Maybe the final one. Campaigns reached predefined maximum bids, so now placement exploration is complete. No traffic bumps are expected anymore. Blacklist campaign bids are randomly changed (between min and max) once per day, to help define bid-traffic curve for remaining placements...
A couple of interesting things have happened over the last few days.
AdsTerra social bar campaign has got the first green day. It's just 3% ROI, and may be a coincidence, but anyway it's inspiring.
PropellerAds in-page campaigns perform worse. High activity is burning about $6-7...
Nah! Change of plans :))) I'm manually removing bad prelanders and offer from the flow, retraining neural networks (to optimize prelander itself, hopefully), and relaunching campaigns after training is complete, though it might take more than a half of day.
Update #8, probably the last one.
All campaigns were stopped several days ago, since after spending about 3-4 days I could not find a quick fix for ML issues. Now it's time to draw some conclusions, fix the problems and prepare for the next campaign.
I'm revealing countries and offers this...
Campaigns still running... Bids are going up every day, ROI stays almost the same. A couple of days ago Leadbit AM contacted me telling that 2 of the conversions were fraud and will be declined by their anti-fraud system, the rest are ok.
Here are PropellerAds campaigns that were...
3 campaigns still running. So far I've managed to split PropellerAds High and Low activity campaigns into tiers by (hopefully) optimal bids. Most placements are still in testing phase, so they are in default tier and their bids are increasing progressively, so campaign costs are...
No changes were made to the last remaining campaign. It did not perform well yesterday, giving 2 conversions and -72.38% ROI. Today, however, I've started a search for better placements by raising bids, since the best creative and offer are already known. From now on, CPM bids will be...
One more addition.
Statistically there is no reason to block Samsung Internet browser, but I will do so since it did not bring any conversions
There is insufficient data for further optimizations. So, I'll just block underperforming creatives and Samsung browser, then will create whitelist...
Addition to the last update.
Here are screenshots from the tracker stats. Conversion layout is different from PropellerAds/Leadbit, since tracker adds conversions to corresponding visit date rather than the date of conversion. Column names are somewhat cryptic, but I hope they are sufficiently...
Had no chance to post update yesterday, so here are traffic source side stats for past 2 days.
And total for all 5 days of running these campaigns:
I've already blocked 3 campaigns out of 4...
Thanks for your offer @DEADZ, I really-really appreciate that. Checked your thread, your findings together with Kintura comments are sound with direction I'm moving to, but Kintura is so far ahead, and mine is just a DIY-level stuff built out of curiosity and personal interest.
Absolutely :) In the ideal world, bid/volume relation (green line) and potential profit (orange line, considering funnel CR and payouts) per placement look like this:
It took me some time to find this example in a different campaign. It's quite rare, I'm mostly having placements like this...