Home > Mobile >

---

The Mobile Cookbook: The Main Course (Part 2) (45)

---

Searing The Meat: Gathering Initial Data


1: Set up 8 campaigns in your tracking, 4 for each angle

Now, set 8 campaigns up in your tracking software. If you’re using the STM Mobile Tracker, you can follow our detailed tutorials to set these up.

Each of these campaigns should be split-testing your offer on all the affiliate networks that have it.

It should be split-testing between your landing page and a direct link to the offer.

Set up 4 campaigns for each angle, named

• OFFER-ANGLE-WIFI-APP
• OFFER-ANGLE-WIFI-SITE
• OFFER-ANGLE-MOBILE-APP
• OFFER-ANGLE-MOBILE-SITE

Replace “ANGLE” there with the name of your angle and “OFFER” with the name of your offer.


2: Start Up The Same 8 Campaigns On Your Traffic Source

And finally, start the same 8 campaigns up on your traffic source.

Before you set them live, ask each of the AMs who have the offer you’re testing for targeting suggestions. If they tell you that one carrier, handset, site, or other targeting option does not work, de-target that in all your campaigns.

Add in all your banners for each angle to each campaign targeting that angle.

As you’ve probably guessed already, for WIFI-APP you should set your targeting options to restrict to only Wifi and App traffic. For WIFI-SITE, use Wifi and Site traffic, for MOBILE-APP use mobile and app traffic and so on.

Give each campaign a daily limit of $7 spend, and use no other targeting aside from country targeting and anything your AMs have recommended. You don’t want to restrict sites, carriers and similar yet.

Bids will vary by traffic source.

On an optimised/smart bidding source using “Smart CPM” or similar, enter 1.5x the payout of the offer. On other CPM or CPC sites, attempt to get an idea of the average CPM or CPC for the country you’re targeting and bid around the same amount. Check tools and ask your rep.

Put them in for approval, and once they’re approved, start them off!


3: Watch out for terrible ads and placements

Until your campaigns have spent $14 each, do not make any changes aside from the following:

1. If, after 5,000 impressions, one banner is showing no conversions and a CTR of less than half of the top 3 banners’ average, cut that banner.
2. If a placement has spent more than 3x the payout of the offer, isn't profitable, and is taking more than 50% of total impressions, kill that placement temporarily. Return it to the campaign after you have spent $14 - this is just to ensure that you see stats from more than one placement. .
3. If a placement is running extremely high CPCs (more than 3x the average) and has spent more than 2x the payout, kill it unless it’s profitable.

Otherwise, wait for your campaigns to spend the minimum before continuing.

---

Simmering: Optimising to Profit


1: Check each of your campaigns for ROI

Check each of the 8 campaigns you’re running to see how their Return on Investment ( ROI ) has been over the initial test.

If a campaign had an ROI below -50% in the initial test, pause it. These campaigns go on our “Not Promising” list.

If a campaign had an ROI of above -50% in the initial test, leave it running. These campaigns go on our “Promising” list.




Our NASCAR_Wifi_App and TESTING_Wifi_Site campaigns have an ROI of +10% and -15% respectively after the initial $14 spend. We put them on the “Promising” list and the other 6 campaigns on the “Unpromising” list.


2: Optimise “Promising” campaigns

Now, for each “Promising” campaign:

1. Cut banners on ROI, as explained in Part 10 of the Getting Started Guide. Set your ROI goal to break-even initially, as you’ve not yet optimised the rest of the funnel. Over time, we’ll raise this.
2. Create more banners until you have 10 banners again, and add those banners to the campaign. Create a few more too, so that you have some ready to swap in.
3. Cut whichever of the lander or the direct link performed worse from your campaign. Then create another 3 landers in very different styles and set them to split-test against the direct link. These landers should be consistent with the overall angle but explore as wide a range of options within that as possible: interactive vs static, short vs long, image-heavy vs plain, etc.
4. Cut your offers as explained in Part 11 of the Getting Started Guide. You may well not be able to cut any offers yet.
5. Check your top devices: Review your tracker stats for the most used handsets on each campaign. Check the resolution of those handsets, then use a tool like Google Chrome Developer Tools or BrowserStack to make sure that your landers work in those handsets.

Set these campaigns running again and leave them for another $15 spend apiece.




For our NASCAR_Wifi_App campaign, we test and find that only 4 of our banners have the potential for break-even. We pause all the rest, make 6 new banners, and add them in. We discover that the Direct Link performed best, so pause our test lander and create three new landers in various styles, ranging from very short to very long, and interactive to static. We check our 3 offers and discover that the F5 offer is the clear winner, so we pause the other two offers. Finally, we discover that most of our visitors are using a Nexus 5 phone, so check that our landers all work on that phone.


3: Check “Unpromising” campaigns for breakout potential

Look over the stats of your “Unpromising” campaigns and check if any individual elements of that campaign massively outperformed the others. Did the landing page massively outperform the direct link? Did one or two banners generate 3x as many conversions as the others? Did one placement or one carrier hugely outperform the others?

If you can identify at least one area where there was an element that massively outperformed the others, cut the campaign down to just the outperforming elements. If not, pause the campaign.

If you cut the campaign down rather than pausing it, then:

1. Cut banners on ROI, as explained in Part 10 of the Getting Started Guide. Set your ROI goal to break-even initially, as you’ve not yet optimised the rest of the funnel. Over time, we’ll raise this.
2. Create more banners until you have 10 banners again, and add those banners to the campaign. Create a few more too, so that you have some ready to swap in.
3. Cut your offers as explained in Part 11 of the Getting Started Guide. You may well not be able to cut any offers yet.
4. Check your top devices: Review your tracker stats for the most used handsets on each campaign. Check the resolution of those handsets, then use a tool like Google Chrome Developer Tools or BrowserStack to make sure that your landers work in those handsets.

Then unpause the campaign and let it run for another $15 spend.



For our NASCAR_Mobile_Site campaign, we find that one carrier, AT&T, had a conversion rate 4x higher than any other. We pause all other carriers on that campaign, then go through the rest of the checklist, then set the campaign running again.


4: Keep checking running campaigns

Continue to check your campaigns for placements and banners which are underperforming. Follow the steps in “Searing The Meat - 3” above.

In addition, if a banner or a placement gets to 5x payout without a conversion, kill it.

If you cut a banner during this process, replace it with a new banner.

If you are winning 90% or more of your bids on traffic, are in first place bid, or otherwise appear to be over-bidding on traffic, lower your bid 10% and wait a day. Repeat if necessary until you are winning a lot, but not all, of your bids, or until you’re close to the second-place bidder.


4: Review After Another $15 Spend

As each campaign hits its second $15 spend, check the following:

1. Check which landing page did best, as explained in Part 11 of the Getting Started Guide. Cut landing pages if it is appropriate to do so (in other words, if they have a 10% or less chance of being the winner).
2. Cut banners on ROI again. If one or more landing pages has been cut at this stage, set your ROI goal to break-even. Otherwise, set it to 30% (positive) ROI.
3. Cut offers again.
4. If this is a “mobile” campaign (as opposed to “wifi”), check if any carriers are massively underperforming. If a carrier has had at least 5x your payout in spend and is converting half as well or less than other carriers with 5x or more payout spend, cut it.
5. Check if any handsets are massively underperforming. If a carrier has had at least 5x your payout in spend and is converting half as well or less than other carriers with 5x or more payout spend, cut it.

If this was originally an “Unpromising” campaign, check its ROI now. If it’s not positive, kill this campaign.

If this was originally a “Promising” campaign”, let it run for another $15 spend. Repeat steps 3 and 4, then kill the campaign if its ROI is not positive after this second round of optimisation.




We let all three of our campaigns run for another $15. At the end of that time, NASCAR_Mobile_Site is running an ROI of +25% - we’re in profit, hurrah! We go through the checklist and keep it running. Meanwhile, NASCAR_Wifi_App is not positive, but since it was a “promising” campaign, we run through the checklist and keep it running anyway, for now.


5: Continue To Optimise

For every positive ROI campaign, continue to add new banners and cut any that have poor ROI, aiming now for +50% ROI or better.

Also, create variants of your best landing page, and test those: test new images, new headlines, new CTAs, new copy, new background colours or images, and any other element of the page that might be relevant. Continue to cut these using the Split-Test Calculator, aiming to continually improve your ROI.

---

And Serve: Scaling To Massive Profit

Once you have a single profitable campaign, you can scale it for massive profits.

Follow these steps once you have a campaign that is stable at more than 30% positive ROI over more than 3 days:

1: Double Spend

Double the daily spend on the campaign and leave it for a day. If this does not result in a significant ROI drop the next day, double the spend again.

Repeat this process until either you can no longer increase the spend, or you are spending 1/10 of your available budget daily.


2: Talk To Your Affiliate Network

Cut down to the top-performing offer. Then, talk to your affiliate manager on that offer, and ask:

• How much of a pay bump you can get on the offer. Specify that you’re aiming to rapidly scale and that you need the bump to help you achieve more volume.
• What revenue threshold you need to achieve to get faster payments. If you’re on monthly payments, you should be able to get to weekly payments if you do more than $1,000 revenue per week. Faster payments still are often available for higher volumes.

3: Duplicate in Other Geos

If the offer is available in other countries, create duplicates of your winning campaign - same landers, same offer, same banners - for those countries. You’ll probably need to get creatives translated - use a website like One Hour Translation for this.

Then run those campaigns for $25 spend (or approximately 10 times the payout, whichever is higher). Cut any campaign that doesn’t at least break even or have a couple of break-even banners initially. From there, optimise these campaigns like the original campaign.


4: Duplicate On Other Traffic Sources

Sign up for a number of other mobile traffic sources. Here’s a huge list of mobile traffic sources.

Set up the same campaign as your winner - same banners, same landing pages, same offers - on those sources. Use the same bid as your winning campaign.



And that’s it! Well done for making it to the end of this massive guide!

If you have any questions, comments, if you've spotted a typo, or for other thoughts, please let me know below!

---

Caurmen, thank you so much for writing this. I had a bunch of questions regarding optimization and this answered them all. It did make a couple new questions come to mind though...

What if neither of the 2 angles initially selected work? Do you just go back into your list of ideas and try with 2 new angles?

What if Angle 1 is 10% CVR and Angle 2 is 5% CVR? Do you go back and pick a 3rd angle to test against Angle 1 again?

The 10 banners we are supposed to create....is it 5 for each angle? Do you make a unique lander for each angle?

---

If neither of the 2 angles work, I'd usually recommend dropping the offer and trying another one. However, if you're convinced that the offer can work for you, try another couple of angles. It's a judgement call.

Angle 1 vs Angle 2: it's not a contest. If they pass the hurdles in the process, keep 'em running. If both of them end up profitable on multiple campaigns to varying degrees - that's a good problem to have

10 banners: you should make 10 banners per angle, and a unique lander for each angle.

Hope that helps! Let me know if you need any more clarification.

---

Awesome guide Caurmen! Thanks for taking the time to write this out One quick question... If you don't know how to design and program different landing pages yourself quickly then where would you recommend to get this done quickly and efficiently? The same for banners? Do you have any personal recommendations?

Warm regards!

---

Do you mean coding... or graphic design?

Banner design is typically something you need to become competent with yourself otherwise it will hold you back often.

Landing page design and coding can be outsourced via oDesk, Elance, etc. or you could use a more IM-orientated service. There are a few posted around the forum - just browse through the buy/sell and design sections.

---

Thanks for the tips Zeno... I'll get to work designing and creating banners myself... for the landing pages I did mean both the design and coding

---

HUGE guide, probably the best on STM!

---

Thanks for that advice Caurmen, it came right on time.

You mentioned that if you cut any banners, to replace them with new ones. I am a little confused about this step. Do they have to be highly different than the banners you already have or can they be just a color variation or something? So far I have been doing a very small variation of one of the winning banners and I don't think its right.

I know you said to make the initial 10 banners as different as possible from each other, is this also true when replacing a losing banner?

Also, how long do these campaigns last on average and how much maintenance do they require? I am having a hard time keeping them profitable for more than a week. Should I be always introducing new banners every few days to keep it fresh?


Thanks for all of your help!

---

When replacing banners, definitely make them quite different.

Changing a colour is unlikely to turn a mediocre ad into a star just as much as someone changing their lipstick won't make them twice as attractive.

---

What if my payout is 2.75, you mean I have to put in smart CPM 4.1? That is a lot!

---

@ims2014 - The replacement banners can be anything provided they're genuinely testing a new idea. So could have a different central image, different design, or a variant on an existing banner you know is working.

Don't test very small variations, though. You're aiming to explore the "problem space" of finding a winning banner as fast as possible, so you need to make BIG steps, not small ones.

As for campaign longevity - it very much depends, but yes, I'd recommend introducing at least a couple of new banners every week.

Hope that helps!

---

@iwarrior - yep, that's what you should put in! Don't worry, it's not necessarily what you'll end up spending.

---

I just went through this a few times, and I'm abit confused about when to cut offers. From Part 11: How To Cut Offers And Landing Pages, you mentioned 4x payout before you cut the offer.

If the payout if $0.50, then offer would be cut at $2 spend with no conversion? Before the $15 max spend per campaign?

---

What if I am not trying to improve the campaign but just keep it going? I have heard you need to add new banners just to keep it running at the same eCPA. I know this is true for adult but I am trying to figure out how true it is for mobile app installs. Do I need to add banners to keep the eCPA at the same level? Will small variations work for this purpose?

I have a couple more questions about banners. You said to always make sure you are testing at least some animated banners. Should I always make sure to test at least some static banners? After my first few campaigns, I have not really been testing static. Should I test static banners every single time?

Also, I am a bit confused about when you are launching a new campaign with 10 banners/angle. I used to test 10 of the same design with 10 different copy variations. After reading your guide, I realized I had been doing it all wrong. Let's say I am testing 10 very different designs, how many variations of copy should I use initially?

Thanks for writing this and all your help Caurmen. This guide was the tipping point for me. You're the man!

---

This is a guideline only and you would normally apply it to a specific banner, or placement.

Test static banners every time! I would suggest you focus more of your design time to static rather than animated banners. Animation can be a large time sink and this slows down how quickly you can roll out different tests. This may be a generalisation, but expect animation to effect CTR more so than the CVR of your angle (which is probably the most critical factor to test first).

For ad copy... I would focus on one ad copy first - or maybe 2 different ad copies if your designs are limited. Your ad copy translates your angle to the user. Thus, I think, for banners at least, this is better separated by the 'angle' category.

---

@quantum27 - The Getting Started Guide is focused on Web, wheras the Mobile Cookbook is for mobile.

Because there are more factors at play, it takes longer to gather reliable data on mobile. As a result, follow the guidelines in the Cookbook, not the Getting Started Guide, on when to cut things.

---

Thanks for that. But I'm still abit confused about when I should drop an offer. From what you wrote, I should drop it if I can't get it any campaigns to be profitable after testing 2 angles?

My AM is telling me the offer is one of the top offer on his list, but I can't get it to work. At what stage should I call it quits on an offer?

---

You would test 2 to start out, but definitely test more if those two fail and the offer is proven.

If the (proven) offer pays $1, I would probably stop at $50-$100 total spend after trying 5-10 angles.

There's no 'statistically best' approach, rather guidelines - you just need to commit to trying multiple angles and giving each one a fair go.

Be mindful of your budget and how much you're willing to invest in each offer and let this dictate how far you go with angle testing.

---

Currently, I'm spending minimum $60/angle. Which equates to $15 for each wifi/mobile/app/site combination. 5-10 angles would come to at least $300-$600/offer.

I don't have a problem spending that much, but want to know if it's money well spent, or should I move on to other offers.

---

I'm a bit more ruthless about offers - if I haven't seen promise ( more than -50% ROI ) from an offer after trying a couple of angles, I'll probably move on.

It very much depends on how much you trust your AM, how much other info you have which says the offer works, etc.

---

Quick question, when should we cut a once profitable offer? Say it had a few good weeks before hitting a bad one. Would you cut it or wait to see if it bounces back? Do good offers occasionally good and bad days with great variation in ROI? Thanks for the awesome guide.

---

I'd give it a week, no more.

If an offer consistently tanks for a couple of days, it's also worth checking in with your AM and checking to see if there are any problems with the links. Sometimes there are technical issues or budget issues that don't get fully communicated.

---

What is meant by placement here?

---

An individual place on a site or app where you ad shows up

---

So for Airpush, would that mean the different ad formats like Inapp, Push ads etc.?

---

Although I'm not familiar with airpush, a placement would be a place where your ad shows up, so for example an ad in angry birds is a placement, an ad in Sudoku puzzles is a placement, a push ad that comes in because someone has army warriors 4000 installed on their phone is a placement, a placement is just somewhere your ad shows up

---

Here do the 5,000 impressions mean 5000 impressions for that campaign or 5000 impressions for a banner?

---

Per banner, per campaign would put it at 500 impressions per banner, which is not a big enough sample size to be confident that the banner will continue to perform at it's current level.

---

They do have a "form factor" selector, but I think it's changed since I last used it, and doesn't appear to allow you to select phones only (although it DOES let you select tablet only...). Decisive folks / other affs, can you shed any light there?

---

@caurmen I have another questions because im getting close to 5000 impressions on some banners. You said "If a placement has spent more than 3x the payout of the offer and is taking more than 50% of total impressions, kill that placement." - Can you please show me what that means and how to kill a placement? I would assume it is to pause the campaign as I don't see any way of killing a placement without killing the campaign (or maybe it's the same thing?).

I also noticed some campaigns having much less impressions then others. Those campaigns that are not generating a lot of impressions also have much lower bids and I have a lower % of wins. The ones that have volume and I have wins of 70% + are bringing in a lot of impressions. Should I be increasing the price on the lower ones to receive more win percentage?

Thanks and looking forward to hearing from you soon

Aris



Ari

---

@1jilliondollars: If you look at the live reports for the campaign, under the app or site section, you'll see an option at the far right called "Blacklist". Tick that to kill the placement.

In terms of placements you should kill: just check each of the two conditions. Sort by total spend (click "Spend" at the top), and look for any placement that has spent more than 3x your offer's payout. So if your offer payout is $1, look for anything that has spent more than $3. Then look at the total number of wins for the campaign, divide by two, and check whether any placements have that many wins or more. If so, blacklist it temporarily, until your campaign hits $14, then un-blacklist it again.

Yes, try raising the bids on the other placements.

BTW, have you read through the Mobile Cookbook: The Appetiser? It's a lot more step-by-step and may help you answer quite a few of your questions without having to wait for me!

---

I appreciate your help and yes I have gone through the appetiser but didn't take action as I liked this version more. I created a follow along now and getting great help from there so I won't need to wait for you here haha

Thank a lot for everything

Aris

---

Awesome guide. Totally love it.

I have some questions:

1# I am new to affiliate marketing but knows the foundation. Is it still recommended to start with the appetizer? I really want to do split testing and make decision based on data

2# there are some numbers mentioned in the guide like $15 spend for each campaigns and stop the placements if ROI is -50%. Why these numbers? Are those just guidelines? Or are you guys using statistic significance of 95% or more?

3# speed is everything. How long does it takes to create all those angels, banners and landers? Lets say for one angle

4# how do you guys monitor your campaigns? Do you guys check your campaigns every hour? What kind of processes do you follow?

---

@whtang - Glad it's useful!

1) Yes, start with the appetiser - that'll get you making data-driven decisions and split-testing your ads. Once you're comfortable with that, move on to the main course. Don't try and do everything at once.
2) They're guidelines based on statistical signficance calculations for the spend range of the guide, as well as usual price ranges on RTB mobile platforms. Obviously since they're for a range of spends and payouts they won't be spot-on for everything, but they're based on math not just instinct. If you are familiar with statistical significance calculations, I'd highly recommend running your own stat.sig tests as soon as you're comfortable with the basic approach - but do sanity-check your results initially against the guideline spends.
3) Depends how much practice you have! It'll take you a fairly long time initially. Once you're experienced with it, your creation speed will increase. Don't try and force it, though - it's easy to take short-cuts like not doing proper angle generation or research, or not getting your text translated by a human. Those short-cuts WILL come back to bite you in the form of unsuccessful campaigns.
4) I generally check a campaign about once or twice a day. If it's high-volume I'll check more often than that just to make sure it's not going to massively overspend and I'm not wasting money on junk placements. Checking a campaign too often usually isn't a very effective use of time, as the data you'll have gathered will be highly biased by the time of day and probably not statistically significant anyway.

---

@Caurmen - thnx for the answers :-). Im going to start my following along this week / next week following the apetiser.

---

@whtang - great! Looking forward to seeing it.

---

First of all, thanks for this awesome guide Caurmen, greatly appreciated! Just a quick question, could we use this system with Exoclick RON? There are so many sites in their inventory you would blow through your $7 daily campaign budget pretty quickly without hitting that many placements.

The system i know for RON on Exo runs a test campaign for 24-48 hours to trigger placements and gather data then either move sites to a profits or low bid campaign. The test campaign remains paused though until you need more sources to add to your other campaigns when sites taper off and become unprofitable. I'm just confused when and where to use each strategy. Thanks

---

Glad it's useful! You could possibly use this guide for Exoclick mobile, but not desktop - the numbers are all wrong for desktop, as mobile has more confounding factors to deal with.

Placements shouldn't be an issue, though - DSPs have a similarly large number of large placements. You just need to keep on top of the spend and pause placements that are eating spend aggressively.

The system you're describing (which I heard described way back in the day as the "Master Blaster" strategy) is another very good one to use on RON, particularly RON desktop (it can work on mobile too but you need to keep aware of the other factors going into mobile like carrier and handset). Either approach can work - pick one that works for you!

---

Ok thanks for clearing that up! So i'll use the "Master Blaster" for RON Desktop and "The Cookbook" for RON Mobile

I probably should of asked this with my first question sorry, but what about using this strategy for premium sites like PornHub Mobile without the RON aspect. Would you follow the same general principles just leaving out the cutting of placements obviously and only running 2 campaigns (wifi and carrier)?

---

Hi Caurmen,

Thanks for you time and effort in creating all these materials and answering questions, truly awesome!

When it comes to 'placement', not all traffic sources include 'placement' in their targeting (e.g. Buzzcity), we can optimize by banners/design/carrier/device etc but not placement.. would you still recommend using them?

Thank you.

---

@ofourone - Yes, they're still worth testing. Check for bot traffic as usual, but there's no reason you can't get to profitability.

It's also worth asking your rep at that traffic source, or their support, if there's any "hidden" way to optimise on placements. Often there is, whether it's your rep blacklisting for you or something they only open up at higher spends.

---

If I'm using adplexity can I skip most of these testing phases and blacklist/whitelist certain Os's, carriers, placements, angles, banners etc. based on the data I'm gathering while spying? Or is there a benefit to testing all of these myself from scratch?

---

Adplexity will certainly give you a start on angles and banners - although you should still test multiple banners.

You can try running a whitelist campaign based on just the placements you see running a campaign, but it's likely to end up giving you quite small volume. Better to test more broadly there, IMO.

As for cues on OSes and carriers - I'd test from scratch initially and see if you're seeing the OSes and carriers that you expected outperforming the others. If so, you can probably trust the results you see in your spying in future on the same geo and vertical, at least for six months or so.

Basically, spying data is useful but there are a number of ways for it to be misleading. I'd always approach it on the basis that it could be wrong, but if testing matches up with your spying data, then you can trust the spying data a lot more.

---

Home > Mobile >