic from the converting placements from yesterday. I'm starting to think that increasing the daily budget significantly screws up your results, like on Facebook. Maybe I need to duplicate the campaigns on PopAds with the same bid and daily spend? I don't know, this probably isn't it, just a theory though.

Maybe it's the fact that I cut iOS. Did I cut it too early?

Or, could just be a bad day. The problem with that is knowing when to stop. If I spend another $50 tomorrow, I don't know if it will perform like yesterday (80% ROI) or like today (-80% ROI)!

On the plus side, I now have a better strategy for testing offers. Starting off with multiple bids seems to be working great, since there is generally a huge difference between the different bids. Before I'd use just 1 bid. The staggered bid approach is better IMO and I'm going to keep using this. Great suggestion Amy!

---

Pop camps are just volatile - the tank in performance could happen without you changing anything. You can try some or all of the following:

-Duplicate your new campaign (where you've blacklisted IOS) and see if it would do better.

-Unpause your original campaign and try to get from -20% ROI to 30%+ ROI via cutting placements (instead of cutting IOS).

-Start a new campaign to just target Android OS alone.

Also - sometimes, when performance tanks, just pausing the campaign for 1-2 days and unpausing it can turn it around. Worth a try.


I don't think that was it, but then you can never be sure - pop camps are just volatile and sometimes you can't find a cause for the effect.


That is another thing I had wanted to suggest: Sometimes if you'd just bear with the loss for one day and keep running, the next day can become normal again.


Very glad you had good experience with the approach! Most people don't realize how important testing bids is - and now you do!



Amy

---

Need to stop slacking off here! Just been quite a rollercoaster lately. But, I've been working on a campaign for the past few days that has shown some promise, so I'll document that in here.

Started off the campaign with 3 offers in the same geo (2 of which were actually the same, but just had different offer pages from the advertiser) and one lander from a previous sweeps campaign. Day 1 I was able to remove 2 offers since there was a clear winner.

12/10
I started to test the best offer + lander combo on that day, but before I began throwing money at it, I saw another lander that I wanted to test, so I threw it into the campaign with the previous best one.
Here's how it performed that day:


Needless to say, it was obvious who the winner was. The new lander that I added ended up killing the previous best one. Plus the offer was at -25% that day with one lander that was at break even, so I knew I should keep testing this.


12/11
For the second test, I wanted to see if I could improve the lander some more, so I made 3 more angles to split test. Added them to the campaign and threw about $30 for the test and came out with this:


Wasn't able to cut anything with this, so I made no changes to the campaign. However, the whole offer ROI did drop from -24% to -70% which sucks. Weekends might just be better for this offer.

I was also able to blacklist two placements that were at a loss of 2x the payout. Hardly put a dent in the traffic volume since this is a big geo.

12/12
Woke up this morning and did a very small test before heading off to work.



Was able to cut 2 landers from this since they had <10% chance of beating the best lander. Then, I tested the final 3 landers at the normal testing time with $30. Here's how that ended up:



So based off of that, I can cut LP3, which leaves me with LP3 v2 C and LP3 v2 A. So going into the future, my plan is to finish up this angle test and determine whether variation A or C is better, and then do one final lander test--the winner of the current test vs. itself, but translated to the native language in that country (current landers are in english). I'll also keep looking out for placements that are in 2x loss. However, I plan on retesting these blacklisted placements when I get the best funnel. Haven't cut anything else since I'm still testing the best lander, but here are some screenshots of the major traffic segment's performances since the start of this lander test:

Carriers



OS


So in a nutshell, the plan going forward looks like this:

Keep cutting placements in 2x payout loss
Test best lander angle
Test best lander language
Observe major traffic segments and if none are green or close to break even at this point, then drop the campaign
Otherwise, retest the other bids with the best funnel (best lander + highest ROI major traffic segments)
Retest blacklisted placements

Noticing that the performance seems to be declining each day. Started off great (-24% ROI) but now it's in the -50 to -70% ROI region. I played around with some of the numbers, and if I can double the CTR or CR, I would be at BE with this payout and CPM. So, hopefully the best lander can achieve it--so far I've noticed the CR has gone up from 1 to 4/5% which is great. CTR overall is definitely low, as you can tell, since it is a more competitive geo. However, the landing page speed is around 0.500ms so I don't think the speed is an issue.

---

Hi!

So the lander that was performing good in the beginning (LP3) is with a different angle than LP3 v2 C? It's interesting how dropped its performance. I'd actually split the campaign and test the angles separately, so that you don't loose your good performing creative. And I'm sure having landers not in English but in native language of your targeted country is going to increase the performance, curious to see it!

In terms of weekends - did you check the time period stats? How is it going, maybe there is some obvious hours to make dayparting?

Your plan on things to do next look good, to add - I think you can already cut all the OS except Android and iOS, as well as carriers with -100% ROI.

Hope it helps!

---

Your plan sounds good, but be sure to also test bids. That can make a big difference.

It's common for pop camps to do better on the first day, then experience a decline in performance. The trick is to make the campaign work in spite of the drop in performance, and you'd do this by optimizing.

And since this is a big geo, there's potential to cut a significant amount of placements/segments and still end up with profits that are worth your efforts. I definitely agree that there's a chance of turning a profit with further testing.

Once you have your best lander, test 3-5 bids and see what happens.



Amy

---

Yup, LP3 is a "spin to win" lander, and the others are just variations of it. Can you explain more about putting the landers in separate campaigns and testing that way as opposed to doing a split test in one campaign?

I think the native language will definitely boost the performance, but I wanted to save that for last so I could save on translation costs.

The offer that I'm currently testing is actually a new offer, so the support at Mobidea said there is no data on it. But when I was split testing this offer with some others, I looked at what hours of the day those offers were converting the most and have been using those hours to test. It's mainly 7pm-9pm, though most conversions have came in between 7-8pm. $30-$40 budget gets spent pretty quickly. I've been trying to spend a bit in the late morning/early afternoon, like $5-$10, but still most conversions come in the evening.


Definitely won't neglect this! Already seen from a few previous offers how big of a difference the bid makes.



----------------------------------------------------------------------------------------------------------------------


Thanks to both of you for your feedback.

Today I tested the offer with another $40 and was hoping to cut down to the best angle, but didn't get much data today.

12/13



Only things I'll do is blacklist 2 placements that are now at 2x loss and blacklist the -100% ISPs and OSs (as suggested above). Hopefully tomorrow is a better day!

---

Hmm...I don't know about that!

The definition of "split-testing" is when you subject all test candidates to the same set of conditions. And testing landers in multiple campaigns is NOT that.

But let's see what advice Mobidea has on this.


Why not just rip a similar lander that is ALREADY in the native language? One that has the same look-and-feel.

Another thing to try is using google translate. It has gotten a TON better since 1-2 years ago.

To maximize accuracy, rewrite all the text into short and direct sentences, and rewrite figures of speech into straight-forward, literal phrases, before using google translate.


Good observation! That's pretty typical for lots of geos and niches (that the after-work hours convert the best).



Eek! I hope this was just a bad day!

If things continue to look bleak tomorrow, pause the camp for a couple days, then resume and see if you get the performance back.

Also - check which placements your conversions were coming from before, and compare against the current placements you're getting traffic from, to see if you're still getting similar traffic volume from the same placements. If not, then you probably got outbid. Clone the camp and increase your bid to see if you can get some of that traffic (and conversions) back.

Keeping my fingers crossed for you...


Amy

---

Yes, I agree you should set the test landers to the same traffic conditions. And of course Vortex is right about Google Translator- the tool works well (just need to be careful with Arabic). There are more free translation tools you can check here

---

Not a good day unfortunately.

12/14 Results
Cost: $40
Revenue: $5.4
Conversions: 2
I checked to see how the converting placements were doing, and they were still getting at least the same level of traffic as they were on the previous days. So I don't think I'm being outbidded. I'm going to pause it for a day or two and see if that helps.

For now, I'm going to look for the next set of offers. Maybe I should try to deviate from sweeps? All I've been doing since day 1 is sweeps, and pretty much every campaign has performed the same as this one.

---

Yup - let us know how it does after being paused for a day or 2.

It may also be your offer. Is there another offer that you've tested previously, that was converting but just not as well as your best offer (that you're currently running)? If so, reinstate that offer to see if the conversions will come back or not.

If you haven't got such an offer, perhaps wait a few days, set up a new camp with the same targeting, and test a new batch of offers.

Or, just pick another geo/carrier to target and test entirely new offers.

Often, pop camps would just die suddenly, for no apparent reason. It could be the traffic or the offer. And sometimes, just starting a new campaign to test new offers may be the best route to go.




Amy

---

So I restarted the campaign today after pausing it for 2 days.

Results 12/17


Started off with a $30 test, and that let me cut the 2nd row lander (I cut based off data from the day I stated this split test, not just today's data). Not quite the performance I wanted to see, but I figured I might as well split test the best lander language. So I did another test with 2 variations of the best lander--one in the native language and one in English. The native language did better and performed at -60% ROI. Also, can you clarify something for me? When cutting OSs, devices, isps, etc, should I wait until I have the best lander and only use conversions from that lander, or can I speed things up a bit and make testing cheaper by using data from all the landers from the day I started split testing them? Seems to me that using conversions from different landers will misrepresent the data, no?

So that basically leaves me with the best lander at -60% ROI. Not too promising anymore, but I'll take your suggestion of trying 3-5 bids with the best lander and see how that goes.

Note: What I do find very interesting is that if I set the date range from day 1 of this campaign to today, my original angle (called LP3 in the posts above) performed at -47%. Although most of these conversions came in just one day. I ended up cutting this lander when I started split testing it against the other angles because it had under a 10% chance of being the best in the new split test. But the best lander from the split test performed at -70% ROI.

---

It depends.

With pop camps though, try to choose efficiency over accuracy in general - pop camps are so short-lived, you need to optimize them as fast as you can.

Basically, just eyeball your stats and estimate if the OS/device/ISP/etc. will stand a chance of being profitable for your best lander. For example, if the overall ROI of the OS/etc. is very negative, and all your landers perform around the same, then it would be obvious that even your best lander won't make the OS/etc. work. And of course you can drill down to OS/etc. -> lander to confirm as well. Obviously, try not to base decisions on a single conversion. 3+ conversions would be my recommendation - any less than that and the ROI can be far from accurate.

Also keep in mind that many traffic segments are related, such that when you cut one low-ROI segment, the ROI of the other segments will increase. So for example, if you see an OS/etc. at close to break-even or better, and you haven't really cut placements yet, then cutting more placements may push that OS/etc. into the green.


Worth a try!


This is exactly why you can't compare landers that were tested on different days.

Pop traffic is volatile, which is really not conducive to split-testing. But we should still aim to do our best to expose the landers to the same set of conditions, by rotating landers in the same campaign, and only comparing stats gathered over the same time period when the landers were all being run simultaneously.

Meaning, I can't take a lander I ran yesterday, and compare to the stats of another lander I ran today. Yesterday or today could have been a good/bad day.

This is also the reason why I always advise not to add landers to a split-test before the current split-test is done (i.e. finished cutting down to a winner).

You've observed that the performance of the campaign has tanked since the initial days - so it really wouldn't be fair to compare the ROI of a lander you're running now, against that of a lander you were running when campaign performance was good.

I know that it can be frustrating how short-lived pop camps are. I've summarized some additional things to try when campaign performance tanks:

https://stmforum.com/forum/showthrea...l=1#post332874

Also - to compensate for the short-lived profits, it helps to explore more traffic sources you can scale to, then scale hard and fast whenever you have a promising campaign on your hands.

Pop traffic can be a wild beast - do the best you can, and if things don't improve, it may be time to start testing another geo/carrier/traffic source/etc. for now - you can always come back to this campaign after a week/month to start things anew. Would probably be better though to create a new campaign instead of re-using the old one.



Amy

---

Home > Mobile > Follow-along Campaigns