Hi,
Imagine you have 10 offers for a test. You start testing 'em, throw out bad offers, keep good ones, blacklist placement, whitelist placements, etc., all good. But... Suddenly, a new bunch of offers comes in. You insert 'em into current campaigns, and see that they convert way better than previous ones.
Now the problem is should you bring back old placements that you previously blacklisted, as there could be a chance that they would work perfectly with current offers? In other words, I'm trying to find a good algo on testing *NEW* offers in the niche I've already worked with. If I start from scratch, I could spend money on previously blacklisted placements and probably lose money. On the other hand, I could miss good placements that worked bad previously just because previous offers were bad. I could use another strategy and re-test previously blacklisted placements but spending half less than in previous iteration.
So, I'm kinda lost here, and I'd appreciate if someone help me with an optimal strategy here. Thanks.
Your logic is right : new offers should be tested in the campaigns that you have already optimized. In case you see they have better performance than the old ones, it's quite possible that some of the already paused targets could make profits with these new offers.
You can do 2 things now basically : look at the stats and un-pause whatever was close to profits. Or, make a new campaign and start from scratch again - in this case you would still want to blacklist the very worst placements straight away - chances are that the rally bad ones can't be brought to life anyways, even with a great offer.
The truth is, after some time, campaign becomes over-optimized... since we keep on cutting, pausing and optimizing ... so it's a good idea to restart it clean at some point anyways. And a good new offer might be the right impulse for doing so.
Make a note when adding a new offer. Compare the different offer performance stats only from that date/time on.
If a new offer performs much better, you can do what Matuloo suggested.
I would test the new offers on the best placements first. Only if an offer is profitable on those, would I re-test some of the previously-blacklisted placements.
Also, I wouldn't suggest to bother retesting 1)small placements or 2)placements that are in loss by hopeless amounts. In other words, just pick placements that are fairly big that are not in major loss. For example if a placement is in loss by more than 10x average payout, chances are it won't turn out profitable even for your best offer, unless all the offers you previously tested were ALL hopeless duds.
As always, you'd need to find a balance. Retest everything = more expensive but you'd be fair to every placement. Retest nothing = less expensive but you'd leave out potentially-profitable placements. The sweet spot may be to retest only the bigger placements that have not lost too much money in the past. There's no need to squeeze every drop of profit from each camp - after a certain point, you'll find that your valuable time could be better spent setting up new camps.
Amy
One additional note: if you're going to retest previously blacklisted placements, it'll really benefit you to do a bot test on those first.
That way you can quickly eliminate the obvious-fraud placements, and just retest the ones that have actual humans visiting them!
What are bad placements?
Most of the time you kill placement because it doesn't convert at all ie. the traffic is artificial coming from a BOT.
If the traffic is from BOTs you can have the best offer in the world and it will never convert.
Why some placements have bot traffic?
Well our good old publishers are trying to game the system, drive robots to their sites and we pay for it.
That's clear.
Then there are placements with humans mixed with bots, which are more difficult to decide if we are going to kill them or no, but your profits/losses ratio should give you the answer.
The fact is that there are just handful of placements that will bring you in money - 80/20.
Retest the placements that gave you conversions, but those that just eat up mountains of budget and you know their are bots .. well stay away from them.