The primary goal of this dashboard is to introduce a methodology for classifying users within an ecosystem based on their activities and engagement from inception to the present. After introducing our method, we apply it to the Avalanche blockchain, focusing on users who have executed at least one transaction over the past three months; however, this classification framework is universally adaptable to any blockchain ecosystem, reflecting its versatile analytical scope. The temporal scope of our analysis spans a 90-day window along the time axis, providing a robust foundation for evaluating recent user engagement. A pivotal consideration in this analysis is the historical activity of users, as assigning scores without accounting for their longitudinal participation would undermine the equity of the assessment. To elucidate this principle, consider the following illustrative case: a user named Alice deposited $1 million five months ago and withdrew the funds three months ago, followed by a $10 deposit two months ago and a withdrawal one month ago. In contrast, a user named Bob deposited $10 two months ago and withdrew it one month ago. If we confine our evaluation to the last three months, disregarding their historical interactions, and assuming no additional activities beyond those specified, both users would receive identical scores. Such an outcome is manifestly unjust, as it fails to recognize Alice’s extended engagement with the ecosystem. Consequently, the foundational step in our methodology is to integrate users’ historical activities, ensuring a balanced and fair scoring system. Although our analytical window is limited to the past three months, we incorporate historical data to enrich our insights. This historical perspective is crucial not only for fairness but also because it serves as a proxy for user loyalty, a critical indicator of sustained ecosystem commitment.

Another essential dimension of our analysis is the diversity of user engagement across various protocols within the Avalanche ecosystem, encompassing a wide array of activities—to name but a few, liquidity provision, staking as delegators or validators, participation in NFT trading, and simple transfers. Each of these activities exerts a distinct influence on the ecosystem’s vitality and resilience, with certain actions—such as staking—proving indispensable for the robustness of a proof-of-stake framework, while others contribute meaningfully yet to a lesser degree. This dashboard undertakes a granular analysis of each activity category, assessing its specific contribution to the ecosystem’s flourishing and long-term sustainability. While our empirical focus is on the Avalanche blockchain, the methodological approach is designed to be extensible to other blockchain networks.

Activities that involve users locking their assets in diverse areas to accrue profit through varying interest rates are particularly vital for ecosystem stability. Within a proof-of-stake context, the most significant asset-locking activity is staking, whether performed by validators or delegators, offering returns based on network participation. Other forms of asset immobilization include, but are not limited to, deposits in lending-borrowing protocols, where users lock funds to earn interest, and liquidity provision to facilitate token swaps, which also generates profit through interest or fees. These examples highlight key mechanisms, though additional unlisted forms of asset locking may exist within the ecosystem. Nonetheless, staking by validators and delegators remains the preeminent activity due to its critical role in network security and rewards. Following a detailed examination of each category, we provide a comprehensive exposition of how we determine the relative significance of each area in the user scoring process.

A further analytical challenge arises when evaluating asset-locking behaviors: focusing exclusively on the staked amount overlooks the duration of the lock, while prioritizing duration alone neglects the amount involved, both of which are essential metrics. To address this, we propose a composite metric that integrates both duration and amount. Hence, we define the following metric:

W = duration × amount   (1)

This metric, denoted as W, is uniformly applied across all instances where users lock their assets within diverse ecosystem sectors. In mathematical terms, W represents the integral of the asset value over time, effectively capturing the area under the value-time graph. The proposed scoring metric, W = duration × amount, is conceptually derived from the power formula in physics, which quantifies energy per unit time (P = W/t). In our context, the balance of locked assets is analogized to power, while W is interpreted as the cumulative work or energy expended by a user within the ecosystem, reflecting their sustained contribution over time. Here is the example:

Example 1:
  Data Points (Step Chart)
- Day 1: $5 (initial deposit of $5)
- Day 2: $10 (adds $5, total $10)
- Day 3: $20 (adds $10, total $20)
- Day 4: $5 (withdraws $15, total $5)
- Day 5: $5 (no change, remains $5)
- Day 6: $0 (withdraws $5, total $0)

To augment our analytical framework, we introduce a scoring system that assigns users a score ranging from 0 to 15 based on their W metric. In the domain of user activity scoring across platforms, it is conventional to employ an increasing, concave function, wherein the attainment of higher scores becomes progressively more arduous as the score escalates. We propose the following scoring function:

Score = 15 (W ⁄ W_max)^k   (2)

where 0 < k < 1, and W_max denotes the maximum W value attained by any user to date. The coefficient 15 is selected to ensure that a user’s score remains within the range of 0 to 15. The parameter k modulates the difficulty of score improvement, with its value influencing the function’s behavior: when k approaches 1, the scoring function approximates a linear model, implying uniform difficulty in score increments across all levels; conversely, as k nears 0, score increments at lower levels become notably easier, while the challenge of advancing scores intensifies as the score increases. In our analysis, we adopt k = 0.5, though this parameter may be tailored to suit the specific dynamics of each ecosystem or platform, thereby enhancing the adaptability of our scoring model across diverse blockchain environments. To provide a visual representation of the scoring function’s behavior, we depict this chart with k = 0.5, enabling readers to observe its characteristic curvature.

Having established our methodological framework, we proceed to apply it to a real-world scenario, focusing initially on users who have staked their assets as validators or delegators. For each user, we compute two distinct metrics: W_history, representing the cumulative W value accrued prior to the past 90 days, and W_current, reflecting the W value accumulated within the most recent 90-day period. Subsequently, we derive two corresponding scores based on these metrics using the scoring function: Score_history and Score_current. We need to determine W_max, the reference maximum W value required for score calculation. For each part, history and current, we use the W_max specific to that period. This approach ensures that all users’ Score_history and Score_current fall within the range of 0 to 15. Now we can apply a convex combination of Score_history and Score_current to compute a final Score_skating, weighted according to their respective influences. For calculating the final score for a user who stakes money as a validator or delegator, we use the following convex linear combination:

Score _staking = α _history * Score _history + α _current * Score _current   (3)

where α_i ≥ 0
and α _history + α _current = 1

We elucidate this methodology with illustrative example.

Example 2:
  Suppose Alice has acquired a Score_history of 10 and a Score_current of 7.
  To compute Alice’s Score_staking, we use equation (3), so we have:

Score _staking = α _history × 10 + α _current × 7

α _history and α _current represent the respective weights assigned to Score _history and Score _current. The coefficients α _history and α _current are adjustable, depending on the relative importance of a user’s historical and current interactions within the ecosystem; by tuning these coefficients, the ecosystem can prioritize the aspects of user interactions deemed most significant for its development.

We now establish an overarching framework for the scoring system, extending its application to categorize Avalanche users based on their interactions within the ecosystem. Initially, we determine the W values—and subsequently the Score_history and Score_current—for users staking AVAX, laying the foundation for a comprehensive evaluation of their contributions. For Score_staking, we establish α_history= 0.3 and, correspondingly, α_current = 0.7. Additionally, we set k = 0.5. However, these parameters, along with hyper-parameters including the scoring function itself, are adjustable and contingent upon the priorities of a given blockchain, allowing for tailored application. It is evident that altering these parameters will correspondingly modify the classification outcomes. Given that our scoring system operates on a continuous scale, we round all scores to the nearest upper integer to ensure that any user engaging in at least one action receives a minimum score of 1. Furthermore, we exclude the top 1% of scores, which may represent outliers such as unique wallets potentially used for staking by centralized entities, to enhance the fairness of the distribution.

In this section, we address another category of activities that differ from those previously explained in detail. Unlike the prior activities, which involve locking assets for a specific duration, these actions occur instantaneously and thus lack a "duration" component. Examples include swapping one asset for another, purchasing an NFT, or transferring assets to another address. Consequently, we cannot define W for these activities. Instead, we consider the total amount involved in these activities as the metric for scoring users. We apply the previously defined scoring function, with the modification that W is replaced by V, representing volume. Thus, the score is calculated using the following equation:

Score = 15 × (V ⁄ V_max)^k   (4)

Following the staking analysis, we extend our approach to evaluate both the lending activities and liquidity provision of each user, recognizing that these activities share a similar nature, as both require users to lock assets within the ecosystem. The methodology for lending and liquidity provision does not differ from the staking section. For lending, we assess users’ locked funds in lending-borrowing protocols, while for liquidity provision, we analyze users who add liquidity on Uniswap Version 3. However, the parameters k, α_history, and α_current remain adjustable independently for each field. This flexibility is incorporated to allow coefficients to be set proportionally to the significance of the different activities within the ecosystem. For the purposes of this dashboard’s lending and liquidity provision analyses, as well as other subsequent evaluations, we adopt k = 0.5, α_history = 0.3, and α_current = 0.7, as our primary goal is to introduce the scoring approach rather than to assign specific scores to individual users.

In this section, we analyze another type of activities that are distinct from those previously discussed, by focusing solely on the number of transactions and using it to score users with greater precision. We did not count them previously because our goal was to analyze the material of the transactions in those parts, regardless of the number. After analyzing most parts, we now count all transactions irrespective of their type. This is due to its importance as a measure of blockchain adoption, scalability, and security. High transaction counts enhance network vitality by supporting diverse use cases and validator incentives through fees, and strengthening resistance to attacks. Moreover, the transaction fees collected provide a sustainable revenue stream for the ecosystem, which can be reinvested to fund development, improve infrastructure, and incentivize further participation, thereby fostering long-term growth. Given the computational intensity of this analysis, we set the initial reference point at the beginning of 2024. The coefficients are similar to the previous part.

In this section, we aggregate our scoring approach and explore its application in analyzing user behavior based on their interactions. Additionally, this methodology can be generalized to other domains not assessed in this dashboard. We have analyzed several key components of the Avalanche ecosystem, which operates a tri-chain architecture comprising the C-chain (for smart contracts), P-chain (for staking), and X-chain (for asset exchange) . We acknowledge that other fields may play a vital role in allocating scores to users.

For example, we can calculate the balance of each user and assign a score based on their balance; however, balances can be considered passive capital compared to assets locked in pools or full nodes as validators, as the former have a lesser impact on the ecosystem’s health compared to the latter. However, as previously noted, the primary objective of this dashboard is to introduce a relatively fair method for classifying users based on their interactions, rather than assigning specific scores. Furthermore, we cannot integrate all data presented in this dashboard for aggregation due to constraints on data access. For instance, we could not map P-chain users’ addresses, who execute at least one transaction on the P-chain, to their corresponding C-chain addresses, as Avascan does not provide an API to access and combine these data. Nevertheless, assuming all such data are available for a specific user, we can determine their score using the following convex combination equation:

β₁ × score_staking + β₂ × score_lending + β₃ × score_LP + ... + β_n × score_transaction = SCORE   (5)

where β_i ≥ 0 for i = 1, 2, 3, ..., n
and β₁ + β₂ + β₃ + ... + β_n = 1.

It is evident that these coefficients (βs) vary based on their importance. For instance, β₁ (for score_staking) is significantly larger than β_n (for score_transaction), as locking assets in a blockchain must be appreciated more than the number of transactions. Similarly, performing a swap is less important than providing liquidity for a pool but more significant than the number of transactions. However, setting these coefficients depends on the ecosystem’s insights and roadmap. For example, if the ecosystem aims to promote a specific action, it can increase the corresponding coefficient to attract user attention to that field. In addition to utilization by data analysts, this approach enables ecosystems to gain deeper insights and guide users along desired paths. It also serves as a valuable tool for allocating airdrops to users based on the scoring system. This can be achieved by showing their score in their native wallet or on blockchain explorers alongside users’ addresses (similar to SNS in Solana) for transparency.

Furthermore, in the aforementioned equation, we can introduce a coefficient β_incentive and assign it a score of 0. When an ecosystem introduces a new DeFi protocol or tool and seeks to encourage user participation, it can assign an incentive score ranging from 0 to 15 based on user interaction with that DeFi or field. The revised calculation becomes:

β₁ × score_staking + β₂ × score_lending + β₃ × score_LP + ... + β_n × score_transaction + β_incentive × 0 = SCORE   (6)

where β_i ≥ 0 for i = 1, 2, 3, ..., n, incentive,
and β₁ + β₂ + β₃ + ... + β_n + β_incentive = 1.

The scoring system proposed in this dashboard represents an initial version with inherent limitations, as it is a raw prototype. For instance, one might argue that applying a fixed coefficient to users’ past interactions (e.g., 0.3 for the period up to 90 days and 0.7 for the last three months) lacks fairness, suggesting that the influence of past interactions should diminish with increasing distance from the present. To address this, we can incorporate TD-Lambda from Temporal Difference (TD) Learning , a technique used in reinforcement learning to weight recent experiences more heavily [Reference: Sutton, R. S., & Barto, A. G. (2018). Reinforcement Learning: An Introduction, 2nd Ed., MIT Press]. A suitable function to reflect the history of user data based on temporal distance could be:

f(day) = e^{0.03 × day} where day ≤ 0   (7)

By applying this exponential decay function, we emphasize recent interactions by multiplying them by larger values, while the impact of past interactions decreases as we move further back in time. We conclude the discussion of scoring, acknowledging that many details remain, and reiterate that parameters and hyperparameters can be refined to meet an ecosystem’s specific needs. As an example, we employed a convex combination, though nonlinear combinations might prove more suitable for particular platforms.

Another commonly reported metric for blockchains is user retention, though approaches vary. We must differentiate between the retention of high-value users—those with significant activity or asset holdings, such as whales—and other users. Our scoring approach, which incorporates historical scoring , not only accounts for user retention but also accounts for differences among users, providing a nuanced evaluation of their contributions to the ecosystem.

This chapter explores insights into pattern recognition in data analysis, focusing on how data analysts can identify and present patterns that enable users to maximize profits while providing ecosystems with insights to implement incentive mechanisms for user retention, thereby enhancing the ecosystem’s overall growth. However, not all data required for pattern recognition are available in on-chain data. To gain deeper insights, off-chain data are necessary, and some may not be quantitative. Given that on-chain data are our primary tool as data analysts, we accept these limitations and aim to identify patterns to the extent feasible.

In traditional economies, tracking "smart money" reveals market trends, indicating reliable and profitable investment opportunities. In blockchain ecosystems, however, user identities behind wallet addresses are typically unknown, limiting us to address-based analysis. To adapt the "smart money" concept, we focus on whale wallets—those with large balances or high transaction volumes—as proxies for influential actors.

Stablecoins, such as USDC and USDT, are pivotal for ecosystems, facilitating low-cost, fast, and volatility-free transactions essential for real-world payments and DeFi. Their flows reflect economic health, driving adoption through integrations. Increasing stablecoin usage signals a pro-crypto environment, potentially bolstered by political support, fostering confidence and positioning a blockchain as a leader in blockchain-based commerce. By tracing stablecoin movements, we gain insights into user behavior.

Our initial focus is on cross-chain bridges, particularly users bridging stablecoins into the Avalanche ecosystem. These users often have specific intentions, as they could have held stablecoins in their original ecosystems if their goal were merely to hold. We analyze the inflow of assets via bridges to understand where bridged funds are spent and the activity patterns of users who bring assets into the ecosystem. Specifically, we examine the behavior of users bridging stablecoins into Avalanche to identify common patterns. It should be noted that we do not trace the exact assets bridged, as this is impractical—users may already hold such assets in their wallets, and exact tracing offers limited behavioral insight. Instead, we focus on user behavior to discern patterns among those who bridge stablecoins.

The following chart illustrates the volumes of the top five assets bridged over the past three months from these blockchains: Ethereum, Optimism, Binance, Polygon, and Arbitrum. To better understand the behavior of different user types, we categorize them into three groups: whales (top 1% by volume), sharks (next 10% of remaining users), and fish (the remainder). Should the scoring system become available, we can assess user behavior more precisely based on their scores.

Hereafter, we concentrate on users who have bridged USDC to the Avalanche network, as identifying patterns for USDC can be extended to other tokens, and our primary goal is to recognize patterns rather than analyze individual tokens. The chart below displays the volume of USDC bridged to the network over the past three months by the three user categories: whales, sharks, and fish. A frequently overlooked aspect in charts with user differentiation is the correlation between the behavior of each category. For instance, this chart reveals a higher correlation between the behavior of whales and sharks compared to fish with either whales or sharks. Additionally, we can assess these charts alongside market trends to determine whether there is any correlation between them.

The table below presents the activity of users who have bridged USDC to the Avalanche network across key segments of the network, focusing on their interactions with the USDC stablecoin. Specifically, we highlight their engagement in the swap and lending sectors, visualized through bar charts differentiated by user type (whales, sharks, fish). Analysis of these charts reveals that whales exhibit a stronger preference for trading in the swap sector, which appears more profitable, despite the higher risk compared to lending their stablecoins for lower returns. Additionally, we present the volume of all USDC transactions and the BTC price over the past ninety days. These visualizations provide valuable insights into the behavioral patterns of whales relative to market trends, with a notable increase in USDC transactions by whales since mid-April 2025, a period when BTC began to rise in value, while fish users exhibit a distinct pattern. These correlations between overall market trends and user behavior facilitate the recognition of behavioral patterns within the network.

At the conclusion of this dashboard, we propose the adoption of more dynamic metrics to better understand an ecosystem, as opposed to relying on passive metrics. Conventional analyses often focus on metrics such as new user counts, total number of transactions, TVL (Total Value Locked), and stablecoin market capitalization. However, as the mission of the IDG initiative requires, we must shift toward metrics that provide more actionable insights. For instance, merely reporting the number of new users offers limited insight to stakeholders. In contrast, by monitoring the behavioral patterns of these users across various blockchain segments, or by evaluating their accumulated scores over time to determine the proportion of high-value users, we can more effectively determine the percentage of long-term users who contribute to the blockchain’s health and stability.

Similarly, reporting the market capitalization of stablecoins alone provides minimal value, particularly when comparing blockchains, as the market cap of a stablecoin on Avalanche is not directly comparable to that on Ethereum. To address this, we draw on traditional economic concepts, such as the velocity of money—an indicator where a high value indicates robust economic activity, while a low value suggests stagnation. Rather than accepting this metric at face value, we adopt a novel perspective by analyzing its application to stablecoin circulation, aiming to uncover deeper insights into blockchain dynamics.We calculate the velocity of money for a token using the formula:

Velocity= total volume of transactions
market cap   (8)

By applying this concept to stablecoins, we derive a velocity metric that measures their rate of circulation through on-chain transaction frequency and volume. This reinterpretation allows us to move beyond traditional usage, providing a nuanced understanding of how stablecoin activity reflects ecosystem vitality. This metric enables more meaningful comparisons of stablecoin activity between blockchains. We have plotted this indicator across the three user categories (whales, sharks, fishes) over the past ninety days , highlighting the distinct roles each group plays in the circulation of stablecoins on the Avalanche blockchain. In conclusion, by reimagining a conventional metric like velocity of money through the lens of user behavior and blockchain-specific data, we offer a powerful tool for stakeholders to assess and enhance the Avalanche ecosystem’s long-term health.

Thus far, we have sought to identify patterns in user behavior by comparing user categories and external factors such as market trends. This approach can be extended to other user groups. For instance, instead of focusing on users who have bridged USDC, we could analyze users who have minted USDT and identify patterns in their behavior. Alternatively, we could calculate the average balance of users in AVAX and stablecoins over a specified period, rank them from highest to lowest, and examine their behavior based on their balance. It is worth noting that our goal in these examples is to track smart money, emphasizing the role of whales in the blockchain context—those with larger asset holdings or who have recently executed transactions with larger amounts. Furthermore, if a scoring system is available and can be applied to these users, it would provide clearer insights into each user group. Our objective has been to track smart money in the network to deliver actionable insights to stakeholders by analyzing user behavior.

We will now examine the Avalanche ecosystem in comparison to other blockchains. A key indicator for comparing blockchains is the flow of assets into and out of the network, prompting us to revisit cross-chain bridges. The charts below illustrate the five currencies with the highest dollar volume bridged to Avalanche, followed by the five blockchains with the highest dollar volume bridged, over the past ninety days . We have refrained from including additional charts in this section, although for a more detailed analysis, these inflows and outflows could be plotted over time or analyzed by user type. Additionally, the minting and burning of stablecoins within the ecosystem could be examined. This perspective enables us to both identify patterns in user behavior and provide network founders with deeper insights into the ecosystem’s strengths and weaknesses by tracking inflows, outflows, and liquidity volume.

Credits

We utilized @charliemarketplace's query for calculating the stablecoin market capitalization on Avalanche, which was integral to our velocity of money analysis.

The query can be accessed at:
https://flipsidecrypto.xyz/charliemarketplace/q/Ma1Dp2lZvEuI/kpi---native-stablesupply