How we model the data for crypto exchanges
The modeling is conducted in three steps:
Step 1: We set up an initial model starting with the daily volumes multiplied by the estimated median fees.
Step 2: We conduct a multi-step iteration adjusting the fee distribution and adding non-fees revenue streams.
Step 3: The model stops iterating when we calibrate for all reference data points captured in the public and non-public sources.
Step 1: Initial model set up
Large array of data is gathered covering exchange volumes, transaction, users, fees, etc. To achieve that we tap on a few dozens data public and non-public sources. We constantly monitor the web for the appearance of the new data points (CEO speeches, conferences, submission to regulators, etc.). The key to success is to cover the broadest data set that is currently available.
Step 2: Data modeling via iterations
We conduct a multi-step iteration adjusting the fee distribution and adding non-fees revenue streams. It allows us to obtained a detailed model of a crypto exchange business linking financials (Income Statement and Balance Sheet), operations (Operational model) and people (Headcount model). The model is running multiple iterations to minimize the difference with calibration points
Step 3: The model is calibrated on reference points
We make a judgment-driven assessment using cross-validation with multiple sources. Thus, we are controlling quality of input data. We are constantly looking for new data to further calibrate the model