Web3 Ad Network Campaigns Bought on Wallet Outcomes

The first question is which networks are real: When I test a fresh set of web3 ad networks, two or three out of eight survive the first month. The rest fail on the same tell: click volume arrives on schedule, and then 40 to 70 percent of those sessions end under two seconds with no scroll event, concentrated in a handful of placement IDs the network will not name. This is why campaigns here get bought on wallet outcomes rather than on the network's own dashboard. A connected wallet that signs a transaction is expensive and countable. A click is neither. How a flight is set up: Server-side click logging before anything else, so the click count belongs to you and not to the network. Invalid traffic screening on page with HUMAN or Protected Media. Then the events that matter get instrumented: wallet connect, first signature, first transaction, and a day-30 retained wallet, pushed into your warehouse and mirrored in a Dune query so the on-chain side can be checked by anyone without asking me for a file. Spindl or Safary handles attribution between the click and the wallet where the stack supports it; Addressable covers the wallet-to-social path when the target is holders of a specific contract rather than a broad crypto audience. Networks I currently run and how they get judged: Coinzilla, Bitmedia, Blockchain-Ads, Persona and Slise, plus direct placements on dapp surfaces where the publisher will take a tag. Each starts on the same small budget with the same creative and the same landing path, so the comparison is fair. Placement IDs get cut weekly against a floor set on cost per retained wallet, not cost per click. A network that cannot deliver placement-level reporting is dropped rather than optimised, because you cannot cut what you cannot see. Typically the survivors are two networks and a short direct list, and the plan concentrates there instead of holding eight open for the sake of coverage. What gets refused: No incentivised traffic, task-wall installs, or airdrop-farming inventory, whatever it is labelled. No cost-per-wallet deals with a network that also controls the wallet definition. No bot-filtered traffic reports written by the vendor selling the traffic. I will not run creative that states or implies a return, a price direction or a yield the protocol does not contractually pay, and if your compliance position depends on me doing that, we should stop at the call. Not for: Projects with no on-chain event worth optimising towards, since without a signature or a deposit this becomes a branding exercise bought through the worst inventory on the internet. Teams that cannot ship a tracking change inside a sprint, because the instrumentation is the work and everything after it is placement hygiene. Anyone who wants the eight-network media plan for the look of it. And launches under a month away, where there is no time to run the sieve and you would be buying blind. What you keep: The blocklist of placement IDs and subdomains, which is the compounding asset here and stays yours. The event schema and the Dune queries. The tested-network verdicts with the numbers behind each cut. If you bring an in-house buyer in later, that package is what they inherit, and it is written for that purpose. I work in English and Portuguese, and I do not take two competing protocols in the same category at the same time.