Wallet-Targeted Campaigns Across Web3 Ad Networks

Wallet targeting is worth paying extra for only when the postback comes back at wallet level. A network that sells holders of a given asset and then reports impressions and clicks is selling an ordinary CPM with a story attached, and roughly one in five of the networks I evaluate survives that test. The campaigns I run are built on the ones that do, and the rest of the list stays in a rejection file you can read. Networks and tools: AdEx, Coinzilla, Bitmedia, Slise and Blockchain-Ads on the demand side. Persona and Addressable for building on-chain cohorts and matching them out to X, Meta and Google where match rates justify it. Spindl or a warehouse-side equivalent for attribution. I do not run all of these at once. A campaign usually settles on two networks and one attribution path, because reconciling more than that costs more analyst time than the incremental reach returns. Audiences worth building: Holders above a balance floor in a named asset. Wallets that bridged to your chain in the last thirty days. Interactors with a competing protocol at a depth that implies use rather than airdrop farming. Liquidity providers above a position size. Dormant wallets from your own product, for reactivation. Each definition is written as a query against on-chain data that you can re-run yourself, not as a segment name picked from a network's dropdown. The limits, stated before you spend: Wallet clustering is inference and it is wrong often enough that cohort sizes should be read as ranges. Users hold several wallets and behave differently in each. Match rates from on-chain cohorts to off-chain platforms typically land between fifteen and thirty-five percent, and the matched portion does not behave like the unmatched portion. Any plan that treats these audiences as deterministic will produce a reporting number nobody can defend in a board meeting. Measurement: One event schema from impression through click, session, wallet connect and first transaction, with the wallet hashed and the join performed server-side. A holdout cell in every campaign of any size, either an audience cell or a geography, agreed before launch. Incrementality calculated after the fact is a negotiation, not a measurement. Decay and exit: A new network is cheap because its inventory has not been bid up yet, and that advantage compresses over two to three quarters. Every campaign I run carries a written exit trigger: the acquisition cost at which we stop, the frequency at which we stop, and the date on which we review whether the arbitrage has closed. Staying in a channel after the reason for entering it has gone is the most expensive habit in this field, and it is almost always defended with a chart. What I refuse: Networks that will not hand over a placement domain list, because a meaningful share of web3 display inventory is made ground. Impression packages sold as reach. Retargeting that passes wallet addresses to a third party without a corresponding line in your privacy notice. And I will not report a wallet-level result that I cannot reproduce from raw event data in front of you. Who this is not for: Products with nothing for a wallet to connect to, and teams without a data engineer who can stand up the event pipeline. If nobody on your side owns the warehouse, the first month of this engagement becomes plumbing that you should build before hiring anybody to buy media. How it runs: One week to audience definitions and network selection, one to two weeks of instrumentation with your engineers, then flighting with weekly reads. Reporting comes out of your warehouse rather than my slides, and the queries stay behind when I leave.