An update to our ERC-20 methodology

Recently, it was brought to our attention that our figures for transaction count and transaction volume for ERC-20 tokens were quite significantly different from those of Etherscan.io. We looked in to this and realized that we were undercounting these figures – dramatically in some cases.

The problem was this: we were ignoring token transfers made indirectly – through the invocation of third-party contracts which in turn called token contracts. Previously, we were only counting ERC-20 transfers triggered by the direct invocation of the contract. This meant that we were undercounting transaction count figures by as much as 58 percent in some cases (VERI), and in extreme cases undercounting transaction value by 93 percent (Gnosis).

We have completely resolved the issue now and made our numbers consistent with Etherscan’s ones. Any differences you see with Etherscan or other aggregator sites will be the result of different methodologies over when the daily close is.
When we noticed the problem, we could have quietly fixed it, probably without many people noticing. However, we prefer to be as transparent as possible, so these are the steps we are taking:

• we are posting this public mea culpa
• we are posting the fix to our github
• we are creating a data archive on our github, which we will update if/when we make another change to our methodology, so its impact will be visible
• we are posting a list of the worst discrepancies

We’ve listed the 10 worst discrepancies by transaction count and transaction value below. Note that these are all-time figures.
Transaction count:

Transaction value:

Our sincerest apologies for this misstep. We will try to do better next time. We would like to thank Stephen Harrington and Martin Skakala for bringing this to our attention.