Home > General > Affiliate Marketing Forum

---

Running Mobile? Boost ROI with Granular Offer Testing! - Part 1 (35)

---

************************************

INTRODUCTION

************************************

With competition levels on pop traffic sources being an all-time high, it's becoming increasing important to test massively but effectively to get the most ROI out of each traffic segment.

Many people know that in general, mobile carrier/cellular traffic converts a lot better than wifi. I'd like to share my way of mass-testing offers in a granular way - by finding the best offer for every major carrier vs. wifi traffic separately. This way you can get more ROI out of each traffic segment.

I've only used this approach with pop traffic so far, but you should be able to do this for display traffic as well.




************************************

WHEN TO USE THIS APPROACH

************************************

The following conditions will make this approach especially effective and suitable:

-You already have a proven lander that works. Without a good lander, you can't gauge how promising the offers are that you're testing. Plus it would waste more test budget.

-Offer payouts are low. Otherwise it would take too much test budget. (Exception: When you don't care about initial test budget because you may just make it all back after identifying the best offer for each traffic segment.)

-You're running on a traffic source that has a lot of traffic for your geo. Otherwise you'd be waiting days to cut offers for separate camps.

I'm going to test some offers in Mexico on Popads to illustrate the process.




************************************

STEP 1: INITIAL TESTING TO IDENTIFY MAJOR CARRIERS

************************************

First thing I do when running in a new geo on popads, is set up 2 test camps to find out how much carrier traffic is available, and which ISPIDs/ISPNAMEs they're under:

Test Camp #1 - I call this the "CAR" camp - With "connection types" set to "cellular" and "connection speeds" set to "cellular/carrier".

Test Camp #2 - I call this the "NOCAR" camp - With "connection types" set to everything EXCEPT "cellular" and "connection speeds" set to everything except "cellular/carrier".

Basically I just use any offer for these - we're not looking to test offers at this point. Bidding a bit above average would be good, because we want to trigger traffic from every carrier. Set a $5 for each camp. Before running these tests, make sure you've added popad tokens ISPID and ISPNAME to your tracker!



Then we drill down into stats. Let's look at the CAR camp first:







Then I did this:

-In Voluum, I drilled down to ISPID -> Mobile Carrier -> ISPNAME, expanded all entries.

-On an excel spreadsheet, I recorded all ISPIDs that contained carrier traffic (except when there was only like 1 impression because that could just be wrong traffic; anyways we would never optimize for small volumes).

-For ISPIDs that had both wifi and carrier traffic, I also calculated and recorded the percentage of carrier vs. wifi traffic.

-I also recorded the ISPNAME for each excel entry. I ended up with this:




I then checked stats for the NOCAR camp. I:

-Drilled down by Mobile Carrier -> ISPID, expand for carriers (basically all entries except "Wi-Fi"), recorded all ISPIDs that contained carrier traffic (except when there was only 1 impression).

-Then I drilled down to ISPID -> Mobile Carrier -> ISPNAME, checked each ISPID recorded on the chart and noted down traffic makeup and ISPNAME for each. Resulting excel chart:






Then, I used PopAd's traffic estimator to estimate the amount of carrier traffic that is available for every entry in both CAR and NOCAR excel charts above. So I went to the CAR camp on popads, went into campaign edit, set the bid to $1 to get top bid, set the "Internet Service Provider" to each carrier one by one and checked the "Summary" tab for the maximum traffic available, and noted the number in the excel spreadsheet. Then did the same for the NOCAR camp.

(Note: If you're only interested in testing either one of mainstream or adult offers, you'll need to target the right categories to get the corresponding traffic estimates. Same thing with other targeting - e.g. if you only want to test android offers then target android only).

To finish the excel spreadsheet, I looked for entries that had both wifi+carrier traffic, and calculated the daily volume for carrier traffic.

The finished spreadsheet:






So what does all this data tell us?

1)There isn't much carrier traffic for Mexico on Popads.

Iusacell may look like it has good traffic volume, but 80% of Iusacell traffic is wifi and only 20% is carrier. Still worth a test - the key would be to have at least one offer that converts well enough for the wifi traffic in order for the ISPID to be profitable overall.

Telcel does have some carrier traffic volume (approx. 5800/day), but for ISPID 366660 then you'll need to be prepared to have 145k wifi traffic "tag-along" with that Telcel carrier traffic. ISPID 959205 is 100% Telcel traffic, but volume is negligible (only <1000/day). Again, still worth a test assuming we can find an offer that converts well enough for wifi.

Nextel has 2k/day max. Not much especially when we separate that out further for adult and mainstream, but if we hit on an offer that converts well for this carrier, that may turn into xx/day - a nice start for a newbie.

Every geo and every traffic source will be different in the amount of carrier traffic available. By spending a little bit of money to set up test camps for different geos on different traffic sources, you can target the largest volumes of carrier traffic and mass-test offers for them. MX on popads unfortunately doesn't have a lot of carrier traffic, but I'm going to continue this example anyways for the sake of illustration. I can't bring myself to show stats for my best geos but you can find out easily with similar testing.


2)On the other hand, there's lots of wifi traffic. Over 95% of MX traffic on popads, in fact.


So what do we do with this info? Please see Part 2!





Amy

---

Hey Amy,

First, awesome thread (can't wait for part 3!). I've been thinking about how to best do this for the past couple of weeks and now I have a blueprint for me to follow!

I'm not sure if it's on my end, but I am unable to view any of the images that you've embedded. Opening them in a separate tab leads to this error message:

"vBulletin Message
Invalid Attachment specified. If you followed a valid link, please notify the administrator"

Not a huge deal as it's pretty easy to visualize the process you've outlined, but I thought I'd give you a heads up anyway.

---

Hello! Thanks a lot. But I can't see images

---

Uh oh! Watch out.

Anytime someone thinks they have a 'blueprint' to making money.......they usually DONT!!!!

If there truly was a mapped out 'process' to make money, why wouldnt someone just hire a bunch of filipinos to replicate the process?

---

Hey John,

Big fan of your blog (I nearly applied to Nutryst awhile back). I'm heading down to southern CA this summer to meet a buddy that I played college ball with who lives near LA. Would love to buy you a beer along the way if you're in the area!

My post above was probably a poor choice of words, thanks for calling me on it. Only meant to express my appreciation to Amy for putting the info together, as I was beginning to go through a similar process myself and she likely saved me a decent amount of time (and maybe some cash).

Looks like my newb is showing...I was under the impression that many affiliates did exactly that!

---

Yep! they do

---

I can't see the screenshots as well, same with those in Part 2

---

Thanks very much guys! It's really strange that images weren't showing - I've uploaded them to imgur this time so they should display alright now! If not please kindly let me know!

You're right about that! "Blueprints" will seldomly work right out of the box. The effectiveness of this method will depend on a lot of testing by the affiliate still.

First of all, the person must already have done previous testing in the particular niche, and know which lander(s) work.

Secondly, it will still require offers to be tested.

Thirdly, I haven't given away the best geos on popads, or any other traffic sources - not every geo will work out, and the more offers you want to test the more money you'll need to spend testing them.


Thus, there's never any guarantee that people WILL make money using this method, and I believe that any members reading this will understand this to be the case. However, it IS an approach I've found success with, and TBH was even a bit hesitant to share - as I'm STILL banking from it and have started to test more and more geos.

Thank you for your feedback!



Thanks bandito and all others for the encouragement. I'll try to do more similar writeups in the near future.



Amy

---

Thanks Amy - this is awesome!

Question: what sort of traffic volume would you need for a $xx campaign or $xxx campaign? And what is too low to bother with?

EG if I take the average bid on popads for Mexico (which is shown as 0.00088 in the popads inventory), then the average bid x 12,000 impressions available for lusacell seems too low... (0.00088 bid * 12000 impressions = $10.56 in daily spend). Even with an ROI of 100% it's only $10.56 in profit.. which seems like it would be hard to get profitable after initial testing costs. Am I missing something here?

---

Hi Brad! You've asked a pretty key question. Unfortunately, I can't recommend a cut-off volume. The actual profit levels will heavily depend on offer payout and conversion rates - which are things you can't really calculate before you actually start testing offers.

So what I would suggest, is to perform the CAR/NOCAR testing for multiple geos at the same time, see which geos have the most amount of carrier traffic, and just start testing from there.

To give you one example of the kind of profits you can make by targeting carrier traffic: One of my buddies is currently targeting one specific carrier for one specific geo on popads. He receives approx. 15k impressions a day, and is making $500+ in revenue, over half of which is profit.

So - get testing!



Amy

---

@Amy:

Thank you for this excellent guide :-)

---

Thanks Amy for the post, is this post only for pop traffic or does it apply for mobile display (banners) as well?

---

I'm only running pop traffic ATM, but there's no reason why the same approach won't work for display traffic! After all, if we're talking about carrier-billing offers that convert well on carrier traffic, then display traffic shouldn't be any different as long as you're targeting carrier and not wifi.



Amy

---

Hi Amy,

Is the purpose of this test supposed to find offers that tailor to specific carriers that could potentially generate higher ROI? I find that in many cases if I go to a affiliate network and I only use offers that target all carriers and all connections.



For ISPID: 499 I see that it's being divided up into two mobile carriers. Is the mobile carrier "Other" in this case Wifi? After creating a CAR and NOCAR campaign I don't see Wifi Mobile Carrier anywhere, I think it's labeled as Other instead. If that's the case I can only see 2% of traffic being Wifi and the 98% being Smartfren.

---

Hey Simon!

Simply put (and I should really add this explanation to the original post haha): There are offers that are DESIGNED to convert well for carrier traffic. These will typically be one-click carrier-billing offers.

The reason why they convert well on carrier traffic, is because they have a different conversion flow. When a visitor is on 3G, all they need to do is click "subscribe" and the conversion is done. But when the visitor is on wifi, they can't go through the one-click flow - they'll need to go through MO/MT where they need to reply to an SMS or submit a pin.

Some offers will contain both types of conversion flows, where one or the other will be presented to the visitor depending on the connection type they're on.

What this tutorial is about, is to find out which geos have enough carrier traffic for them to be worthy of the mass-testing, and also which ISPIDs contain carrier traffic. Popads is just weird in how they categorize ISPs - some of the carriers can't even be targeted exclusively without wifi traffic tagging on. And this is the kind of stuff you'll need to know before you even start testing offers. (e.g. If the only way to target CarrierX is to target ISPID-12345 which will give you 30% CarrierX traffic and 70% wifi traffic, then you can probably forget about testing CarrierX offers).

And yes - Voluum used to have a row for "wifi" but it's been switched to "others" since. I'm assuming they're the same thing. If you find out different please let me know.

So in terms of testing offers: You could target carrier traffic which can give you nice little pockets of profits, or you could target wifi traffic for volume and scalability. Or you could do both: Test wifi offers first, get profitable, then start directing the carrier traffic from your existing camp to one-click offers to potentially increase your overall ROI further.

Sometimes I talk so much that the main point I'm trying to get across, gets lost. So please ask for further clarification if needed.



Amy

---

Could I use ripped LPs for this testing? I sort on adplexity to show me the longest running first, I guess they are profitable because they're running long.

Thanks

---

You could. But if you're wanting to test landers first, it would probably be best to take a couple of AM-recommended offers (i.e. proven offers) to find the best lander. And THEN use that best lander to mass-test offers. If you throw a bunch of ripped landers in with a bunch of offers to test at the same time, chances are it would cost you more money to cut down to the winners.


Amy

---

Thanks for answer. Will 3 offers and 3 landings rotation work for this, or run like 3 recommended offers on one Lander, and then repeat the process with few landers and then choose winners?

---

You can use whatever arrangement you want, but please keep in mind that the more offer+lander combinations you have, the more money you'll need to spend before you can find a winning lander+offer.

On the other hand, the more offers and landers you test, the higher the chances you have of finding a profitable offer+lander.

So how do you strike a balance between these 2 opposing trends?

My answer is to use 2-3 PROVEN offers first (recommended by AMs) to test all of the most common landers (e.g. you can rip all the different landers you see on adplexity), find a winning lander and offer. And THEN using the best lander (and the best offer as "test control"), mass-test offers.

That's the most cost-effective way I know of to test extensively.

Once you find a good offer, you can go back to testing landers - test your own original angles, or lander themes, or both!

If you test 3 offers on only 3 landers, and all 3 landers suck, then even if there's a good offer amongst the 3, you may not be able to discover it. Running 3 offers on 1 lander would be even worse - for the same reason.

It doesn't mean your approach won't work. If you have prior experience in a niche and KNOW the 3 landers work (or the 1 lander works), then by all means just use those to mass-test offers. I'm assuming you're new to the vertical and have no idea what landers will convert. I'm sure you'll exercise your own judgment with this, now that I've pointed out some considerations.



Amy

---

Thanks for answer, everything is much clearer now So I'll test recommended offers on riped landers, find the best lander, then test more offers on that lander and find best offers, and then with that offer test some new landings using the previously best LP as a benchmark

---

Hey Amy, i want to ask you a question. Maybe I am wrong as I am new to propellerads and popcash, I found that they dont provide the options for us to select which carrier we want to target on. So what is your approach for those TS?

Another question is, lets say I found a good lander + offer combo and I want to move it to other traffic source that does not have the targeting options that popads has. The combo might need to cut certain placement before it is profitable in popads. So what should i do in other TS?

---

Haha you've summarized my long rant into a single sentence! Verrry nice!


Popcash does not have carrier targeting as an option, but propeller does - so please talk to their support if you can't see how to do that.

For TSs that don't have carrier targeting, you'll just have to find a way to monetize the "tag-along" wifi traffic. So you'd need to either send the wifi traffic to a monetizer like Monetizer.co or YTZ, or test offers in the traditional way to find one that converts well on wifi traffic.

OR, you could just stay away from TSs that don't have carrier targeting altogether. No need to try to make everything work - you just need to make enough things work to make money.

Regarding scaling to other networks: Yes you'll often need to do some tweaking to make them work. What I would do is keep the same targeting - for example if I found out on my original TS that the offer converted like shit on IOS, then I wouldn't bother testing IOS on the new traffic source (exception: if by targeting android I'm getting a much higher ROI than on my orig TS, which means the traffic is of higher quality, I may test other OSs on the new TS). I would test several bids again because the bidding landscape for each TS is different. And yes, you'll need to cut placements at the new traffic source.

Regarding cutting placements though: If you've tested a few bids, and even the bid that gives you the highest ROI, isn't remotely close to breaking even, plus you don't have any big placements eating up your budget and not converting - in that case I may just give up on scaling that camp to that particular TS. The amount of work required to make it work may not be worth it. That's a call you'll have to make.

Personally I don't like to spend a lot of time and effort trying to make difficult camps work. I'd much rather spend that time testing more offers or scaling to more traffic sources, identify the easy camps and just manage those. Everyone's approach will be different though.



Amy

---

Hi Amy, I would like to ask about your "bid testing" approach here. Please forgive me if I am asking at the wrong thread. I cant find any of your threads that explain this except those small posts you left on others' thread, which I think is not nice to ask there.

I saw your reply on this thread.

Can I do it on this way?

1. I create 3 cam with different bid. Set $10(5xpayout) to each campaign and let it run.
2. Once the budget is finished, I go back and see whether they bring me any conversions. I will stop those campaigns that do not bring me any conversions. For those with conversions, I will let it continue and finish my 10x payout.
3.At this stage, if a campaign does not hit ROI -50% or more, I will either drop the campaign or find another batch of offers for testing. (I remember you mention this before that it is hard to make a campaign that is too far from break even to be profitable by cutting placement only. right? )
4. If a campaign has a good ROI, then I pick it up for further testing and optimization.

If I do this way, can I use the data I spend money on these 3 campaigns for optimization? or I only can use the data from the highest ROI campaign for optimization?

Harrison

---

Hi Harrison! There is no right or wrong way to set up testing. But let me point out some considerations here:

With my "staggered bids" approach, the goal is to find out which bid will give the highest ROI, so that we can use that bid when doing further testing (split-testing offers and landers etc.) 10-20x payout may or may not be enough to decide whether we can give up on the campaign or not!

To decide how much you need to spend to see if the camp has hope, you need to think about how many offers and landers you're testing, and how many traffic segments are accepted.

For example, if you're testing 3 offers and 5 very different landers - that would be 15 combinations already! Obviously, running 10x payout will not be enough - all you need is one of those combinations to be near break-even to optimize the camp further to hit green.

And if the offer accepts android+ios traffic, wifi+carrier traffic, WEB+WAP traffic....then you'll want to run enough traffic to each major segment. For example, if android is profitable and not the other OSs, you can still make a lot of money because android can bring you enough volume.

As for whether or not you can use data from all 3 campaigns for optimization decisions: Probably yes. Just quickly check if you see opposing trends: e.g. let's say you're cutting landers, make sure the lander you're cutting, isn't the best-performing lander in any one camp. If you see opposing trends it would be best to cut based on the individual campaign's stats.



Amy

---

@Vortex

Say we don't yet have a proven lander for a particular geo. Mass testing landers upfront would be too budget-intensive with this method, but what are your thoughts on just spying/ripping a popular lander for a particular geo, running the above tests, identifying the best traffic segments + offer combos, and only then mass testing additional landers on our most promising combos. It shouldn't make a huge difference if the lander isn't proven as long as we're using the same one across all traffic segments & offers - we should still be able to find the most promising combinations, correct?

Hope you understand what I'm getting at

---

Yup I get your point! However, by using a single lander, you run the risk of it NOT performing very well - in which case you'd be spending a lot more money than you'd otherwise need to if you were using a winning/proven lander.

And mass-testing anything will often not be as budget-intensive as you may believe. Often, you'll see a couple/few candidates leading the pack in performance, which will allow you to cut a lot of the other candidates early on.

The worst thing that can happen when you're mass-testing is for many candidates to perform around the same - but I've almost not seen that happen.

Why not just try it once and see how you like this approach? Then tweak your approach as desired. That's how people develop their unique test approach over time.


Amy

---

Just to check, is this $5 test affected by the timing that you're running the test? Eg. does the # of impressions from each carrier get affected if you run it at a certain time? morning/afternoon/evening time

From my understanding of the guide, the numbers will vary but won't make up that much of a difference, right?

---

I haven't actually compared traffic levels from each carrier from hour-to-hour, but my guess is that the percentage of each carrier that makes up the total traffic should be similar throughout the day. Think about it: What would cause users of CarrierA to use their phones more during certain hours compared to users of CarrierB?



Amy

---

Thank you for this tutorial.
I added tokens ISPID and ISPNAME to Voluum, but when I am trying to sort by ISPID and ISPNAME I see only "Unknown" ISPNAME and in most cases "Other" Mobile carrier. What I am doing wrong?

---

I found the answer why ISPNAME not shown, I have forgot to select "track".

But what is "Other" mean, is it Wi-Fi?

---

Yup - common mistake I've also made before! Glad you found and fixed it!

Yes - "Other" used to be labeled "Wi-fi", then later got changed to "Other". I remember having asked Voluum about this, but forgot what they said (I have the worst memory...) I definitely remember that most of the traffic is wi-fi, but a small portion of it is other stuff that doesn't fall under the other categories.



Amy

---

Hi Amy

Can you please guide me how do i add popad tokens ISPID and ISPNAME to Voluum?

Thanks

---

"Parameter name" and "name" are not that important - you can basically put anything there. Just make sure the placeholders are the same, and that the "Track" checkboxes are checked.



Amy

---

Great post as always @vortex!

I have encountered carrier/wifi discrepancies before I have read this post. I was confused when I was targeting a carrier on a traffic source but my tracker was showing that half of it was wifi. I have asked both the traffic source and the Voluum support. This is what they have answered. "Our platform is using XYZ type of database which is the best". Something in between of those lines So basically they are using different databases and everybody claims that theirs are the most accurate in the world. Typical marketing lol

As I can see in this post, it is assumed that the Voluum is the right one to identify if the traffic is wifi or carrier. What about other trackers? Would this type of traffic testing would be valid?

---

That's a really good question!

It's been too many years since I "decided" Voluum's database was "more accurate" than PopAds'. I don't remember exactly how I came to that decision, but I seem to remember that Voluum stats made more sense in terms of which carriers the conversions fell under, i.e. an offer that accepts carrier A would show conversions under carrier A in Voluum.

In the end, it's valid conversions we're after, as our goal is to maximize profits. So we should always optimize towards profitable traffic segments - whether those segments are detected by the tracker, or segments passed from the traffic source to the tracker.

Meaning: If there's a discrepancy in how the tracker is identifying the traffic vs. how the traffic source is identifying the traffic, treat these as two different variables - not unlike how you would treat "OS" and "browsers" as two different variables - and keep the profitable traffic segments only, cutting unprofitable segments.

And by VALID conversions, I mean conversions that the aff network or advertiser will deem valid, that they'd be willing to pay you for. So if there's a concern that the aff network or advertiser may not pay for the conversions because they're not deemed to be from the correct carrier, it would be good to pass both variables to the aff network (by appending them to the aff link when adding the offer to the tracker), so that the aff network or advertiser can tell you which variable values they'd be willing to pay you for.



Amy

---

Home > General > Affiliate Marketing Forum