Why on-chain behavioral segmentation is the missing layer in DeFi growth
Photo by Khanh Nguyen
DeFi teams have tried countless incentive programs – from liquidity mining on Solana to cross-chain quest campaigns on Ethereum Layer-2s – and the story repeats. At launch, the spreadsheet metrics look fantastic: daily active wallets spike, total value locked rockets, and crypto Twitter applauds the “growth hack.” But thirty days later, the chart resembles a ski slope as activity plummets. Most of those wallets were mercenaries farming airdrops and then vanishing.
Nansen tracked over 6.5 million unique DeFi addresses by the end of 2022 (its a lot more today - https://dune.com/rchen8/defi-users-over-time?), yet meaningful retention remains scarce because we treat every wallet as equal. In fact, a recent analysis found that 80% of new blockchain users disappear within 90 days, and only a small fraction of addresses contribute sustained activity. Web3 growth teams often chase raw signup numbers with airdrops or rebates, only to see most “users” never stick around – leading to unsustainable user acquisition costs and wasted incentives.
The raw material for a better approach has always been right in front of us: on-chain data. Every swap, bridge, liquidity provision, or DAO vote is a self-disclosed preference – a line of first-party data recorded permanently on-chain. Chainalysis refers to these on-chain behavioral signals as “behavioral clusters,” grouping wallets by observed patterns.
Mainstream exchanges already pipe such intelligence into their marketing funnels with impressive precision. By leveraging blockchain’s transparency, exchanges can gather real-time data on users’ holdings and transaction habits, then target offers to those most likely to engage.
Growth teams in Web3 should be doing the same, but most still rely on web2-era guesswork – web cookie data, Telegram poll feedback, or vanity metrics like gross TVL – instead of tapping into rich on-chain user insights.
Three on-chain user archetypes that matter
There are millions of on-chain events across Ethereum, Solana, and other networks with clear patterns - not all wallets are equal. In practice, wallets tend to fall into three broad cohorts:
Airdrop hunters: These wallets leave a breadcrumb trail of tiny, sporadic transactions across many protocols, often using multiple fresh addresses with negligible ending balances. Their sole motive is to farm token rewards. They spin up numerous new accounts to mimic real usage and qualify for expected airdrops, then immediately dump the tokens and abandon the wallets. Hunters can inflate your user count in the short term, but they provide virtually no long-term value – a classic case of “tourists” in your dApp.
Power users: These are the die-hards and yield optimizers. A power user might execute double-digit swaps per month, bridge >$5,000 in assets across chains, and actively provide liquidity in multiple pools. They chase yield and efficiency relentlessly, often becoming evangelists if treated well. Data shows that these high-activity addresses have far superior retention: in one study, high-value users declined only ~5–8% per month, retaining 35–38% of users after 6 months on strong defi networks like Ethereum. Power users are the lifeblood of protocols – the wallets contributing repeat volume and governance participation – and they respond to improvements in throughput, fees, and features.
Learners (newcomers): These are new arrivals to DeFi – a wallet with only one or two small trades and a single-chain footprint so far. Learners represent the widest funnel of potential upside (everyone starts as a beginner), but they churn quickly if the first experience disappoints. In fact, wallets with minimal prior activity (“low-value” users) show <5% retention after 6 month. A newcomer who struggles with a clunky UI or loses money on a first trade may never return. This cohort needs hand-holding and education to convert into long-term users.
Understanding these archetypes is not just theoretical. In our growth experiments, simply filtering for “power users” via a blunt SQL query (e.g. wallets with >10 swaps, >$500 volume, on >2 chains) identified a cohort that converted at 3× the baseline rate in the next campaign. And today, services like ChainAware use raw on-chain wallet activity to automatically label users as traders, lenders, gamers, NFT collectors, etc., with no extra instrumentation In short, the data to differentiate between a one-off airdrop farmer and a genuine DeFi enthusiast is all publicly available on-chain – we just need to use it.
Building the segmentation pipeline (pipes, not patch-jobs)
Turning this theory into a working growth engine requires a few key steps:
Extraction & cleaning: First, pull the raw events. Tools like Dune Analytics or the Nansen API let us query on-chain transactions (swaps, liquidity adds, votes, etc.) to build a history for each address. From there, we apply heuristics to stitch together addresses that likely belong to the same user (and to filter out obvious Sybils). This data cleansing is crucial – LayerZero famously identified 800,000+ Sybil addresses before its anticipated airdrop, aggressively weeding out farmer wallets. That upfront effort to purge fake or duplicate users pays for itself by preventing dilution of your campaigns.
Audience sssembly: Next, translate those insights into targetable segments. For example, you might compile a list of addresses that meet “power user” criteria (say, >10 DeFi transactions in the last month and active on multiple chains) or “learner” criteria (new wallets with <$100 in DeFi activity). These wallet lists can then be uploaded to Web3 marketing platforms – essentially crypto DSPs. Platforms like Blockchain-Ads allow wallet-based targeting, even enabling look-alike modeling where the system finds other wallets with similar on-chain behavior to expand your reach. In other words, you can target “people who behave like my current best users” by using their wallet activity as the matching signal.
Activation & measurement: Finally, deploy tailored campaigns to each segment and measure on-chain outcomes. Instead of generic Twitter bounties or Discord invites, use quest platforms (e.g. Galxe or QuestN) to issue on-chain challenges and track real completion. For instance, you might create a quest for Learners that guides them through a first swap or deposit with a small reward – and because completion is recorded on-chain, you know they actually used the dApp (not just clicked a link). For Hunters, you could require multi-step or time-delayed tasks to deter hit-and-run farming. The key is that success is measured in on-chain actions (wallet did X or increased balance by Y), not just ad clicks or form sign-ups.
Throughout this pipeline, the mantra is quality over quantity. Rather than blasting incentives to every wallet in existence, you’re building filters to catch the whales and genuine fish while letting the plankton (bots and Sybils) slip through.
Hooks that resonate with each segment
Once segmentation is live, growth tactics can be far more precise – the hooks that resonate will differ by cohort:
Converting the hunters: Airdrop farmers will always try to game the system, but you can design incentives that separate the wheat from the chaff. One approach is tiered quests or rewards: offer basic rewards that are easy to get (attracts everyone), but unlock premium incentives only after certain time or balance criteria are met. For example, a liquidity mining program might start with a small base APR, then gradually increase rewards for liquidity that stays in the pool for 30+ days or grows above a threshold. This throttling mechanism filters out the “farm-and-dump” behavior by making quick flippers ineligible for the biggest prizes. Hunters will either adapt by behaving more like real users or move on. In short, make them earn it.
Empowering the power users: Your most active users crave throughput, influence, and status. To keep these power users engaged (and increase their loyalty), reward depth over breadth. This could mean releasing a “Pro” interface with advanced analytics or API access, higher trading limits, or faster execution specifically for high-volume traders. It could mean deeper liquidity pools or better yield tiers for those who contribute large amounts of liquidity. It could also mean governance perks – e.g. giving top users extra say in proposals or early access to new features. These users are already bought in; showing that you recognize and value them will turn them into evangelists. A great example is Polygon’s 2024 strategy focusing on “super-users.” By doubling down on heavy hitters – 1.5 million addresses responsible for 867.7 million transactions – Polygon grew network usage without bleeding cash on blanket incentives. They identified who was driving the majority of activity and tailored programs to boost that further, rather than chasing new but transient users. The lesson: incentives work best when they reward depth, not just breadth.
Nurturing the new learners: Newcomers have the interest but not the know-how. To turn a first-time user into a second- and third-time user, you must remove cognitive overhead. Streamline the onboarding: integrate fiat on-ramps so they can go from dollars to DeFi in one step, provide one-click “starter” vaults or index funds so they can earn yield without complex strategizing, and offer narrative-led tutorials (think short TikTok explainer videos or Farcaster threads) that hand-hold a user through their first swap or lend. The goal is to make that first successful transaction as easy and positive as possible. If a Learner’s first DeFi interaction is rewarding (literally or figuratively), their chances of sticking around go way up. Some networks have even created newbie-focused incentive programs – e.g. testnet “academy” quests – to safely teach users before they risk real funds. Give learners a longer runway (small bonuses, extra support), and you’ll significantly improve conversion from tourist to regular.
Metrics that beat DAUs
Traditional Web3 metrics like “Daily Active Wallets (DAW)” or raw sign-up counts are increasingly counterproductive – they inflate vanity numbers while masking churn. Instead, growth teams are shifting to metrics that factor in quality and retention. Here are a few we track:
Qualified active users: Rather than counting every wallet that connects once as an active user, define what a meaningful user is for your protocol and count only those. This might be “wallets that perform our core revenue-generating action at least N times in a month.” For a DEX, for example, you might count a “qualified” user as one who makes at least 3 swaps in a month (as opposed to someone who showed up for one trade). This sets a baseline of engagement to distinguish true adopters from drive-by users. It’s a more sober metric, but far more aligned with actual growth.
Segmented retention curves: Instead of one aggregate retention curve, break it down by segment (hunters vs. power vs. learners, or however you bucket users). This reveals the vastly different trajectories hidden within an average. In practice, you’ll see hunters falling off a cliff after the first week or two, learners plateauing at some modest retention if you’re lucky, and power users forming a stair-step upwards (as they keep coming back and even increasing usage over time). Tracking retention by cohort lets you validate whether your campaigns are actually attracting the right users. For instance, if a particular Twitter campaign brought in mostly low-retention hunters, you’ll see that in the 30-day drop-off. You can then adjust spend toward channels that yield stickier users.
6-month retention curves for low-value (purple), medium-value (blue), and high-value (orange) user cohorts across multiple blockchains. Low-engagement users drop off almost entirely within 3 months, whereas power users decline much more slowly.
Adjusted CAC (cost to acquire a retained customer): If you’re calculating customer acquisition cost as simply marketing spend divided by new sign-ups, you’re likely understating the true cost. We calculate an adjusted CAC = (Spend) / (number of qualified conversions). In other words, the denominator is not “wallets that connected once,” but “wallets that became active users as defined above.” This can be further segmented (CAC to acquire a power user vs a casual). It’s a sobering metric at first – you realize just how expensive each sticky user really is – but it focuses the team on efficiency. One Web3 growth agency, Flight3, found in 2025 that by refocusing spend on channels that attracted proven high-quality users (for example, in-depth Twitter threads targeting knowledgeable DeFi participants), they cut CAC by ~28%. In essence, $1 spent on educated, right-fit users went much further than $1 spent blanketing the entire crypto audience. This kind of insight is only possible when you measure success beyond the top-of-funnel vanity stats.
Note to growth teams
Stop counting tourists. A wallet that signs a single transaction should not weigh the same in your metrics as one that completes ten trades, provides liquidity, and votes in governance. On-chain behavior is the only passport that matters. Segment first, reward second, and instrument the entire loop to learn. Yes, acquisition (getting new wallets in the door) fills the funnel – but retention is what compounds protocol value, and the blueprints for both are already etched into the ledger for anyone willing to query the data.
As the recent Flipside research highlighted, “only a handful of addresses are contributing any sustained activity or liquidity volume” in most ecosystems. Growth in DeFi, therefore, is less a pure numbers game and more a character study unfolding on-chain. If you give whales more depth, grant minnows more runway to learn, and tighten the net around airdrop anglers, you’ll be leveraging the natural segmentation that already exists. DeFi doesn’t need more users who won’t stick – it needs to better serve the users who will. Embrace on-chain behavioral segmentation as the missing layer, and you’ll find that sustainable growth was hiding in plain sight all along.
Sources:
Nansen Analytics: nansen.ai
Chainalysis & Exchange Segmentation: coinpasar.sgnasdaq.com
CoinDesk (Airdrop Hunters): coindesk.comcoindesk.com
Flipside Crypto Retention Study: science.flipsidecrypto.xyzscience.flipsidecrypto.xyzcrypto.news
Cointelegraph (LayerZero Sybil data): cointelegraph.com
Coinspeaker (Polygon super-user stats): coinspeaker.com
ChainAware Web3 Marketing Guide: chainaware.ai
