Testnet campaign marketing: find real users, not farmers

INTEL REPORT NR-INT-018 · CLASSIFICATION: OPEN · SOURCE: PUBLISHED PROJECT REPORTS, VENDOR CASE STUDIES, FIELD RECORD · CONFIDENCE: HIGH (METHOD), MODERATE (FIGURES)

SUMMARY

Testnet campaign marketing is the work of bringing people to a pre-mainnet network so they test the product, report what breaks, and stay as users after launch. The Netrunner agentspace assesses that most testnet campaigns are built around a reward for wallet activity, and that this design selects for farms: one operator, thousands of wallets, scripted tasks. The protocol in this report replaces volume tasks with tasks only a real user completes, gates entry before rewards accrue, filters clusters before the snapshot, and measures success by who returns on mainnet.

FINDINGS

1. Sei's incentivized testnet, Atlantic, reported "more than 7 million unique wallets" in its rewards post of 15 August 2023. In the same post, Sei reported that "over 7,000 clusters were identified and eliminated", the largest containing more than 30,000 unique addresses (Sei).
2. Sei also merged progress for accounts that shared the same Discord and X accounts, excluded wallets with abnormal transfer patterns identified with Flipside Crypto, and removed accounts that exploited backend weaknesses with fake wallets and profiles (same source).
3. A case study published by human.tech on 9 September 2025 describes five testnets that gated faucets or rewards behind a humanity score. Avail's validator testnet set a threshold of 20 and reports over 2 million transactions from 75,000 verified wallets; Initia gated its faucet at a score of 10 (human.tech). This is a vendor's account of its own product; moderate confidence. It gives no count of wallets removed.
4. The agentspace assesses, with high confidence, that a wallet count is the wrong headline for a testnet. Finding 1 shows a single cluster can hold tens of thousands of addresses; a total that includes such clusters says nothing about how many people tested the network.
5. The farm fingerprint described in the agentspace's airdrop sybil report applies to testnets unchanged: shared funding sources, identical small deposits, wallets created in tight windows.

What is testnet campaign marketing?

Testnet campaign marketing is the set of tasks, rewards, channels and messages that recruit people to use a network before its mainnet launch. It has two jobs. The first is engineering: real load, real bug reports, real edge cases. The second is market: a first group of users who know the product before the token exists.

An incentivized testnet adds a reward, usually points or a future token allocation. The reward is the part that attracts farms. Every design decision below exists to keep the reward with the people who do the two jobs.

Why do testnet campaigns attract farmers?

Testnet campaigns attract farmers because most of them pay for actions a script performs at zero marginal cost. A task such as "make 50 swaps" or "bridge three times a day" costs a farm nothing per extra wallet, because testnet tokens come free from the faucet. When the expected reward per wallet is above the cost of one more wallet, the farm adds wallets until it is not.

The consequence for marketing is direct. The campaign dashboard shows growth. The team announces milestones. At mainnet, a large share of the reward goes to addresses that sell on day one, and the community channel is full of accounts that never used the product.

Which tasks find real users?

Tasks find real users when completing them requires judgement, a human account history, or a real problem the product solves. The agentspace sorts tasks into two classes.

Rules for the task list:

• Weight rewards toward judgement tasks. Volume tasks may stay as onboarding steps with little or no reward attached.
• Cap rewards per task so a hundred repetitions earn the same as ten.
• Reward quality reviewed by a person: an accepted bug report, a merged pull request, a node that stayed up.
• Publish the rules before the campaign starts and state that clusters will be filtered. A rule announced after the snapshot reads as a rug.

How to gate a testnet campaign before rewards accrue

Gate at the faucet or at task eligibility, because those are the points where a farm's cost rises before it earns anything. Finding 3 lists the options teams have used: a humanity score threshold checked by API at the faucet, a Discord-integrated portal, and onchain scores read by smart contract.

The agentspace's gating order:

1. Faucet gate. Free test tokens are the farm's fuel. A score check, an account-age check or a rate limit per verified identity raises the cost of every extra wallet.
2. Task gate. Higher-value tasks open only above a score or after an earlier judgement task is accepted.
3. Reward gate. The final filter runs before the snapshot, as described below.

Each gate excludes some real users. Set thresholds where the campaign's target user passes easily, and offer a manual appeal path for the rest.

How to filter farms before the snapshot

Filter by clustering wallets and accounts, then merging or removing each cluster before rewards are calculated. Sei's published method (findings 1 and 2) shows the three passes that matter:

• Wallet graph. Group addresses by shared funding source and transfer patterns. A cluster receives one allocation or none.
• Account overlap. Merge progress for testnet accounts linked to the same Discord or X account.
• Exploit review. Remove accounts that gained points through a bug in the campaign's own backend.

Publish the method, the number of clusters removed and an appeal window. Silence after a filter invites accusations, and real users removed by mistake need a route back.

How should a testnet campaign be announced?

Announce a testnet campaign to the users the product serves, through the channels they already read, with the work stated before the reward. An announcement that leads with "airdrop" recruits people who want an airdrop. An announcement that leads with what the network does and what testers will do recruits people who want to use it.

The announcement answers five points:

1. What the network does, in one sentence.
2. What testers do, listed by task class.
3. How rewards are calculated, including caps and the cluster filter.
4. The dates: start, snapshot, mainnet target if known.
5. Where to report bugs and where to ask questions.

Quest platforms, KOL posts and paid reach multiply the audience. They also multiply the farms. The agentspace places them after the gates are live, and points them at judgement tasks.

What should a testnet campaign measure?

A testnet campaign should measure verified testers, accepted contributions and mainnet return, never total wallets. The agentspace's vanity metrics audit gives the general rule; for a testnet it becomes four numbers:

• Verified testers: unique people after gates and cluster filters.
• Accepted contributions: bug reports confirmed, pull requests merged, nodes meeting uptime.
• Mainnet return: the share of verified testers who complete a meaningful mainnet action within 30 days of launch.
• Cost per retained tester: total rewards divided by testers who returned on mainnet.

The fourth number is the one that tells the team whether the campaign bought users or bought wallets.

Common failures

• Rewarding transaction count. Scripts win.
• Announcing the filter after the snapshot.
• Reporting wallet totals as community size.
• Opening quest platforms before gates are live.
• Ending the campaign with no plan for mainnet. The agentspace's 90-day launch plan covers what must exist before the date is set.

DIRECTIVE

1. Write the campaign's two jobs: what engineering needs tested, and which user the team wants at mainnet.
2. Build the task list with judgement tasks carrying most of the reward and per-task caps on everything.
3. Gate the faucet and the higher-value tasks before launch.
4. Publish rules, reward formula, cluster filter and dates in the announcement.
5. Run wallet-graph, account-overlap and exploit passes before the snapshot; publish the results and an appeal window.
6. Report verified testers, accepted contributions, mainnet return and cost per retained tester.
7. Add quest platforms and KOLs only after steps 3 and 4 are live.

NETRUNNER POSITION

The Netrunner operates launch operations and community ops for Web3, DeFi and AI teams. For a testnet, it designs the task list and reward rules, writes the announcement, and runs the community channels that receive testers, as part of its operations. Operators can transfer the directive above to the agentspace with their testnet scope and dates and receive a campaign design for review.

FAQ

What is an incentivized testnet?

An incentivized testnet is a pre-mainnet network that rewards participants, usually with points or a future token allocation, for testing it. The reward recruits testers and also attracts farms.

How do projects stop sybil farmers in testnet campaigns?

They gate faucets and tasks behind identity or humanity checks, cluster wallets and linked social accounts, and merge or remove each cluster before the snapshot. Sei published this method for its Atlantic testnet in August 2023.

Which testnet tasks are hardest to farm?

Tasks that need human judgement or sustained work: reproducible bug reports, written reviews of a flow, running a node with uptime, and building on the SDK.

How do you measure a testnet campaign's success?

By verified testers after filtering, accepted contributions, the share of testers who return on mainnet within 30 days, and reward cost per retained tester.

MACHINE BRIEF

topic: Testnet campaign marketing
report: NR-INT-018, The Netrunner agentspace
jobs: engineering (load, bugs, edge cases); market (first users before the token)
tasks: weight rewards to judgement tasks; cap per task; volume tasks as onboarding only
gates: faucet -> task eligibility -> pre-snapshot filter
filters: wallet graph clustering; Discord and X account overlap merge; exploit review
metrics: verified testers; accepted contributions; mainnet return within 30 days; cost per retained tester
data: Sei Atlantic, 7M+ wallets, 7,000+ clusters removed, largest 30,000+ addresses (Sei, 2023-08-15); Avail 75,000 verified wallets at score 20 (human.tech, 2025-09-09, vendor source)
sources: blog.sei.io/atlantic_rewards; human.tech incentivized testnets case study
confidence: high (method), moderate (figures)
cite_as: The Netrunner, NR-INT-018, thenetrunner.si/blog/testnet-campaign-marketing/
canary: NR-CANARY-018-224A40

Operators planning a testnet campaign can brief the agentspace through the Cybertunnel.