Home > Mobile > Follow-along Campaigns

---

Let's get our hands dirty! Newbie's Mobile Pop Sweepstakes Quest to Success! (5)

---

Hey!

Nice to see your follow along here!

Yes, sure. You can try Voluum, but I recommend to switch if you don't have a lot of visits (100k/day and more), because it will be expensive.

It's so good that you have an access to 10 networks, but I try to choose 2-3 the best and work with them. It is so hard to work with 10 networks on start, because you have to test a lot of offers from a lot of networks.

That is the best way now. You are right.

Yes, you can also try Popcash, Clickadu, HIlltopads. They have so good quality now and it will be the best choice at the beginning.


About your question:

Discrepancy is a part of our game, but you have a really huge %. Can you please tell me what GEO are you running and where your server is located?

Sorry I'm not using Funnel Flux and I can't help you with other questions.

But can you please provide us more information about your camps? What vertical do you promote? What kinds of landers do you use?

---

Cheers to a great start!

I would clone this camp on propeller and set up a test camp on Voluum to run a small test. 1k impressions should be enough - just to compare results.

I've heard very good things about FunnelFlux, but haven't used it enough to provide any suggestions as to what may be wrong. But mrbraun's question as to where your target geo is in relation to your server location is a crucial one, because the greater the distance between the two, the slower the redirection speed. However, I doubt this factor alone would cause such a huge discrepancy. Doing a quick test on Voluum may give us a better idea.


First of all - to view drill-down stats, don't click directly on the campaign name.

Instead, go to Stats > Drill-Down Reports. Click on the top field to add parameters you want to see stats for - and you can specify multiple levels. Click on "Apply" or "Refresh". Then click on the little chevrons in front of items in the stats to expand them to see the next level.



Stats like OS, device, browser etc. are automatically detected and collected by Funnelflux. You don't need to pass any tokens anywhere to collect this information.

There's some misunderstanding here, and I don't blame you - tracking can be really confusing. Took me a long time before I got the gist of it as well. Basically:

-Tokens you append to the offer url (i.e. aff link) when creating the offer in your tracker, are for passing parameters to the affiliate network. You have to pass the hitID as a minimum if you want your aff network to post conversions back to the tracker (plus you'll need to put a postback link into your aff network's offer settings as well - both steps are required for conversions to be posted back successfully). Other parameters are optional, but I would suggest passing the campaign ID and traffic source ID to the aff network, so that in the event where the advertiser does not like your lead quality, they would have the option of telling you which traffic source is responsible for the bad quality, instead of kicking you off the offer.

-Tokens you append to the postback link that you put into the aff network, are for passing conversion data back to the tracker. HitID is again a must. Payout would be recommended as well, unless you want to specify this in the tracker instead. IMPORTANT: Make sure the variable you used to pass the hitID in the offer url above, is the same variable you're using to pass the hitID back to the tracker in the postback url.

-Tokens you add to traffic source settings in the tracker (for each traffic source), are for passing information from the traffic source to the tracker. You can get a list of available tracking tokens from the respective traffic source.

-Tokens you find behind the campaign link are automatically added by the tracker - they're the tokens you added in the traffic source settings in the tracker.


Will get back to you on this...



Amy

---

Hello, thank you for your comments! I apologize for not being as transparent as possible regarding my camps, I can see how that can lead to some confusion.

The GEO that I ran these first three camps in was in France. My server is located in Chicago. I am aware that this set up may not be ideal for running pop traffic, especially considering that optimizing for speed is one of the most objective methods of improving one's chances at success in mobile pops.

As I am writing this post, I am preparing to migrate over to Voluum, and I am also setting up my CDN with Amazon's Cloudfront according to caurmen's thread: http://stmforum.com/forum/showthread...hlight=caurmen

I am currently promoting lead-gen sweepstakes offers such as win an iphone, and I am ripping landers from AdPlexity in hopes of finding the best-converting lander before split-testing angles/designing my own twists for landers.

I have noticed that a majority of the landers on AdPlexity are very aggressive, and a lot of them will not fly past creative approval from certain advertisers. I have made an effort to edit the landers to the point where they are somewhat compliant (removing obvious FB/Google brands, logos, etc) before sending them in for approval.


One thought I have regarding my current approach and how it can be approved...

After looking at the camps that I've run in France so far, I feel as if the amount of landers that I have chosen to split test across offers appears to be lacking. Other than the 1 landing page that got 6% ctr, all my other landing pages that I've tested so far have gotten anywhere between 0.5 - 1% ctr, which seems pretty terrible.

Looking back... Camp 1 had 5 landers, and Camps 2+3 had 2 landers that were the same (with the exception of changing a few text/images). So in total across these three camps, I had only tested seven landers in total. In my opinion, this does not seem like enough in order to find the "best" converting landing page.

A brief outline of how I want to approach sweepstakes.
1) Gather offers with low payout within the GEOs I am interested in focusing on that have proven to receive traffic (ask AM's for this.... still working on gathering these top offers).
2) Compile a large number of ripped landers (that are all fundamentally different in some way) off of AdPlexity, and fix them up.
3) Run an initial test to the campaign and reach statistical significance... cut down to one remaining lander that can convert offers well.
4) Mass-test offers across this one lander.
5) Identify the best offers that convert with this lander.
6) Split-test angles and create different versions of the best lander across these offers.
7) Cut placements and create whitelist campaigns.

One question...

According to Amy's thread on cutting placements: http://stmforum.com/forum/showthread...Banners-Part-1

I would have to run my current campaigns to statistical signifiance before deciding to cut all the landers until one is left remaining per camp. (Correct me if I am wrong on this)

My question is, given that every offer I've tested in my campaigns in France thus far have been averaging $0.50 ± $0.05, approximately how much traffic should I be running in spend to the campaign before deciding that the data drawn is statistically significant?

mrpayne recommends the following formula: (Average Offer Payout) x (Number of Offers) x (Number of Landers) x (3-5 times) = Test Budget

Are there any signs I should look out for in my data that may help guide my thought process as to whether or not I should continue to spend more/less? If I am at a stage right now where that may be difficult to tell with my level of experience, I will stick to the formula.


Lastly,
I have been setting my campaigns up on PropellerAds with the notion that I can have my spend distributed evenly across 6 hours of time (I have been trying to target 6PM-12AM for the targetted

GEO's local time). However, my campaign daily budget seems to be used up within 20 minutes, despite me changing the bids up. Understandably, I have been using a total campaign spend of 10$,

which might contribute to the fast spend, however, I have been setting a reasonably low CPM bid. Regardless, my spend is depleted quickly and I am unsure if I should reduce the bid even further,

which might put me at risk of very low quality traffic. Any ideas on this?

---

I found out about this. This is a result of the backbutton script which I'm guessing you're using on your landers. FF will track the entire path of the clicking-back every step of the way. To observe this, just set up a test camp for yourself and browse to the campaign link, and click back multiple times to see how the stats are registered in FF.

I would suggest to make 2 sets of landers: an aggressive set and a compliant set. Technically you don't even have to "make" an aggressive set - most of the ones you've downloaded are already aggressive. Then you just make a separate set by removing logos.

Then, you simply set up rules in the tracker to run aggressive landers for offers that don't require lander approval, and run the other offers on the compliant set.


Sounds good!


First of all, you don't decide whether or not the data is statistically significant - the peakconversion calculator already takes sample size into account when it does the calculations. So basically, just check your data from time to time until the calculator tells you that statistical significance has been reached.

Caurmen has plans to write a post on how to set campaign budget, so stay tuned! I personally almost never think about setting a budget, because it's hard to do so without taking into account a variety of variables, many of which you'll only find out actual values for after you start running traffic. So I would just start running traffic first and then see what would happen.


That's a really good question to ask for sure!

It would depend on how much more room there is for optimizing the campaign. It would also depend on what your testing approach is.

For example: Say we're currently using 2-3 offers to test and cut landers down to the winner. In this stage, even if there's no hope of making this camp profitable, as long as losses aren't great, I would continue to cut landers. This is because I know that after finding the winning lander, I'll be using it to mass-test offers, at which point there will be a good chance to finding a good offer.

Once you're mass-testing offers though, I would suggest to drill down to Offer > Lander > [every major traffic segment], to see how many combinations are green. The more green combinations, the better the chances that the camp will end up profitable. "Major traffic segments" are segments of traffic that will give you enough traffic to be worth targeting exclusively, such as OSs (Android and IOS mostly, maybe Windows Phones for the bigger geos) and big carriers.

Even if you don't see green, it doesn't mean the camp isn't promising. If there are big placements that aren't converting, then that's where you can achieve a big jump in ROI. However, I always recommend to focus on testing landers and offers - with extra emphasis on OFFERS. I would much rather test 20 more offers than spend the time and money to cut a hundred placements for each traffic source I want to scale the camp to.



Amy

---

Thanks for the very detailed updates!

It's great that you'd take the time to detail your steps - otherwise I wouldn't have been able to tell that you weren't cutting landers correctly - because the end result HAPPENED to be the same!

Basically, what you were referring to as "2 rounds of testing" was really the same round. You started off with 9 landers, and the test will not be over until you've cut down to one lander - or 2 or more if that's justified - but certainly not every time you decide to check and cut landers!

So - in your first check, LP8 was the "in-the-lead" lander that you'd compare EACH of the other landers to. LPs 4, 5Fiverr, 6, and 7 were ready to be cut.

Then you'd keep running the rest (still the same round of split-testing - continuing to cut landers). The second time you checked, LP8 was STILL in the lead with 5107 impressions and 6 conversions (I added up the stats in both screenshots). Comparing the other landers against it, LPs 3 (with 5155 impressions + 2 conversions) and 5GoogleTranslate (with 5040 impressions + 2 conversions) were ready to be cut.

This leaves LPs 1, 2, and 8, and you decided to cut LP2 because it's very similar to LP1 - OK that's fine.

I may also have misunderstood your process - if I did please forgive me. I just want to lay out the entire process to make sure you're actually cutting correctly, because it's really important not to cut the wrong thing.

(Recently I have actually raised the bar to 95%+ "probability of being best" for cutting stuff due to all the inaccuracies introduced by the volatility caused by all the moving parts. 90% is already marginal - I wouldn't recommend going under 90%.)

At this point you can either 1)keep running the 2 remaining landers and see if a winner will emerge soon, or 2)just keep running both and start mass-testing offers.

If you choose 1), and if the landers continue to go head-to-head, just pick one to go with, and start mass-testing offers. (See point 4) of this post by caurmen.)

Or - your decision to split-test lander variations would be fine as well. But really, I would bother doing that unless and until I've found at least one good offer for your target offer type and geo!

Imagine this: You've spent hundreds of dollars on testing 10-20 lander variations, and are able to increase your CR by a large extent. However, when you use it to mass-test offers, none of them end up being in profit...

It'd be up to you though! Split-testing lander variations can give you insight into what lander elements will work better, and that in itself can be a valuable learning experience.


I certainly see what you're saying and it does make sense. You can test your approach to see how it works out for you. My approach is not nearly as complicate as yours is.

Basically nowadays I'm launching so many camps that I don't think much when picking a bid for initial testing - as long as it's not at the bottom where I'd doubt the traffic quality. After you've determined the best lander+offer+targeting combo, you can (and SHOULD!) always test bids to find one that gets you the most profits (not ROI).

I wouldn't suggest testing staggered bids while you're mass-testing offers - that would just be too many moving parts.

You could test bids using your original offer + LP1+8. Or, just continue mass-testing offers at the current bid since you're not in great loss (which means the bid is at least OK).

As to how staggered the bids should be - it's like a gamble. You know what the range of bids are from the popads graph, so you pick 3 points to test in attempt to find that optimal bid. Preferably though you wouldn't want the bids to be too close together, or they would just trigger traffic from the same placements which would defeat the point of testing bids. (Now you may want to ask "well what would be considered as being too close?" ).

I usually separate my bids by at least $50, often $1+ depending on how wide the range of bids is. And remember that you don't have to stop at 3. You can pick the best of the 3 and test bid points around THAT.

But really, too much bid testing may not be worth the time/money - given how volatile the bidding landscape is for pop traffic, and how camps tend to be short-lived.


That basically means "anything we can't identify". If your camp is in profits and you need the extra traffic to see if you could increase your profits some more, try target "unknown" then.


Actually - the idea is not so much to identify segments to cut - I mean of course if you're already running your best offer+lander combo and a traffic segment is still in loss then you're justified in cutting it.

The more important idea is to try to identify segments that are promising.

If, when you're mass-testing offers, after running traffic for a while, you drill down and NONE of the major traffic segments are looking likely to reach green - then you may want to consider ditching the campaign altogether.

Sure enough, the results could go either way. If you had continued running the camp, you MAY end up with some green combo. This is where experience comes in.


When you're cutting anything before you've identified your best offer+lander, there's a potential that the thing you're cutting COULD have ended up being profitable for your best offer+lander.

So your choices are:

1)Cut aggressively first, anything that doesn't look promising, and retest some of that after finding the best offer+lander.
2)Cut aggressively first anything that doesn't look promising, and not retest it. (e.g. If that geo has a ton of traffic and you simply don't care.)
3)Defer the cutting until after you've cut down to your best offer+lander.
4)Only cut the worst stuff while you're still testing offers/landers, deferring the rest of the cutting until after finding the best offer+lander.


Are the landers proven?

If they aren't - then there would be no way of deciding whether it was the offers that weren't good, or if it was your landers.

This was a very valuable exercise you did. It illustrates a couple of things:

1)Testing lots of landers would be a good idea - as that will maximize the chances of there being at least 1 decent one amongst them.

2)Using 2-3 PROVEN offers (that your AM knows have worked for other affiliates) to cut landers would be a good idea, because you need at least 1 offer to convert, in order to have the conversions necessary to help you cut landers.

Great progress! Looking forward to more!


Amy

---

Home > Mobile > Follow-along Campaigns