Content
@
0 reply
0 recast
0 reaction
shazow
@shazow.eth
would the devs plz do something? 💜💜💜 https://warpcast.com/kenny/0x2c86fdc6/quotes
1 reply
1 recast
4 reactions
Dan Romero
@dwr.eth
We do algorithmic reply bumping. Strict rule based performs worse across every retention metric.
1 reply
2 recasts
3 reactions
shazow
@shazow.eth
Any intuition you can share about the current reply bumping approach? Also were you able to check out this commissioned piece? Any feedback on the analysis? https://paragraph.xyz/@yesyes/checking-the-comment-velocity-of-real-users
1 reply
0 recast
0 reaction
Dan Romero
@dwr.eth
Algo is machine learning. It chooses what you’re likely to engage with. Assumption in 1st paragraph in article is incorrect. We changed it because reverse chronological reply bumping feeds are way worse for retention. Can read the rest later.
2 replies
0 recast
3 reactions
yesyes
@yesyes
The assumption in the first para was taken directly from your post https://warpcast.com/dwr.eth/0x5a5a09ab Though, i should mention that the article is a little outdated now. Will update the metrics soon.
1 reply
0 recast
1 reaction
Dan Romero
@dwr.eth
That’s channel feeds not main feeds. :)
2 replies
0 recast
2 reactions
yesyes
@yesyes
Yeah I realized that after rereading it lol. Will correct it in the article. I think the rest of the article still stands strong, I just objectively looked at the data.
1 reply
0 recast
1 reaction
Dan Romero
@dwr.eth
Right, not disputing the data. What you’re missing is new user retention, which we can measure based on app usage and time spent. You can partially do it with protocol data using public signals like a reaction. The reply bumping main feed performed worse than the algo feed.
1 reply
0 recast
0 reaction
