Chapter 28 · The Nature of Crypto Growth

A Real-World Problem

A team reviews quarterly growth. A beautiful curve fills the projector screen.

Week 1: 2,000 daily active addresses, stable. Week 3: an incentive campaign is announced; daily activity reaches 12,000. Week 6: tokens will be awarded for historical interactions; daily activity reaches 80,000 and volume rises thirtyfold. Week 9: snapshot day. A record 110,000 active addresses, with screenshots everywhere in the community. Week 10: the snapshot is over. Daily activity: 1,500.

Lower than before the campaign.

Three explanations emerge, each with supporters:

• “The market is weak; it is the broader environment.”
• “The airdrop was too small to retain people.”
• “The product itself needs rebuilding.”

They share one problem: none is verifiable, and none leads to a concrete next step.

Yet this can be measured. Every transaction, every address's first appearance, and every interaction over those ten weeks is publicly recorded on-chain. Chapter 23 taught you how to obtain this data.

The real question is not why activity fell, but two more specific questions:

First, how many of those 110,000 came to do something they already wanted to do? Second, how much did the incentives cost, and what did they buy?

The second has a precise answer, and line four of the income statement in Chapter 26 is its numerator.

This chapter turns those questions into repeatable procedures.

Thought Experiment

A gym opens in a new neighborhood.

Month one: the owner gives out 5,000 passes for three free months. Lines extend into the street, reception is overwhelmed, and social feeds fill with check-ins. The gym receives 900 visits a day.

Month four: the free period ends. 120 people renew. Daily visits: 140.

Three months of rent, equipment depreciation, and trainer salaries bought 120 members. Total cost divided by 120 is the acquisition cost of one real member. If that exceeds the membership fees one person contributes over three years, the campaign lost money—however lively the photos looked.

Add three details, each corresponding to a real on-chain phenomenon.

Layer one: people come solely for the gift. Registration includes a towel. Someone uses friends' and relatives' IDs to obtain 300 passes and 300 towels, never visiting once. They are not a fraudster; they found a ten-minute opportunity yielding three hundred towels.

Layer two: change the rule to sixty workouts for a gift. The owner expects this to identify enthusiasts. Instead, a new occupation appears: hired students swipe in daily, sit for ten minutes, and leave. Completing the count costs less than the prize, so it will happen.

Layer three: the owner says only that there will be future rewards. Short-term results are impressive: an unknown threshold encourages more visits, and an unknown prize encourages imagining the best outcome.

But a delayed cost follows: on announcement day, most people will be disappointed. Everyone sets their own expectations; reality offers only one result. Afterward, they may stop coming and tell others the gym is not worth visiting.

The beautiful curve is no longer mysterious. It measures the appeal of rewards, not the quality of the gym.

The question that matters throughout is: how many people return voluntarily after the free period ends?

Your Decision

You have a growth budget and four ways to spend it.

What Happens

The four options converge on the chapter's central statement:

Incentives buy first use; retention must come from the product itself.

Split growth into five layers to see where each tool acts:

The final row is the table's center of gravity. Money can buy the first four directly, but cannot buy the fifth. A growth plan's value depends on the proportion of arrivals it retains.

This explains why successful growth so often precedes disappearing users: the first four layers were maximized while the fifth remained zero.

One clarification is essential: this chapter does not oppose incentives.

They are necessary and irreplaceable for bootstrapping (a two-sided market with neither side cannot start itself; someone must fund the first side) and ownership distribution (giving tokens to real users rather than only investors and the team).

The question is never simply whether to distribute rewards, but whether people remain afterward, and whether you measure it.

Build a Model

Split Users into Four Layers

The four-layer distinction from Chapter 23 becomes a standard procedure:

Only all four together answer who the users are. Reporting only layer one effectively reports nothing.

Two Numbers You Must Calculate

First: real acquisition cost.

Cost per retained user = Total incentives valued at distribution-time prices
                         ÷ Addresses still active 90 days after the campaign

Use layer four rather than layer one as the denominator. Results often differ by one or two orders of magnitude, and only the former can be compared with lifetime user value.

Second: the retention curve's shape.

Anchor day zero at the campaign's end and plot activity for the following 90 days. There are three shapes:

The stable level in the second pattern is the campaign's real output. Subtract the pre-campaign baseline to obtain net growth.

The Arithmetic of Incentive Farming

Use the designer's perspective, or you will never prevent exploitation.

The rule is simple: positive expected reward minus cost means the behavior will happen and scale.

Expected reward = Expected prize value × Probability of receiving it
Cost            = Fees + capital tied up + time + address and identity preparation

Designers can change only three variables:

Remember: any clearly describable threshold can be satisfied at scale. Effective design therefore makes scaling uneconomic, rather than simply making thresholds harder—for example, a per-address cap so a thousand addresses do not earn a thousand times as much, or long-term funds that make capital costs grow with address count.

Also, these participants are not enemies. They rationally follow your rules. If your rules reward manufactured volume, that is the outcome you asked for. You get what your rules specify, another formulation of incentive design in Chapter 20.

Verify a Large Community with On-Chain Data

The question left by Chapter 22 now has a procedure. Any claimed community of tens of thousands can be tested with four queries:

These queries are straightforward and together take less than an hour. They turn an unverifiable slogan into four citable numbers.

What It Is Called

Hands-on

Day zero:       A specific date, UTC
Window:         30 days before and 90 days after
Addresses:      Main contract 0x… verified in an explorer; mainnet only
Events:         … specific event names
Deduplication:  By initiating address; aggregator users count under the aggregator
                (include this in uncertainties)

select
  date_trunc('day', block_time) as day,
  count(distinct tx_from)       as active_addresses,
  count(*)                      as tx_count,
  sum(amount_usd)               as volume_usd
from <platform event table>
where contract_address = <verified contract address>
  and block_time between <day zero - 30 days> and <day zero + 90 days>
group by 1
order by 1

Pre-campaign baseline = Average daily active addresses in the preceding 30 days
Campaign peak         = Highest daily activity in the window
Stable level          = Average daily active addresses on days 61–90 afterward

Net growth = Stable level − Pre-campaign baseline

-- Layer one: addresses interacting during the campaign
-- Layer two: among them, first transactions predating the campaign (not created for it)
-- Layer three: among them, at least one interaction unrelated to incentives
-- Layer four: among them, interactions continuing 30 days after day zero

select
  date_trunc('week', first_tx_time) as cohort_week,
  count(*)                          as addresses
from <your address first-transaction-time table>
group by 1
order by 1

Layer-one basis: Total incentives ÷ Active addresses during the campaign = ______
Layer-four basis: Total incentives ÷ Addresses active 90 days after zero = ______

Real net growth from this campaign:   ______ daily active addresses
Real cost per retained user:          ______
Uncertainties, at least three:        1. …  2. …  3. …

AI Lab

First, design an eight-week incentive campaign for this protocol:
<What it does, current users, and the behavior to grow>

Include:
1. Target behavior and why it matters
2. Thresholds and weights for individual contributions
3. Distribution form and schedule
4. Total budget and valuation method
5. Post-campaign success metrics

Second, add a section titled “How this plan will be exploited.”
From an incentive-farming team's perspective, give at least five concrete methods:
- Exact actions
- Approximate cost per address
- Difficulty of scaling to one hundred addresses
- Whether the plan blocks it

Third, explain why these people will return after the campaign.
If they will not, say so directly.

Rules:
- Avoid vague instructions such as “strengthen anti-Sybil measures.”
- Explain on-chain execution for every defense.

Real-World Case

Change One Variable

Takeaway Questions

Chapter Check

Cost per retained user = Total incentives ÷ Addresses active 90 days after campaign end

Expected reward = Expected prize value × Probability of receiving it
Cost            = Fees + capital tied up + time + address preparation