NumiaEngage
Marketing · CRM · Wallet Analytics · ClickHouse
Protocols can't talk to their own users. Wallets are pseudonymous addresses. No email, no phone number, no way to send a message. Most Web3 growth strategies default to airdrops: distribute tokens broadly and hope the right people engage. Without targeting, a big share of incentives ends up with farmers and bots. Protocols want precision, but traditional marketing doesn't work when your entire customer base is anonymous.
We already had the data infrastructure indexing wallet activity across the Cosmos ecosystem. At some point we started asking: what if protocols could use that data to actually reach specific wallets, not just run analytics on them?
Getting distribution right
Good targeting is useless without a channel. Email doesn't work here. Social doesn't work either. The only reliable way to reach a wallet user is through the wallet itself.
We partnered with Keplr (the dominant Cosmos wallet) and Leap (multi-chain and growing fast). Campaigns reach users through native notifications and in-app placements inside the wallets people already use every day. No new app installs. No spam.
Getting these partnerships across the line took more effort than writing the targeting code. Wallets are protective of their users, and they should be. We had to prove that our campaigns would be useful, not another notification firehose blasting irrelevant messages. We went through months of negotiation on message frequency limits, content guidelines, and user opt-out controls before either wallet agreed to integrate.
Targeting
Every campaign starts with on-chain data. Protocols can target wallets that swapped, staked, or bridged specific assets. They can filter by token holdings above a threshold, or find users active on competitor protocols. We also built behavioral models that identify power users, dormant wallets, and new entrants, plus lookalike targeting that finds wallets behaving similarly to a protocol's best existing users. Audiences refresh daily as new transactions land, so a campaign targeting "active DeFi users who haven't tried our protocol" doesn't go stale.
dYdX was our launch client. They target traders based on trading volume, position sizes, and activity on competing perpetual exchanges. Attribution tracks back to on-chain activity, so they can see which campaigns drove real conversions. After dYdX, other protocols followed: Quasar and Tower for vault and DeFi strategies, Stride for liquid staking campaigns, Osmosis for DEX liquidity programs. Each one uses the same targeting primitives but builds audiences around their own user profiles.
Real-time notifications
Alongside campaign targeting, we built a separate event-driven notification system. Something happens on-chain (a token transfer, a governance vote, a liquidation) and a message fires in milliseconds. Protocols hook into it via webhooks with structured JSON payloads and automatic retry logic.
This covers the time-sensitive stuff: governance vote reminders, staking reward alerts, liquidation warnings, re-engagement triggers when power users go dormant. Messages that only matter if they arrive on time.
Architecture
Wallet activity from 25+ chains flows through our indexers into ClickHouse, the same cluster that powers our APIs and analytics products. Behavioral models score addresses in the same system and update continuously. Campaign distribution connects to Keplr and Leap through their partner APIs. Attribution happens entirely on-chain.
We only use public blockchain data. No PII collection. Wallet profiles are purely behavioral, never identity-linked. Users can opt out through their wallet settings. Protocols actually prefer this. They've been burned by vendors who over-collect, and they appreciate knowing exactly what data we touch.
Results
dYdX, Osmosis, Stride, Quasar, Tower, and several other protocols across the Cosmos ecosystem run active campaigns through NumiaEngage. Hundreds of campaigns have gone out through Keplr and Leap, each one targeting specific wallet segments rather than blasting generic messages. Multiple chains also use the notification system for governance alerts and staking events. The public-data-only approach, no PII, clear opt-outs, has been easier to sell than we expected. Protocols that were skeptical of on-chain marketing came around once they saw we weren't building identity graphs.
What I learned
On-chain targeting is a completely different animal from web2 marketing. No cookies, no device IDs, no login sessions. You're working with wallet addresses and transaction histories. That sounds limiting, but it forces you to target based on what people do rather than who they claim to be. The signal is noisier in some ways but more honest in others. We threw out most of our assumptions from traditional martech and rebuilt from scratch.
NumiaEngage wouldn't exist without the data platform we'd already built. The ClickHouse cluster, the indexers, the behavioral models, all already running. We layered a new product on top of infrastructure that was proven and paid for. Starting a marketing platform from zero, including the data layer, would have taken years instead of months.