running programmatic buys through an rtb setup and im convinced a meaningful slice of my spend is going to placements that are basically worthless, bot heavy or stacked or just garbage inventory dressed up to look fine. the reporting looks ok on the surface but my post click behavior is terrible on a lot of sources.
how are people cutting the junk without nuking their volume to zero. do you blocklist aggressively and accept smaller scale, do you buy only curated deals, how do you even identify which placements are the rot when everything claims to be premium. feel like im funding fraud half the time.
stop looking at click metrics and look at a real post click action, even a soft one like time on page or reaching a second screen. bots click fine, they dont behave. rank every placement by that downstream signal and youll see the junk separate from the real stuff fast. then blocklist the bottom ruthlessly. yes your volume drops, but the volume you lose was never converting anyway so your effective cost actually improves.
curated and private deals exist for exactly this reason, you trade some scale and a higher cpm for inventory that is actually what it says it is. for adult specifically your clean open exchange options are limited anyway, so leaning on a few direct or curated sources you vetted yourself often beats trying to police the whole open firehose. police what you must buy open, but move your core budget to sources you trust.
watch out for the domain spoofing angle too, a lot of the garbage declares a premium domain it isnt actually on. if your dsp supports it, check sellers against the published seller lists and drop anything that doesnt reconcile. thats a huge chunk of the dressed up garbage you described, it claims to be a known site and simply is not.
one thing that helped me, dont try to clean the whole account at once. take your top spend placements, split them into keep and kill based on real behavior, and just do that tier first. the long tail of tiny placements is mostly noise and youll chase your tail trying to judge each one. fix where the money actually goes and the rest barely matters.
seconding the rank by downstream action approach. id add, build the report so a placement has to earn its way back in, not just get blocked once. inventory rotates names constantly, the junk you block today resurfaces under a new label next week. a standing rule based on behavior catches the reincarnations, a one time manual blocklist goes stale in days.
one more thing, whatever you blocklist, write down why. future you will wonder whether a placement is actually bad or whether you killed it in a bad mood six months ago, and without a note youll just re test it and waste the spend proving what you already knew.
the trap in rtb is judging placements on cost per click when the only number that matters is revenue per session from that source. i build the report so every placement shows what it actually returned, not what it cost to get the click, and the garbage collapses to the bottom instantly because bots and stacked frames generate sessions worth nothing. rank by that, cut the bottom, dont agonize over each line.
ill give you the sell side view since nobody has yet. a lot of what you call junk is real inventory being resold through three hops that each take a cut and blur where it came from, so by the time you buy it you cant see the actual site. buying closer to the source, direct or one hop, isnt just cleaner inventory, its the same impression without the layers skimming and hiding it. the open firehose is mostly resold resold inventory.
one thing from the cheap geo side, some of the junk is not bot, is just real traffic from a geo that never convert for your offer but the exchange mix it in. before you call everything fraud, split by geo, sometimes is depend of the geo not the placement. i waste weeks blocking placements when the real problem was two tier3 geo i should never been buying for that offer.
right, and make that signal one you record server side, not a pixel the placement itself can fake or that fails to fire on the junk anyway. ive seen buyers rank by an onsite event that bots were happily triggering, so the garbage looked fine in the very report meant to catch it. the measuring point has to sit somewhere the traffic source cant touch.
i treat a new rtb source as guilty until it proves out, tiny spend, watch the real behavior for a few days, scale only what survives. the mistake is opening wide and then trying to clean a huge dirty account after. start narrow and let placements earn their way up instead of blocking your way down out of a mess. way less work and way less spend bled in the meantime.
this is the shift i needed, thanks all. i was policing the open firehose line by line and losing, when the real move is rank by downstream behavior recorded somewhere the source cant fake, move the core budget to curated and near source inventory, and only skirmish on whats left open. also the geo point stung a little, gita, because two of my worst sources were a geo problem i was blaming on fraud. reworking the whole report now.
A community for adult-industry professionals. Independent reviews live at
xwebmasters.com.