Galxe and Zealy Quest Campaigns With Sybil Screening First

On the last four quest campaigns we ran across Galxe and Zealy, between 61 and 78 percent of completions came from wallets our screen classified as farming clusters. None of them were paid. The screen is written, agreed and published before the reward is announced, because a filter designed after the fact is a filter you will be argued out of. The screen: Clustering runs on funding graphs first: wallets funded from the same exchange withdrawal inside a time window, or from the same intermediate hop, are treated as one participant until proven otherwise. On top of that we look at gas history entropy, contract interaction diversity outside your campaign, and account age weighted against first interaction with your contract. Identity signals sit as one input among several, never as the whole test: Human Passport stamp thresholds, Galxe Passport where a campaign genuinely needs a KYC tier, and reputation scores from Nomis or Trusta read as evidence rather than as verdicts. Every rule carries a stated false-positive cost, and there is an appeals path with a human reading it, because a screen with no appeal will eventually exclude a real user with an unusual wallet. Quest design: Tasks map to the product action you want repeated. If the retained behaviour is a deposit, the quest is a deposit, at a size that means something. Follow-and-repost tasks are excluded from reward-bearing campaigns; they price your token against a bot's marginal cost and teach the audience nothing about the product. On Zealy we design the XP curve and sprint structure and we staff the manual review queue, which is where most campaigns quietly fail, since unreviewed submissions become the reward distribution. On Galxe we set OAT and points logic and decide with you whether the credential should sit on-chain through the credential API or as an off-chain snapshot, which changes both cost and what you can prove later. Distribution: Merkle claim contract review, claim window length, and the sweep policy for unclaimed allocation, agreed in writing before launch. Where cliffs or streaming are used we work with Sablier or Hedgey rather than bespoke vesting code. The reward split between screened participants and the excluded pool is your decision; our position is that a partial allocation to borderline clusters buys less peace than people expect. What we do not do: We do not buy completions, and we do not work with agencies that supply them. We do not write your token's legal structure, tax treatment or jurisdiction analysis, and we will ask to see counsel's position before we start. We do not run points programmes where the conversion is undefined; if you cannot say in one sentence what a point becomes and roughly when, we will decline, because that campaign spends trust you will need later. We do not chase quest-platform leaderboards as an objective. Leaderboard position is bought by farmers more cheaply than by anyone else. Who this is not for: Protocols whose real goal is a headline participant number for a fundraise. We will produce the smaller true number and you will not enjoy the meeting. Teams unwilling to publish the screen before the reward. Teams who want the campaign to run in three days; the screen needs contract history to read, and a campaign launched into an unprepared claim flow costs more in support than it saves in time. What you get at the end: A participant file with cluster assignments and the rule that fired for each exclusion, the forkable Dune query behind it, and a retention read at day 30 covering how many screened participants performed the target action again without an incentive attached. That last figure is the one we ask to be judged on.