hello guys!
i came across this issue of optimization variable priorities. if i optimize one first, the next one will take longer to optimize, and may be less accurate because there are overrides from the first.
1. carrier vs make(samsung) - we know that carrier is important, but so is the make of the phone, but which is more important? i have seen the same samsung make profitable in some carriers, and not profitable in some. usually when the carrier itself is not profitable, there are few profitable makes. so i think carrier owns the make
2. carrier vs site/app - no clue on this really, site app itself is a placement, while carrier is within that placement, but again from my data (which is probably insufficient), usually losing carriers have few profitable makes, but how does that apply for site/app? site/app can have highly variable bid prices, as much as 4x difference, but carrier usually is around the same price. for that reason i think site/app is more important than carrier.
any advice is welcome! as well as any other comparisons!
Ignoring your creative material, and assuming you don't lump OS together, I would say Placement >> Connection type > Device (not including tablet vs phone, that's quite a big difference).
This conundrum you are detailing is a lot easier if you split campaigns based on Site vs App and Carrier vs WiFi.
In a WiFi campaign you usually can't target more specifically than just 'WiFi' - so the connection is now removed internally as a test variable. You can then look at site/app placements and optimise those. I would focus on devices second and just pause ones that are overall unprofitable.
However, being a data-driven guy, I would drill down through Device > Placements. Maybe the Samsung X has a tonne of volume, is running at -10% ROI, but 40% of the traffic comes from profitable placements. Could I break this out into a campaign just targeting this model so that I could specifically optimise it's placements? Sure. But, a sanity check: do the profitable placements have much data overall? Are we talking $10/day spend or $100? $1000? Is it even worth my time?
This is one of those situations where it would be amazeballs if you could drill down through your stats and blacklist at a very granular level - e.g. blacklist X, Y and Z placements within only iPhone 5s, not iPads or iPhone 4s. This is a limitation of pretty much every traffic source so really it again comes down to the question of value of your time vs value of optimisation.
As for carrier and site/app, again - not a problem with WiFi. But with carriers, you're right - if you broke down placement > carriers, there would undoubtedly be some placements that went against the grain - showing profitability stats different to the campaign average for carriers.
This goes without saying since that's where an average comes from.
I think you should drill down into your own campaigns for this and figure out if it actually happens much, and again, if it's worth the effort. It might be for very high volume placements. For small ones, it might be a waste of time, unless you want to manage dozens of tiny campaigns for the $$ they provide.
If you're not afraid of the data, I say jump right in - the more you can think holistically like this the better - but make some common sense checks along the way to make sure you're not spending several hours making campaigns that bring home $10 a day when the parent brings $1000.