Home > >

---

From Red to Green: a Newbie's Journey on Push Campaigns (17)

---

Hey STM Friends,

This is the 2nd follow along. In the 1st FA, I was running a DE dating offer but got kicked out from the offer. https://stmforum.com/forum/showthrea...ffic-DE-Dating

Now I chose to run sweepstakes on Push traffic, this is the campaign details:

Offer type: Sweeps SOI, win phone
Geo: TH
Traffic source: PropellerAds
Offer Number: 4 offers from 3 different networks
Landing Page: 3 different types---Quize, Spinwheel, Giftbox
Payout: 0.3~0.56$

According to twinaxe's suggestion: the testing budget should be 4*3*0.4*10=48. Let's take a middle payout number 0.4 to calculate.

I set up the campaign with Classic push, CPC 0.012$. When I wrote the FA, this campaign has been running for 5 hours, 10 conversions, ROI -79.86%, spending 20.06$. This is the stats of offers.

Offer 4 seems to have a problem after it generate 1 conv. Affiliate manager told me the offer link expired. So Pause it for now. We have a total of 3 offers. The testing budget we need is 3*3*0.4*10=36$

From the stats we could see the LP CTR is amazing but CR is horrible. Not sure what causes the problem. Maybe the offer is too old? Saturated? Maybe there's snicky fishing code that steals my traffic? I really don't understand why the LP CTR is high but CR is low.

According to what I learned from twinaxe, the first step we need is to find out a good converting offer and use that offer to test the lander. But in this phase, offer 2 seems to converts great but has a low payout. Offer 1 converts not so well but has a higher payout.

It made me puzzle here: Which offer should I use to test landers?

This is what I think, in this phase, ROI is not important. So, we shall pick offer 2 to test landers. After We found out the best lander, use that lander to test these 3 offers again.

Do you agree?

---

The thing about TH is that it is one of the most pricey GEOs up there at Propeller, which is good - it means that it converts and there is some money to earn...

...But it is also bad because it requires higher bids and more money in general before finding out winning placements and/or landing pages + offers.

Be careful with your total ad spent when promoting something there, as payouts are usually below 0.50$ and it might require a lot of conversions on working and profitable funnel to get all the money back.

TH is on the top 5 list of Adplexity Mobile landers when you type "Iphone". You can get close to 50% CTR with the exceptional lander that you find there (there are plenty of them to test)

About the horrible CR - well, try thinking about all those wheels that are now spinning simultaneously for tens of thousands of Thai people.

Use statistical significance.

Here is an online tool that can help you: https://marketing.dynamicyield.com/bayesian-calculator/

Cut out underperforming landing pages and/or offers if their chance of winning is well below the others.

---

First of all: TH is hard.

Doesn’t mean it can’t work, just temper your expectations of cheap traffic and high volume.

Just use the spinner.

Gift box works OK but spinner is better overall in TH.

TH has loads of traffic.

Cut aggressively. Like 1/2 to 1/3 payout for any zone that doesn’t convert.

Offers are all roughly the same (iPhone, S20, 7/11, KFC) Just find the one that converts better.

I find that iPhone usually performs best. And only 1-2 on each network are consistent.

Then test that one against other networks. e.g. Goztha, Advidi, Clickdealer, etc all have decent TH sweeps.

---

Thanks for your kind advice. This tool is so cool.

Sure, I used this tool to check the offer performance after running for another 12 hours:

Clearly, offer B converts best among the 3 offers I'm running.

If we put it in the tool, offer B is a clear winner as well:


So I'd take offer B alone to set up a new campaign and test landers.

---

Hey Jaybot,

It's so nice of you to share the valuable experiences with me. You're right. I just discovered there're very few advertisers and offers here in TH. All you can run in Sweeps every day is iPhone, Samsung, 711, Shopee, etc.

Already worked with Advidi and Clickdealer. Goztha refused my application so they might expect at least intermediate affiliates to join. It's fine. The offers now are sufficient to have a test.

---

As I mentioned in the former thread, I'll run offer B to test landers. The testing budget is now: 3*1*0.4*10=12$

After running this campaign for about 8 hours, this is what we can see in the lander stats:


There're actually only 3 landers. The lander C has 3 verions for various carriers. I set up a rule in Binom to show them for relavant users. So, each user has an equal chance to see 1 of 3 landers, but when traffic comes to lander C, Binom will show users one of the variations of Lander C based on its carrier. Thus, we need to sum the stats of 3 versions of lander C to compare with the other 2 landers.

This is what we can see in Calculator:


Lander C has the biggest chance of winning but can't be called a winner according to the 95% significance level. Clearly the stats are not enough.

The next step, I think would be copying the campaign and launch it again. Let's see Lander C could stand out. Not sure if this is the right move. What do you think?

---

Hey Guys,

After testing landers for about 3 days, we now have the 95% significance level. The winner came. As you may still remember, I tested 3 landers, just paused lander A this morning as it's performing the worst and save the traffic to B&C. Finally Lander C stand out.

When I wrote the post, this is what I can see:


So for now, I'll use this winning lander to go back to test all 5 offers. Hope we'll have a working funnel soon.

---

Yup, C is the winner, but A is not that far behind... with a different offer or a set of Ads, it could still perform close to C.

It's always better to have more than 1 LP running, even when one of them is converting a bit less than the other.

What is the ROI like for these 2 LPs?

---

Lander A: -68.21%
Lander C: -34.9%

It's even clearer when we look at the ROI for these 2 landers.

---

Yup, when you posted the ROI, it looks much clearer.

Was it a totally different type of lander?

Try to make a few variations of the winning lander, to see whether you can up the performance.

If you decide to go with that one lander, try to give it some more traffic to see if the performance holds, because usually, the performance goes down once a single element get's more exposure. If the performance holds, move on to testing more offers to find some promising ones.

---

Yes, they're totally different type of landers.

You're correct about that. When I use the lander C only to test offers, the overall performance dropped. Is there a way to solve this?

---

These days, I'm using the winning lander to test offers.

This is what I could see after testing for 12 hours:


If we look at the stats in this tool, offer D&E are bound to lose. So, I'll pause these 2 offers and let other 3 offers continue running.



Just as I mentioned earlier, I thought the overall ROI would be so much better than this because I'm using Lander C, the best lander from former test to test the offers. But in the end, the overall ROI dropped to -58%.

It's so weird. Is it because too many users saw this lander?

---

It´s a bit tricky in this situation because the offers have different payouts.

Based on the SS calculator offer D is worse than offer A but based on ROI offer A is worse than offer D.

That´s the problem with such calculators, they only work with the data you feed them.

They can help with statistical significance in terms of CVR but the real results can still be different.

You can see it as well with offers B and C, in the calculator C is better than B but in tracker stats offer B is doing better.

I would definitely kill offer E and then it depends a bit what to do.

The offers all convert so when there is enough traffic available to have enough volume left after optimization just send a bit more traffic to the remaining offers.

---

Hey Twinaxe,

After sending more traffic to these offers, the performance looks still pretty similar. I really don't know which one to kill this time. Based on what you taught us in the Push Tutorial, the testing budget should only be 20 USD(5*1*10*0.4), but so far, I've spent more than $430 to test these offers after getting the best lander.


Offer A&C seems to have higher probability to win.



In this case, I thought even sending more traffic to these offers will only get similar results. Is it a good choice that I pause offer B and continue testing creatives with offer A&C together? Also, do you guys think if this campaign has a chance to be green?

Thanks.

---

Wow, this wasn´t that obvious from your last posts.

I reply with more details when I am back on my PC.

---

Hey twinaxe,

It is from the last post. I'm currently testing the offers after getting the winning landers.

---

Yes, I only saw it from your last post but from the posts before it seemed that you only spent way less.

One important lesson: Only wait for statistical significance when it´s worth it.

When you keep it running for too long you risk to just find the winner amongst the losers.

In other words, you only find the offer that is losing least money compared to the others.

But the statistical significance calculator doesn´t know and can´t tell you anything about profitability, it can only compare the clicks and conversions and declare a winner there.

When now 2 or more unprofitable items perform pretty similar then in worst case you could keep the tests running forever without haveing a clear winner and at the same time keep losing more and more money.

Good thing is that your offers are converting somehow so that not all money is lost and you could collect some data already.

What you could do is to take the best offer A (it´s best from the SS calculator and tracker stats) and run that offer with the best landing page.

Then also check the stats for all these offers combined and try to find placements that spent much money without any or only very few conversions.

Blacklist these placements then and run a test with the offer/LP combination.

-45% ROI isn´t that bad when there is enough volume available, then it´s still possible to get the campaign running when you have a good combination and get rid of the worst placements.

Set a limit for the new test of about $8 or so and when blacklisting the worst placements doesn´t show better results or when it would cut too much traffic then better move on.

---

Home > >