Home > Mobile > Follow-along Campaigns

---

Master the game of Mobile CPA (17)

---

Hello onleads

I have to say you have done pretty well, you are very close to break-even and some combinations already are green, good job!

Now let's take a look at what you could do to improve the numbers.

1. The bid - not sure if the numbers in the screenshot are accurate, but the higher bid was actually giving you cheaper clicks, right? This is often the case, higher bids translates to better CTR and that can result in cheaper clicks. Even the initial ROI was better with the higher bid. Assuming that the numbers are accurate, you should consider running at higher bid.

2. LPs - LP3 really seems to be the best, at least from this small sample of traffic. LP4 is coming close tho, so I would continue with these 2. It is a good idea to run LPs that are close in performance for a bit longer for sure, but in this case LP3 wins by far so you didn't make a bad decision. The thing is, you don't want to have just one LP, so either try to make a variation of LP3 or test LP4 some more... or come up with a totally new LP as you need to drive the performance up a bit.

3. Offers - offer 2 is obviously the best for now, so keep that one and try to find a similar one or the same on another network and split test them.

4. OS - not sure why some of the androids perform that bad, usually some versions perform better than others, but not like this. I just checked some of my campaigns and 5.0 and 5.1 usually do well, 2.0 sucks in most of my campaigns for example. It would be a good idea to analyze this on a per placement basis - are there some placements that send an extraordinary large % of 5.0 or 5.1? These could be bots. If it's not the case and distribution is balanced across many placements, there will be some connection with the way the offers are setup and in that case, cut these low performing OS versions.

5. Placements - analyze placements and cut those that have extremely low performance - you don't have enough conversions to cut based just on that data yet, but you can cut placements that sent you a decent amount of clicks and none of them clicked through the LP. You should also consider running some BOT tests

Once you do this changes, let the campaign run for a while and then look at the data again.

You can try to run this campaign on other sources for sure, but first try to improve the performance some more. Not sure what GEO this is, so hard to say if it can be a large money maker or not. It can certainly be your first profitable campaign and that's what matters at this stage.

---

Hey Matuloo,

Thanks for your input.

1. I dont understand what you mean here. I am bidding at $6 CPM and Popads always charge me in between $3-$4. So should I increase my bids a bit?
2. OK, I will add in 2 more landers to my campaign, 1 with some variation of LP3 and 1 with a total new lander. Lets see what I can get after that.
3. I have already tested all the offers for this GEO from all the affiliate network that I have. So not much choice now. Will apply to new networks.
4. Will learn and run a bot test soon.

Question Is there such things that sales are slow on weekend? I found that my campaign does not perform well Friday and Saturday? Going to test Sunday soon.

Cheers

---

I'm talking about this screenshot :


You have lower CPV set there for the HIGH campaign, so I assumed it was giving you cheaper traffic. Did I get something wrong here?

And yes, some offers do perform worse on weekends, while some actually work better on sat and sun... these are things you will learn as you gain more experience.

---

Oh, when starting out, I was bidding 0.002 for low bid cam, 0.005 for medium and 0.008 for high bid. Then I got the results above. I think I understand what you mean now. Will increase my bids. Thanks for the helps!

Cheers

---

Yup, play with the bids a bit, you need to find the sweet spot - best possible price per hit/click but still good quality.

One more suggestion : since you will be doing more changes now, in the future try to make one change at a time and then get some data to see if it helped or not.

---

Ya, this is what I want to know too. If I change something to my campagin, how long I should keep it running before I carry out another batch of data analysis? Can I include the old data before the changes into the new analysis? or just analyze the current new data will do?

I have another question tough. If I really found that weekend does not convert well, can I stop the campaign on weekend at this stage? Anyway, I just started this campaign 4 days ago and this weekend will be the first weekend for testing.

---

You should run long enough to reach at least some level of significance, it's hard to say "how long", some people like to reach statistical significance again, some people (my included) react faster. Old data becomes irrelevant from a large part - when you cut a banner or an LP, data you have for those are not so important now. It still has some use - to spot trends for day-parting, to judge general offer performance, to have it as reference for banner/LP creation etc...

The best is to be running 24/7, but when its not possible, day-parting comes into play. Before you cut the whole weekend, make sure you have been running on it for a few times, the performance can vary from week to week and from offer to offer.

---

Its a good idea to let it run for a while, so you can see what effect the changes you have done had made to the campaign.

Its good to optimize fast, that's true, but this is only possible when you run larger volume. On top of that, you need to see what the day-to-day variations will be etc... some parts of the optimization process cannot be rushed. In some cases, especially with low volume GEOs and placements, it takes me even a week or two to fully optimize a campaign.

---

@matuloo, regarding the bot test, do you run the test before you start a campaign? I think I should run it before I jump into a new geo so that I can find out which placement to avoid. right?

---

I start my second cam today. Lets call it C0004 in future.

General Info of C0004

1. 6 offers, recommended by my AM
2. Direct Linking because the offer page itself is a lander with the landing page style that we can usually see from spy tool.
3. GEO - traffic is cheap for this geo.
4. BOT Test - 90% human for most of the placement. (My first bot test and i dont know whether I have done anything correctly here. )
5. Day parting from starting: 1800 to 0000

My question
Should I start testing 6 offers at the same time? Although they are landers but their vertical is different. My plan here is to test the best offer and traffic segments combo for this geo. Dont know this is a good plan or not.

---

Bot test : its a good idea to run it when you enter a new source or a new GEO. Once you are familiar with it, its not necessary to run it all the time, occasional checks should be enough.

So you raised the bids but the conversion actually went down. This happens sometimes too, the first thing to do would be to move them back where they were to see if the conversions return. If that's the case, you would know where to stay for the future.

Your campaign didn't start working worse because people are used to the LPs, that would be too fast. LPs can burn out, but not within a few days.

I see the emotions are taking over, but they should not It's not gonna help you.

Look at what you did : you paused the bad OS and picked the best LPs, you also changed the bid. That's quite a lot of changes at once and maybe you should have waited longer. Cutting the OS's was a good decision, they were not giving you any conversions. The bid level didn't prove to be a good idea tho. So first of all, move it back and collect some data. Then you can test it with more offers.

As for C0004 - 6 offers for such a small time frame will be too much I'm afraid - how much traffic is it getting? I'm also not sure about the direct-link strategy, properly done pre-selling always increases conversions. Let's wait for some numbers and post them here.

---

@Harrison



If you're running in any of those geos like the above then it's probably good to consider this approach:

1) If you're targeting a large geo like United States, you can setup 2-3 recommended offers from the AM and rip 5-10 landers. In the first phase of testing, you want to find out the little pockets of profit first. This way you can reduce the costs of your campaign on finding what works in the beginning. This is what I think Matuloo is really emphasizing in his post above! So for example:

Let's say Offer A(10 Leads) is doing the best against Offer B(5 Leads) (Assumptions: Same Payout) - I'd run Offer A
Drilldown the lander data for Offer A. Lander A,B,C,D,E all have 2 leads each=10. - Nothing to cut here. Move on.
Under Browsers you see Google Chrome is generating 8 leads and is green while the rest of the browsers are in red - Only Target Google Chrome. <---- Now you can apply the same idea towards other targeting factors like OSNAMES, ISPNAMES

This way you're only targeting things like i.e. Google Chrome, AT&T(ISP) to see if there's more profit potential. Once you let your data run more, you can eventually find which lander does the best. Once you find which lander does the best, you can open up all your targeting options again to see if you generate better conversions. When you open everything up and found your best lander/offer combo then you can start scrutinizing your targeting factors again by cutting things again.

Simon

---

Hey Simon,

Thanks for dropping by and sharing such a great info with me. The list that you gave me is really a golden list!

Ya, I am most or less doing the same thing like you say. But I dont dare to cut too much of placement because I afraid there will be not much traffic left for me to test new landers and offers.

---

In my previous post, I said that I may have found a working lander offer combo.

I decided to create a separate campaign for it and test it. I still have some questions here:

1. Should I release my day parting and test all the hours since I treat this as my winner combo?
2. Do I need to add more landers to this campaign? or just focus the current working lander offer combo? Anyway, I am still testing the new lander in my old cam there.
3. Should I move this combo to other TS now or I should test and collect some data here from popads before moving out?

Thanks

Harrison

---

Just a quick update for my C0005 campaign. I found that the traffic is a bit less for me to test the campaign quickly. So I decided to extend my day parting time from 1800-0000 to 1600-0000. I hope this will give me more traffic to test the campaign. Thanks

---

Wait, I see more leads form carriers in the above screnshots, Maxis, Umobile ... those are all carriers. Or is this the DE offer you are talking about?

We really need to clean up this follow along

1. check the offers you are pushing, cut wifi traffic in those campaigns where you push PIN submit offers.
2. keep wifi in campaigns that use SOI/DOI type offers.
3. Do not mix the offers, keep PIN submit offers in one campaign and the rest in another one.

In case I'm not wrong, the Malay campaign is the one that seems to work with carrier traffic, so keep it that way.
In the DE campaign, looks like the leadgen offers are doing better so stick to that and don't buy carrier traffic there.

Did I get it right?

Do not transfer anything to another traffic source until we sort this out.

---

Yes you are right. lol.

---

Home > Mobile > Follow-along Campaigns