Testing confusion, 3x payout, collecting data??? (6)
So I am currently testing angles and I'm confused on how I should proceed. I read somewhere that I should spend 3x the payout before I cut it. I actually spent 5x payout without any conversions, but I spent that in 4 hours and paused the campaign. Did I pause this too soon? Should I have stretched the 5x payout over 24 hours to collect more data? The payout is $2.30 so I spent about $10. If it helps I am running sweeps, the geo is Australia, and the traffic source is PopAds. Thanks for any advice.
with pops you need to cut at least 50% of placements to get anything near of break even. often especially the big placements perform really bad. could be that 2-3 bad placements are pushing 80-90% of your traffic. With pops 10$ is nothing. If you want to succeed with pops, you're more in the range of 100$-500$ to test an offer. Once you know the bad placements of a traffic source, it will be less. you need to work your way through shit metric tons of bad placements. Cut out bad placements after 1000-3000 impressions without conversion (stretched throughout the day).
ps: if you're on low budget, that list should help to speed things up http://stmforum.com/forum/showthread...pads+blacklist
Originally Posted by sebastian_r
Whether $10 is enough to test an angle will depend on several things - e.g.:
-How well your offer is converting. I would actually suggest that you split-test offers from the beginning. There is a sh*t ton of AU sweeps offers. I would ask my top 3 favorite AMs for their top AU sweeps offer - that would be 3 offers to test.
-How good the traffic quality is. If for example one of the placements was responsible for a big chunk of that $10 spend and it's a bad placement (btw you can't tell at this point whether it's bad or not), then maybe you should spend more.
Those would be the biggest factors I could think of.
Actually - I'm going to suggest this approach if you don't have previous experience with sweeps.
1)Pick 3 offers as was suggested above. This way you can minimize the chances of working with a dud offer.
2)Do a lot of lander spying and rip 5+ landers that you see often which look the most promising to you, and mod them up with the various ROI-boosting code snippets and CTR/CR tricks. This way you can establish a sort of benchmark, and minimize the chances that your lander is bad.
3)Start a test campaign with the 3 offers and the 5+ landers. Keep checking offers and landers for statistical significance and cut until you're left with the best offer and lander.
4)Start 2 separate campaigns.
Campaign A: Compile a bigger list of offers (another 3-5 maybe) and use your best lander so far to test them.
Campaign B: Use the best offer so far to mass-test landers. You can take the most promising lander(s) from your initial test, improve them further, swap to other themes, completely rewrite the ad text etc. And you can also do more spying and ripping and modding and add those in. And of course you can design landers from scratch as well.
And of course, every time CampA uncovers a new best offer, swap it into CampB. And every time CampB uncovers a new best lander, swap it into CampA.
Throughout your testing, you'll want to blacklist bad placements. Just try to only blacklist the worst ones so that you don't stifle the amount of traffic you get, because the more traffic you have the faster your testing will go. Once you have a good offer + lander combo and your camp is green, you can then cut placements / devices / OSs / times of day etc. that aren't giving you a certain ROI (I like to set mine at 30%). You may also want to retest some of the placements you blacklisted from before because they may be profitable now with better offer + lander. After that you can scale to other networks.
In short, it's easier to split-test and cut comparatively worse-performing candidates than trying to answer a question like "is $10 sufficient to test an angle for a $2.30 payout offer". The reason being that there is just too much uncertainty. Whereas with the approach above you can minimize those uncertainties. Hope I haven't made it more complicated than it needs to be.
Amy
Wow thanks Amy! This method of testing sounds way better than mine lol. I will implement this and see how I do. One question, since I am testing offers from the beginning, for the 5+ landers should I keep all the angles and headlines the same? Or should they all be different?
Glad you found the post helpful. 
For the initial landers, it would be better for them to be as different as possible to "cast a big net" so to speak. This would be a good approach to testing other things as well - cast a big net at first, then zero in on areas that look the most promising. For sweeps for example, you have the quiz type lander, wheel of fortune type lander, 3 products to pick from type lander, etc. Theme-wise you have the fb theme, system message themes, whatsapp theme, snapchat theme, etc. So your initial testing should include at least one of each. Then, for example if one of the landers is performing better than the others, you may want to use the same copy with different themes, and/or use the same theme with different copies, and/or test it with different headlines, etc. Hope that helps!
Amy
Home > Paid Traffic Sources > Mobile