a16z: How to identify real users from 220 million active addresses?

Author: Daren MatsuokaEddy Lazzarin

Translation: Blockchain in Vernacular

As part of our Crypto 2024 report, the team spent a lot of time assessing the current state of the crypto industry. As the industry matures and more and more applications come online, we want to understand how many people are actually using cryptocurrencies. This is a complex question because the most obvious and easily quantifiable usage metric — active addresses — can be easily manipulated. Below we share some insights.

In the traditional software world, the concept of “user” is widely understood. Of course, there are many ways to measure user quality. In fact, there is an entire field of user growth analytics devoted to this topic. But at the most basic level, users can be categorized as "Daily active users" (DAU), "monthly active users" (MAU), etc.

In the crypto space, the situation is even more complicated. This is because on the blockchain, user identities are pseudo-anonymous. One person can easily create and control a so-called "Sybil attack" (a group of different identities, called " Public key addresses") operate on the blockchain. (There are many legitimate reasons for this, such as for privacy, security, or other purposes.) Therefore, it is difficult to know how many addresses a person actually uses. (Conversely, , multiple people can also use the same address through multi-signature, master account, and various account abstraction protocols.)

Until recently, the most popular blockchains had very limited capacity, resulting in high transaction fees. This created a natural barrier that prevented people from creating and using hundreds or even thousands of addresses in bulk, as it would cost a lot of money. But as crypto infrastructure has become more scalable — via L2 Rollups and emerging high-throughput L1s — transaction costs on many blockchains have dropped dramatically, approaching zero.

But isn’t the cost of creating multiple identities close to zero in traditional Internet applications? This is largely true. For example, a person can easily create and use multiple email addresses. But the key difference is that there is a strong Motivations drive people to do this.

The crypto industry has long rewarded early adopters by issuing tokens. Today, new protocols often launch their token circulation supply through “airdrops” - activities that provide token rewards to a pre-determined set of addresses. Usually, these addresses list It is generated based on historical on-chain transaction records. Some people may try to manipulate the system by creating multiple identities and conducting transactions. In the industry, this strategy is called "airdrop farming."

Given these behaviors, it is clear to us that the 220 million monthly active addresses we measured in September 2024 do not equate to 220 million people or users. (Note that addresses active on multiple EVM chains are only counted once toward this count.) 220 million in total.)

So, how many active users are there? 10 million? 50 million? 100 million? This is the question we are trying to answer. Next is our methodology.

1. Method 1: Filter active addresses

One approach we took was to filter out addresses that were likely controlled by bots or were part of a Sybil attack. There are multiple ways to do this through on-chain analysis and forensics, here are some of the ones we explored:

1) Filter out addresses that received funds from the distribution contract - the only purpose of the distribution contract is to receive funds and automatically distribute them to multiple different addresses. Although there may be some false positives, this activity indicates that these target addresses All come from the same source and are therefore related to each other in some way.

2) Filter out addresses with a balance close to zero during a certain period of time. For example, if you are looking for actual monthly active users in September 2024, you can try to exclude addresses with a balance close to zero on both September 1 and September 30. This criterion indicates that these addresses are temporary. While bots and Sybil attackers may clean up their balances after performing operations, real users usually keep some balance in their wallets to pay for future transaction fees.

3) Analyze the distribution of addresses that made one, two, three, four, five or more transactions in a specific time period. Addresses that made only one or two transactions are low-quality users at best and robots at worst. or Sybil attacker. This approach works best when data is aggregated over a long period of time.

4) Filter out addresses that conduct a large number of transactions in a short period of time. When humans use wallets or application interfaces, they can only reasonably process a certain number of transactions in a given time, while robots can conduct more frequent transactions in a shorter period of time. trade.

5) Selectively include addresses tied to identity protocols that usually require some setup cost. For example, addresses with ENS names, Farcaster IDs, and other social identity links are likely to be actual users.

These are just some of the patterns on the chain that may indicate bot behavior and are by no means exhaustive. We welcome suggestions for improvements based on this.

2. Method 2: Infer from wallet users

Another way to estimate monthly active users is to look at off-chain data sources. The most obvious place to look is at wallet users.

In February 2024, popular crypto wallet MetaMask reported 30 million monthly active users. They define a monthly active user as “a user who loads the MetaMask extension or opens the mobile app at least once in any consecutive 30-day period.”

Assuming we want to estimate the users who actually transact, the next step is to determine how many of MetaMask’s users actually transact. In 2019, MetaMask reported that on a given day, about 30% of active users confirmed an on-chain transaction. (This is the latest available estimate.) If we apply this ratio to monthly active users (MAU), then there are approximately 9 million users transacting monthly through the MetaMask wallet product.

Next, we need to understand MetaMask’s total wallet market share across all blockchains. While this exact data is not easy to obtain, we can make some reasonable guesses based on the existing information. For example, we can use the data from mobile analytics companies to Sensor Tower data estimates MetaMask’s share of the mobile wallet market. (Due to commercial service agreements, we are unable to disclose specific numbers here.)

Once we estimate MetaMask’s market share, we can simply extrapolate the total number of cryptocurrency users from the 9 million monthly active trading users we derived earlier. We can then compare this result with the result from method 1. and see if they are in the same range.

We can also further refine our estimates by analyzing proprietary data that other wallets and infrastructure providers are willing to share and cross-validate it with the data derived above.

3. Other considerations

We also need to consider that some people use multiple addresses and wallets to transact. While this will not significantly inflate the number of users like bots and Sybil attacks do (after all, there is a limit to the number of wallets a person can reasonably use), based on some Under reasonable assumptions, further deduplication still makes sense.

On the other hand, there are also cases where one address may be associated with multiple real users. The collective account of a trading platform is an example. Moreover, with the popularization of account abstraction protocols and smart contract wallets, the situation will become more complicated. These considerations are not included.

Final estimate: 30 to 60 million real monthly active trading users

Based on our analysis using the various methods described above, we estimate that there are currently around 30-60 million real monthly active crypto users. Obviously, this range is quite wide, but it is the best estimate based on the available data.

It’s worth noting that this only represents 14% to 27% of the 220 million monthly active addresses we measured in September 2024. It also only accounts for 5% of the 617 million global crypto holders reported by Crypto.com in June. (Global crypto holders are people who own cryptocurrencies but do not necessarily trade them on-chain.) This gap suggests that it is necessary to convert existing crypto holders, most of whom are passive holders, into The opportunity to convert to active users is huge. As major infrastructure improvements enable new and compelling applications and consumer experiences, currently dormant crypto holders are expected to re-emerge as active users on the chain.