Spider based on tags for maybe the last 6 months, taking powered up Steem and other coins in as a factor maybe? Maybe consider comments and the likelihood of them getting a reply and upvote from the author and random users. It depends on the platform. If it was a trading platform, I might have to actually consider various times in their wallet history when they transferred money to an exchange, then later transferred money back. I guess the best metric would probably be interaction with the community based on tags. So, if it was trading, I would check trading related tags, who commented how much, how likely they were to get a reply, how likely they were to get an upvote on that reply from the OP and for how much, and how likely they were to get a random upvote, weighting it against repeated upvotes to attempt to avoid bots.