Farming an airdrop is described as free money and it is a job with an unstated wage, unstated hours, and a payment the employer decides after the work is done. Understanding what the payer is buying makes the arithmetic clearer, and the arithmetic is the part nobody does.
A project distributing tokens to users is doing customer acquisition, paid for in its own equity rather than in cash. It wants usage that makes the protocol look alive, holders who have a reason to care about its governance, and a distribution wide enough that the token is not visibly concentrated in the founders and the early investors.
That framing answers most of the questions people find confusing. The criteria are secret because publishing them would cause everybody to satisfy exactly those criteria and nothing else, which would produce activity that looks like usage and is not. They change after the fact for the same reason. And they reward behaviour that resembles a real user rather than behaviour that maximises a metric, because the metric was a proxy for the thing being bought.
It also explains the asymmetry that farmers find unfair and that is entirely deliberate. The payer decides the amount, the date, the eligibility and the vesting, after seeing what everybody did. You perform the work first and learn the wage afterwards, and there is no obligation on anybody to pay anything at all.
You are doing unpaid work for a company that will decide afterwards what it was worth, if anything. That is the arrangement, and it is not hidden.
Every serious distribution now runs filters designed to exclude one person operating many addresses, since a distribution captured by a handful of participants running hundreds of wallets defeats the purpose. The filters look for shared funding sources, similar transaction timing, identical behaviour patterns, and addresses that interacted only in the ways the criteria rewarded.
They work, imperfectly, and the imperfection cuts both ways. Genuine users get excluded because their behaviour resembled a farmer's, which is unsurprising given that farmers were deliberately imitating genuine users. Appeals processes exist, they are slow, and the outcome is at the project's discretion because nobody was owed anything in the first place.
The practical consequence is that the strategy of many small addresses has become substantially worse than it was, and the strategy of one address behaving like an ordinary user has become relatively better. That is exactly the shift the filters were designed to produce, and it will keep moving as both sides adapt.
The activity is not free and the costs are real. Every transaction pays a network fee, and a farming pattern that involves dozens of interactions across several protocols accumulates a meaningful sum, all of it paid in advance of any possible payment. Capital deposited to qualify is capital not doing anything else, for a period nobody has specified.
Then there is exposure. Qualifying frequently requires holding an asset, providing liquidity, or leaving a balance in a contract, each of which is a position with its own risk. A farmer who provides liquidity to qualify is exposed to divergence between the two assets, and a farmer holding a volatile asset to meet a threshold is holding a volatile asset. Neither of those was the intended trade.
And there is the largest cost, which never appears in any calculation because it has no invoice: the time. Hours spent researching criteria, executing transactions, tracking eligibility across projects, and monitoring announcements are hours, and any honest assessment of whether the activity paid has to price them at something.
A distribution creates a large number of holders who acquired the token at no cost and have no reason to keep it, which is a specific and predictable market structure. The selling on the first day is heavy, it is concentrated, and it is entirely rational: somebody who received something for free and wanted cash will convert it, and a substantial share of recipients are in that position by design.
Projects respond with vesting, cliffs and lock-ups, which spread the selling rather than removing it and create dated events afterwards where the same pressure arrives on a schedule anybody can read. Those dates are public and they are one of the more reliable pieces of information available about a newly distributed token.
The pattern this produces is well documented and unsurprising: heavy initial supply meeting demand that was manufactured by the distribution itself. Whether it recovers depends on whether the protocol has genuine usage underneath the farmed usage, which is the question the distribution was supposed to answer and frequently answers badly.
The current form of most campaigns is a points system: activity earns points, points are displayed on a leaderboard, and a future distribution is implied without ever being stated. The implication does substantial work, and the absence of a statement is the point of the design.
What a points programme actually commits to is nothing. There is no obligation to distribute anything, no defined conversion rate, no date, and in most cases explicit language reserving the right to change everything. Participants behave as though a rate exists because a leaderboard makes a number feel like a balance, and the number is a score in a game whose prize is undeclared.
This is worth stating without cynicism, because the arrangement can still be worth participating in. It is worth participating in as a lottery ticket bought with time and fees, not as a receivable. The distinction changes how much of either you should spend, and it is the distinction the interface is designed to blur.
A points balance is a score, not a claim. Nothing has been promised, and the leaderboard is built to make that easy to forget.
Announced distributions attract a second industry, and its methods are consistent. Fake claim pages appear before and after every real one, ranking in search results and in social feeds, and they ask you to connect a wallet and sign a transaction that grants a spending permission rather than claiming anything. The loss occurs later, from an address that signed one thing.
The second pattern is the unsolicited token that appears in a wallet, appearing to be worth something, whose sale requires interacting with a contract designed to drain the wallet in the process. Receiving it is harmless; interacting with it is not, and the appearance of value is the entire mechanism.
The defence is procedural. Claim only from a link you navigated to through the project's own established channels, never one that arrived. Read what you are signing and specifically whether it is a claim or a permission. And do it from an address that holds nothing else, since the whole category of loss requires the signing address to be worth draining.
Receiving tokens creates a record-keeping obligation in most jurisdictions, and the practical difficulty is not the rules but the data. Tokens arrive with no accompanying document, at a value that has to be established as of a date, from an entity that sent you no statement. Reconstructing that a year later, across a dozen projects, is a genuinely hard exercise and it is why people leave it undone.
The habit that solves it costs a minute per event and is impossible to reconstruct afterwards. Record what arrived, on what date, at what address, and what it was worth when it arrived. A spreadsheet is sufficient. Doing it at the time turns a difficult reconstruction into a lookup, and the difference is the difference between a task and a problem.
The same applies to costs. Network fees paid in pursuit of a distribution are part of the picture, they are recorded on the chain, and they are far easier to total as they happen than to extract from a year of transaction history. What your local rules make of any of it is a question for somebody qualified in your jurisdiction; having the numbers is the part that is entirely up to you.
The calculation that makes the decision is short and rarely written down. Estimate what a distribution might be worth, discount it heavily for the probability that there is none, that you are filtered out, or that it vests over years. Subtract the fees you will pay with certainty. Subtract the value of the capital committed for the duration. And put a number on the hours.
Done that way, the activity is worthwhile for a small number of participants and not for most, and the ones for whom it works are usually those who were going to use the protocol anyway. That is the group the distribution was designed to reward, which is not a coincidence: the payer is buying genuine usage, and the people who supply it cheaply are the people who wanted it.
The version that reliably does not work is committing capital and hours to protocols you have no interest in, on the expectation of a payment nobody has promised, with criteria you cannot see and filters designed to exclude exactly what you are doing. That is a real description of most farming, and it is worth reading before rather than after.
Because publishing them would make everybody satisfy exactly those criteria and nothing else, producing activity that looks like usage without being it. The project is buying genuine usage and a wide distribution, and a published rule would be optimised against immediately. The same reason explains why criteria change after the fact.
It is unpaid work for a company that decides afterwards what it was worth, if anything. The fees are certain, the capital is committed for an unspecified period, the hours are real, and the payment is discretionary. That can still be worth doing, and it is not free.
Filters designed to exclude one person running many addresses also exclude genuine users whose behaviour resembled a farmer's, which is unsurprising given that farmers were imitating genuine users. Appeals exist, they are slow, and the outcome is discretionary because nobody was owed anything.
No. A points programme typically commits to nothing: no obligation to distribute, no conversion rate, no date, and usually explicit language reserving the right to change everything. A leaderboard makes a number feel like a balance, and it is a score in a game whose prize is undeclared.
Because the distribution created a large number of holders who acquired the token at no cost and have no reason to keep it. The selling is heavy, concentrated and rational. Vesting spreads it rather than removing it, and the vesting dates are public and worth reading.
Two patterns. Fake claim pages that ask you to sign what appears to be a claim and is actually a spending permission, with the loss occurring later. And unsolicited tokens appearing in a wallet whose sale requires interacting with a contract designed to drain it. Claim only from links you navigated to yourself, and use an address holding nothing else.
What arrived, the date, the address, and its value at the moment of receipt, recorded at the time. Tokens arrive with no document and no statement, and reconstructing that a year later across several projects is genuinely hard. Also total the network fees as they happen rather than extracting them later.
For a small number of participants, usually those who were going to use the protocol anyway, which is exactly the group the distribution was designed to reward. Committing capital and hours to protocols you have no interest in, expecting a payment nobody promised, under hidden criteria and filters aimed at what you are doing, is the version that reliably does not work.