Insight
Can a Tokenized Community Keep People Around After the Rewards Stop?

Run one cycle with rewards and one without. Then measure who still shows up to do something another member values.
Imagine a group that gets 300 submissions during a bounty week. Two weeks later, the prizes are gone and a first-time member asks for help. Nobody answers. The campaign measured whether it could mobilize people for an incentive. The unanswered question tells you what the community still has to build.
There’s evidence that incentives can help with retention. An analysis posted to the Optimism governance forum estimated that receiving 50 OP in its fifth airdrop increased retention by 4.2 percentage points at 30 days and 2.8 points at 60 days. That result applies to the analysis near the study’s eligibility threshold. It measures activity by addresses in that setting and should inform, rather than stand in for, evidence from a different community. Optimism Collective analysis
Give the reward a job it can finish
A bounty can pay someone to do a first useful thing: translate an onboarding guide, host a discussion, or help compile a member resource. Design the next step before paying for the first one. Will the translator meet the members who read the guide? Will the discussion host be invited back by people who enjoyed it? Will someone use the resource and ask the contributor a follow-up question?
For Normie, one experiment could reward members for proposing conversation prompts in week one. In week two, run the same format without rewards and invite people back to vote, explain their choices, and respond to someone who chose differently. Track which prompts generate a real exchange between members, and which members return to start discussions on their own. Compare with an unrewarded group if the audience is large enough, and keep notes on how each group was recruited. Normie project at Martini Labs
The two-cycle experiment, week by week
The Normie setup generalizes to almost any community program. Keep the format identical across both cycles, so the reward is the only thing that changes:
| Week | What happens | What you measure |
|---|---|---|
| Week 1 — the rewarded cycle | Pay for one defined first action: a prompt, a translation, a hosted session, a resource. Announce the end date up front. | Who acted, what the work was, and the total spend. |
| Week 2 — the unrewarded cycle | Run the same format with no prizes. Personally invite week one’s contributors back. | Who returned, and whether members talked to each other or only to the team. |
| Week 3 — the quiet week | The team posts nothing and arranges nothing. The room runs itself, or it doesn’t. | Unprompted actions, questions answered by members, and reuse of week one’s work. |
The quiet week is the part most programs skip, and it is the part that tells the truth. A community that only moves when the team pushes it is a distribution channel with better branding.
Design the next step: a ladder of roles
Retention gets easier when coming back has a shape. A simple ladder gives each contributor a next step that is a role, not another bounty:
- Contributor. Does a first useful thing — the action the reward paid for. The program’s job is to make sure somebody notices it and uses it.
- Host. Runs a session other members attend: a discussion, a working hour, a review of new resources. Hosts return because people expect them, and being expected is a stronger pull than being paid.
- Steward. Looks after something specific over time — a guide that needs updating, a channel that needs a greeter, a resource list that needs pruning. Stewards hold the memory of the community, and they are the people who answer the first-time member’s question.
Notice what changes at each rung. The contributor is paid in tokens. The host is paid in an audience. The steward is paid in responsibility and standing. Only the first of those stops working when the budget does.
What not to reward
Rewards teach people what a community values, so pay for the wrong thing and you will get more of it. Three habits cause most of the damage. Don’t reward raw volume — a bounty per post buys posts nobody reads. Don’t reward attendance by itself — showing up is the floor, not the contribution. And don’t keep paying the same people for the same first action; that trains members to wait for a bounty before doing anything at all. The test for any reward is simple: if this payment worked perfectly, what behavior would the community have more of — and would members still value it unpaid?
Collect evidence that people have a reason to return
- Unrewarded return rate: Of the people who took a first meaningful action, what share did it again in an unrewarded cycle?
- Member-to-member help: How often did a member answer a question, give feedback, or host an event without the team arranging it?
- Contribution reuse: Who used the guides, playlists, templates, or other work the rewards paid for?
- Cost per returning contributor: How much did the program spend for each participant who repeated a useful action?
Read the four metrics together
Each metric can mislead on its own; together they describe what you actually built. A strong return rate with weak member-to-member help means people come back for you — your content, your team, your announcements. That is an audience, and it leaves when you stop publishing. Strong contribution reuse with a weak return rate means the work outlived the workers: the resources are valuable, but the contributors found no reason to stay, so recruit the next cohort around the resources themselves. A low cost per returning contributor alongside thin help usually means the program is efficient at producing activity and poor at producing relationships. And if all four come back weak, the bounty bought a spike, not a community — spend the next budget on a better first action and a clearer next step, not on a bigger prize.
Points, payouts, reputation, and voting rights can each serve different purposes. In a Decentralized Engagement Organization, specify which behavior the incentive is meant to start, what useful work it produces, and what happens when the incentive ends.
A bounty can bring good people into the room. Give them a reason to come back when nobody is paying them to.
Questions founders ask
Does this mean rewards don’t work? No. The Optimism analysis above is evidence they can move retention, within its setting. The argument here is narrower: a reward is a starting mechanism, and you only learn what it started by running a cycle without it.
How large does a community need to be to run this? Large enough to compare groups is ideal — the Normie experiment above notes a comparison group if the audience allows it. If yours doesn’t, run the cycles in sequence instead, keep the format identical, and write down how each group was recruited, so you know what else changed between them.
What if nobody returns in the unrewarded cycle? Then the experiment worked — it told you the first action didn’t connect members to each other, before you spent more money finding out. Redesign the next step, usually the invitation into a role, before you raise the reward.
The short version
Rewards are good at starting behavior and bad at explaining it. Give every bounty a job it can finish, design the next step before you pay for the first one, and run one cycle with the money off. Then read the four metrics together: who returned unpaid, who helped another member, whose work got reused, and what each returning contributor cost. What survives the unrewarded cycle is your community. Everything else was a campaign.
Call to action: Add one unrewarded cycle to your next engagement experiment and decide what a meaningful return looks like before it begins. Use the DEO Builder Guide to map roles, tasks, and incentives. The guide is the direct follow-up to this article.
Internal reading: How Normie applies the community model

Weekly · Free
The Martini Shot
One shot of AI, crypto and future-tech signal every week — what matters, why it matters and what MAX thinks.
No spam. Unsubscribe any time.



