There is a moment in every launch review where someone pulls up the Telegram. Sixty thousand members. Eleven people talking. Nine of them are the moderation team.
Everyone in the room knows what happened. Nobody says it.
The market has gotten better at spotting this. Chainalysis estimated suspected wash trading on decentralized exchanges at up to $2.57 billion in 2024 in its 2025 Crypto Crime Report.
When LayerZero ran Sybil filtering ahead of the ZRO distribution in June 2024, it removed 803,093 wallets, close to 59% of applicants, leaving 1.28 million of the original 2.08 million. Linea’s filtering removed roughly 517,000 of 1.3 million claimants.
Fake numbers have become a detection problem, and the detectors are public.
The harder path produces smaller numbers that survive contact with an unlock. Fewer wallets, fewer impressions, fewer members, and a project that still exists in six months.
Here is how that path works.
Understand what bought attention actually costs
Bought signal has a price, and the price arrives later than the invoice.
CHEQ found that bots or fake users accounted for 75.85% of ad traffic to client websites from X during Super Bowl weekend in 2024.
Independent estimates put bot accounts at between 9 and 15% of active users on the platform overall, with sharply higher concentrations in high-attention conversations. Those are the conversations your launch will be sitting in.
So three things happen when you inflate.
Your own targeting degrades, because retargeting pools fill with wallets that will never buy anything. Your KOL benchmarks become useless because you cannot tell which account generated a real reader.
And exchange listing teams, who run their own wallet analysis, price your application on holder quality, and they do not look at your follower count.
The Nansen review of the ZKsync airdrop found 41% of the largest recipients had sold their entire allocation. Attention that was bought behaves exactly like that, only faster.
Build proof before you build reach
Momentum that holds comes from having something to point at. Most launches get this backwards and spend the pre-TGE window manufacturing noise when they should be gathering evidence.
Ship something people can use before the token exists. A testnet, a limited mainnet, a tool that solves one narrow problem. Usage data from a working product is the only marketing asset that cannot be purchased.
Publish the tokenomics early and completely. Supply, allocation, cliff dates, vesting curve. Teams that withhold this until launch week are read as hiding something, because most of those who do are.
Get audited by a firm people recognize, then publish the findings including the unresolved items. A report with three medium-severity notes and clear remediation reads as more credible than a spotless one from a firm nobody has heard of. Traders have learned to check who signed the audit.
Name your investors and partners with permission, and skip the logo wall of firms who took a call once. Founders check. So do the funds you want in the next round.
Write documentation a developer can follow without asking you a question. This is also how you get into AI answers, because language models cite sources that explain mechanisms clearly.
How earned distribution compounds
Rented distribution ends on the day the invoice ends. Earned distribution compounds.
The distinction matters more than it used to, because the discovery layer has changed. People now find projects through research aggregators, AI assistants, and other builders, and all three weigh substance over volume.
- Publish original research. Announcements get skimmed and forgotten. A dataset, a benchmark, or a teardown of how your mechanism differs will get quoted by people you never paid.
- Get your founder into technical conversations. Podcast appearances, conference panels, and long-form written interviews create secondary citations that keep working for years.
- Feed the research layer directly. Messari profiles, Delphi Digital coverage, Kaito attention tracking, and DefiLlama listings shape how funds discover you, and each one runs on documentation quality.
- Build relationships with reporters months before you need them. A journalist who has already written about your category will take the call. A cold pitch in launch week will not get one.
- Support other teams’ launches without asking for anything. Founders remember, and founders have audiences.
Design incentives that survive the claim
Most launch incentive programs are optimized for the wrong number. They maximize participants, then discover that participants and users share almost no overlap.
Sybil resistance is the first filter, and it belongs in the campaign design from day one. Applied at the end, it becomes an argument with angry farmers.
LayerZero’s self-reporting mechanism let farmers admit what they had done and keep 15% of their allocation, and it surfaced a large volume of activity that on-chain analysis alone would have missed.
Beyond that, a few design choices change outcomes.
Weight rewards by duration held. Action counts are trivially gamed, and duration is expensive to fake. Someone who provided liquidity for six months is a different person from someone who did nine transactions in a weekend.
Distribute in tranches with continued-participation requirements. A single cliff distribution guarantees a sell wall on day one, and everyone reading your snapshot can see it coming.
Make part of the allocation claimable only through product use. If the reward requires actually using the thing, the reward selects for people who want the thing.
Cap the total percentage going to speculative farming. 10 to 15% of supply to a broad airdrop produces a very different holder base than 40% does, and the difference shows up on the first unlock date.
Publish eligibility rules before the snapshot. Retroactive rule changes generate more anger than a smaller allocation ever would.
Let contributors carry the message
The projects with real momentum have people arguing on their behalf in threads the team never sees. That does not come from a rewards dashboard.
Identify the twenty people who show up consistently and give them something real. Early access, direct contact with engineers, a say in roadmap sequencing. Twenty invested contributors outperform twenty thousand passive members.
Pay moderators and pay them properly. Unpaid volunteer moderation collapses the first time a scam-link campaign hits your Telegram at 3am.
Run technical AMAs where hard questions get answered on the record. Deflected questions become clips, and clips travel further than any announcement you write. One founder answering a hostile question well does more for credibility than a month of scheduled posts.
Localize into the markets where your users actually are. Korean, Turkish, Vietnamese, and Portuguese-language communities move independently from English-language crypto Twitter, and organic advocacy in those markets is harder to fake and cheaper to earn.
Hire a native speaker who actually trades. Translation vendors produce copy that reads as foreign to the people you are trying to reach, and the difference shows up in the first week.
Give the community a job. Governance votes with real consequences, bug bounties, translation, documentation contributions. A community with nothing to do becomes a price chat within a month.
Measure what cannot be purchased
Impressions can be bought for a few hundred dollars. Track the metrics that resist that.
Holder count trend measured separately from price. Rising holder count during a drawdown is the strongest signal in the entire dashboard.
Median holding duration for wallets acquired through each campaign. This tells you which channel brought buyers and which brought exit liquidity.
Ratio of product questions to price questions in your Telegram. When the ratio moves toward product, you have users. When it moves the other way, you have a chatroom about a chart.
Returning wallet share week over week, which no amount of paid reach will move.
Branded search volume and citation frequency in AI answers for your category terms. Both reflect whether anything you did created recall.
Unpaid mentions from accounts you have never contacted. Count them manually if you have to. That number is the real one, and it is usually far smaller than the dashboard suggests.
One more worth tracking: how many people in your Telegram can explain what your product does when a newcomer asks. If the answer is only your moderators, the message has not travelled.
Top token marketing agencies with credible token launches
A short reference for teams looking for outside help, with publicly documented work.
Blockchain App Factory ran OpenLedger’s go-to-market strategy over 12 months, from pre-TGE through mainnet, reporting 47 million campaign impressions, more than 6 million registered testnet nodes, and $182 million in launch trading volume. Its published case studies also cover network campaigns for NEAR, Polygon, Hedera, Moonbeam, and Fuel, as well as token launches for LEARN 360, ZAPME, ENERSTAGE, and JPACA.
INORU handles token launch marketing and Web3 go-to-market for early-stage teams, with community operations, campaign execution, and launch support across ICO and IDO formats.
OMNI Agency focuses on Web3 growth and launch campaigns, with community building, influencer coordination, and creative production for token projects.
Turnkeytown offers token development alongside launch marketing, which suits teams that want the product build and the go-to-market handled by one vendor.
ICODA is a crypto marketing firm with a long history in token sale campaigns across PR, SEO, tracker listings, and paid acquisition for launch-stage projects.
The first ninety days
Launch week is the least important week. What happens after it decides everything.
Days 1 to 14 are for support. Stop promoting and start answering. Answer every question. Fix every reported bug publicly. The people watching now are deciding whether to stay.
Days 15 to 45 are for shipping. One meaningful product update inside this window resets the narrative from token to project. Nothing shipped in this window means the story becomes price by default.
Days 46 to 90 are for the second cohort. The launch audience has sorted itself into holders and sellers by now. This is when earned distribution starts pulling people who were never in the campaign at all.
Somewhere in here your first unlock arrives. Keyrock analyzed more than 16,000 unlock events and found roughly 90% produced negative short-term price pressure, with the decline typically starting about 30 days ahead of the date. Plan a real announcement into that window and say something on the day itself.
A launch built on bought attention has nothing to say at day 60, because there was never anything underneath it. A launch built on proof has a product update, a growing holder count, and people who showed up without being paid.
One of those is momentum. The other is a bill that comes due.
Kim Marks 
Nina Smith 