DeFi Is Renting Its TVL and Calling It Growth

You paid for TVL. Do you know what you bought?

Every incentives team knows the chart. Emissions start, TVL rises, dashboards light up, and the program looks like it worked. Then emissions end and the line returns to where it began. The protocol did not buy loyalty. It rented balance.

The harder questions come after the chart. Who was paid? Where did their capital come from? What did each dollar that stayed actually cost? Most programs cannot answer, because they were never designed to.

The program paid for the wrong event

Most incentive programs still reward arrival. Deposit, bridge, stake, mint LP, claim. The moment capital appears onchain, the payout logic starts doing its work. That makes the economic signal easy to read. If the reward is flat and front-loaded, the rational strategy is to arrive early, collect as much as possible, and leave when the payout no longer clears the opportunity cost.

That behavior is often described as abuse. We think that frames the problem backwards.

Farmers are not exploiting a bug. They are responding correctly to the incentive as written. If a program prices the deposit, it should expect deposit-seeking behavior. The issue is not that participants learned the rules. The issue is that the rules paid for the wrong event.

Look closely and a flat rate gets four things wrong at once. It pays for the wrong event. It pays whoever shows up, rather than the capital the protocol wanted. It leaves no record of where that capital came from. And it does all of this at a cost per retained dollar that nobody computed, because nobody could. The rest of this piece takes those four in turn, and shows how a Rabbithole campaign is built to get each of them right: targeting, structure, attribution and cost.

Targeting: decide who earns before the money moves

When a campaign disappoints, the usual response is enforcement. Teams reach for sybil detection, clustering, minimum holding periods, manual review, retroactive exclusions, or clawbacks. Some of those tools are useful. None changes the fact that enforcement happens after the program has already advertised an exploitable payout curve.

Enforcement is targeting after the fact, and it is expensive, adversarial and late. The protocol pays people whose capital was never meant to stay, then pays again to identify them once the money has started moving. Participants adapt, rules get more complex, detection gets more expensive, and the budget leaks before anyone has decided who counts as legitimate.

Eligibility is targeting before the fact. A Rabbithole campaign sets who can earn before a single reward accrues: net-new capital only, a balance floor, a cap per wallet, a record of real onchain activity. Balance alone is a weak filter, because a large position can still be opportunistic. Conditions that describe behavior, such as wallet age, activity history and position held over time, align eligibility with the kind of participant the protocol actually wants to keep rather than with a temporary concentration of capital.

None of this needs an adversarial process. It needs a decision, made in advance, about who the program is for.

Structure: reward modes that combine

If the objective is retention, the reward should accrue through time rather than trigger at the point of arrival. In a time-weighted program, the wallet that plans to leave quickly has a weak reason to enter. The wallet that intends to stay can justify the same campaign differently, because its reward grows with the holding period rather than with the initial deposit. That changes the shape of participation before enforcement enters the picture.

Time-weighting is one lever, not the whole toolkit. Rabbithole runs five reward modes that combine. Time-weighted accrual scales with time in position. A cliff holds rewards at zero until a named day and then unlocks them in full, which filters out capital that never intended to stay. A milestone adds a step bonus at a named day, so reaching day 30 or day 60 is worth something. Net-new only pays for balance that is new to the target position, so capital rotated from elsewhere in the same venue does not qualify. Caps put a ceiling on any one wallet and on the total, so one depositor cannot take the budget.

Each mode prices a different behavior, and a campaign is the combination chosen for the outcome, not a rate. A protocol that wants durable liquidity in one pool might run time-weighted accrual with a cliff and a cap. A protocol opening a new market might want net-new only with a milestone. The modes are the same. The structure is chosen for the result, and that is the difference between an incentive and a strategy.

Structure does not require a lock-up. Retention and lock-ups are not the same thing. Capital that cannot leave proves there is a contract. It does not prove there is conviction. Rabbithole keeps custody with the participant and leaves the exit open at all times, because an open exit keeps the signal honest. If someone stays, they are choosing to stay while rewards accrue under the conditions the protocol set.

Attribution: every deposit traced to its source

Most incentive dashboards report a total. TVL went from here to there. The total says nothing about where the capital came from, which campaign brought it, or whether the wallets behind it look like the ones the program was designed for.

Tracing every deposit to its source: by wallet, by region, per campaign. It turns one number into a map. A protocol can see which markets responded, which cohorts stayed, and which part of the budget produced the capital that is still in place a month later. It can compare campaigns on the same basis instead of reading a chart and guessing.

Attribution is also what makes targeting and structure improvable. Without it, a program is a bet placed once. With it, the next campaign starts from evidence.

Cost: what the budget actually bought

A retention program should produce reporting that matches the objective. The most useful view is not peak TVL during emissions. It is what remained after the campaign had time to separate transient capital from committed capital, and what each retained dollar cost. That is the view Rabbithole reports.

The metrics are simple. What share of campaign capital was still in place at 30, 60 and 90 days? What did the protocol spend per retained dollar of TVL? Where did the retained capital come from? And how much of the budget went unspent because caps and cliffs kept it off capital the protocol never wanted? Those numbers say more about whether the design worked than any headline TVL spike ever will.

That last figure matters more than it looks. When the structure is doing its job, budget that would have gone to capital that leaves is not spent at all.

Good incentive design still needs a good offer

None of this rescues a weak product. A retention program on something nobody wants to hold retains nothing. Better structure is not a substitute for product-market fit, yield quality, or treasury discipline.

What it does remove is the structural subsidy paid to the capital least interested in staying. That is the first design problem many programs need to solve, and it is the one Rabbithole was built to solve.

About Rabbithole

Rabbithole is the capital retention layer for onchain markets. A protocol picks the pool, defines the outcome, and decides who qualifies and how rewards are structured. Rabbithole builds and runs the campaign, routes the TVL, traces every deposit back to its source, and returns unclaimed budget. Custody stays with the participant, the exit stays open, and there is no lock-up.

Protocols that want to run a campaign can learn more at rabbithole.gg.

New York, US

Christopher@boost.xyz