How do you deal with low payout offers and massive numbers of placements? (8)
How do you deal with low payout offers and massive numbers of placements?
Hi Ladies & Gents,
I am still a bit green with mobile so I am finding new challenges which I haven't had to deal with in my career. My background is mostly Email, PPV and highly targeted display. Mobile has been fun, rewarding, stressful and absolute misery all at the same time so I am hoping for a little insight to help find some balance.
The challenge before me is how to deal with the vast numbers of placements with big traffic available in mobile. I am sure many of you cut your teeth in the display world so it might not seem as intimidating to you but this blows my mind.
And now for a recent example:
I started working with a smaller payout offer ($1.20 US) and built out a campaign for pop traffic to gather some test data. My strategy was pretty straight forward. Test some different angles and landing pages to see what's working. Once I have a good idea of the right direction to move in I would then run some splits from the winning angles and landing pages while running each placement up to around 3x the offer payout while trimming the fat.
Please feel free to chime in on this if you feel my testing methods are way off. If there is something I am missing trust me, I want to know about it!
Hopefully I am not crazy and so far you are thinking to yourself "Yea, so what? What's the damn problem here?".
Well, the problem I am running in to is the number of placements is just huge, it's exploding my budgets and stealing away my youth and ability to experience a nice night's rest. This particular example had over 4,000 unique placements sending traffic. ROI is all over the map on this one. Overall at the end of my testing it was about -75% (and getting better) which isn't shocking when you look at the sheer number of placements which are not even close to significance or 3x the payout (whichever comes first). I simply didn't have the stomach to continue.
Carrying this test through to the end would mean a cost of over $14,000 and running at -75% ROI turns my stomach.
I guess my question is, what am I missing?
Should I just be trying to find bigger wins and higher ROI to begin with?
Should I be somehow cutting the bulk of placements and focus my efforts?
Any advice would be greatly appreciated guys, thanks!
Nick
with that many placements you can cut quicker. if something has 1.5 payout or more with 0 conversions I'd cut it. you may end up cutting some slight winners but in my experience the biggest ROI campaigns wont suddenly jump up from 0 to 100% ROI. Obviously there's a tradeoff here and if you have the budget, test to a higher payout.
I would try to buy only from the bigger sources where you see some performance. then try to optimize (landingpages, creatives, handsets, os etc.) and if those big sources get breakeven/roi positive, you can try to buy on the smaller sources again and search for the gold 
Are all the placements sending roughly even volumes of traffic?
I'd guess that they're not - if that's the case, then concentrate on killing the really big losers, then look at the overall performance of your campaign. If you aren't seeing close-to-positive ROIs after a couple of rounds of killing off placements and testing new landers, then move on to another angle or another offer. As Constantin says, you're unlikely to jump from massively negative ROI to massively positive, at least once you've killed the bot-ridden / otherwise crap placements.
Have you run a bot test on this traffic, btw? I'd strongly advise doing that early, particularly if it's a new traffic source to you.
If the placements are all throwing even volume, then the advice above doesn't work - that's much rarer, though. Inverse-square law rules all. Let us know if that's what you're seeing!
Originally Posted by caurmen
LESSON LEARNED!
I had skimmed over your post about finding bot traffic just the other day, again without giving it much thought. This afternoon I went back to that post and ran some quick tests and sure enough, the numbers didn't add up. I snapped open my IDE and went to work on a quick and dirty utility to help me flush this all out using the same methodology. Javascript firing off a request shortly after page load assumes a high probability of a human being behind the wheel, not perfect but well within reasonable assumption. The results are still flying in and at first I nearly fell off my chair, mouth hanging open. In the following screenshots I have removed the placement identification to protect the apparently guilty. Rows in green represent greater than 80% human traffic while rows in yellow represent placements with less than 80% assumed humans.
As I mentioned in my previous post, there's a lot of placements here so I zoomed out my browser to illustrate the big picture.
Once I have gathered some more data I fully intend to discuss this with the traffic source. Let this exercise be a lesson to everyone here, don't make any assumptions about quality.
You don't know what's there until you take the time to check it out for yourself.
Thank you again Caurmen, this certainly put things in perspective for me.
Nick
Glad it was helpful!
Yeah, bots are a bit of a plague right now. Hopefully the increase in awareness of detection amongst marketers will start to encourage everyone to get smarter about spotting and banning them.
FYI - even a somewhat bot-ridden placement can sometimes be profitable. Generally I set my filters for "just don't touch this with a bargepole" to a bit lower than 80% - 60% or 50% maybe.
I'm still developing my methodology on this one, but I've seen a few placements with a 60% or so ratio of bots to humans still become profitable. Obviously be a bit more willing to cut more bot-tastic placements faster, though.
( Also, bonus points for the BSG reference! )
Thanks for all the info in this thread fellas! Is there an easy way I can bot test my traffic without writing a script / code? I just started with 5 new sources this week and I'm eager to test them out against each other. The amount of bots that show up will probably tell me right away who I can trust.
@sportstraffic - not really, but the scripts are very easy. See http://stmforum.com/forum/showthread...367#post180367
Home > Paid Traffic Sources > Mobile