Neynar

Celebrating the experimental launch of hubs.neynar.com

Seeing some confusion about scores. A few things to note:

1/ 

1. Scores are not evenly distributed, higher scores are increasingly difficult to get. No one would expect everyone with access to credit to have a perfect 850 credit score. Same with social scores. For developers, we recommend starting with a threshold between 0.5-0.6 and moving up/down as needed.

2. If you're building with scores, there's no need to set a super high threshold. That artificially reduces the number of users who can interact with your product. Imagine only giving loans to people with 850 credit score, pretty much no one would have access to credit.

3. Only 27k+ users have scores 0.7 or higher at this time, the protocol has 800k+ users. There's really no need to worry about a low scores unless you're actually in the bottom digits. You can see a live dashboard of the distribution here: https://bit.ly/3ZQItzH. We are working on model improvements and we expect more users to move up the score ladder.

Have to apologize in advance for bringing this up again. Boz at facebook used to talk about how "communication is the job" and I feel like the learning only gets more pronounced over time. 

Thus, PSA for all Neynar developers - *start with a low score threshold* 
unless you actually want a very small cohort of users to use your product. 

Someone with a 0.77 score contacted me yesterday and they were prevented from taking an action. As mentioned in the linked cast below, 0.7+ is top 3% of accounts on the network. It's already ~27k accounts (of 800k+ accounts) at that threshold. It's not that many users.

It might feel like your product only needs users with scores 0.9+ but that artificially restricts to ~2500 users! It's likely not what you want. Start with a low threshold and increase as needed.

If this confusion remains, we might have to change the shape of this product. Maybe next week.

Have to apologize in advance for bringing this up again. Boz at facebook used to talk about how "communication is the job" and I feel like the learning only gets more pronounced over time. 

Thus, PSA for all Neynar developers - *start with a low score threshold* 
unless you actually want a very small cohort of users to use your product. 

Someone with a 0.77 score contacted me yesterday and they were prevented from taking an action. As mentioned in the linked cast below, 0.7+ is top 3% of accounts on the network. It's already ~27k accounts (of 800k+ accounts) at that threshold. It's not that many users.

It might feel like your product only needs users with scores 0.9+ but that artificially restricts to ~2500 users! It's likely not what you want. Start with a low threshold and increase as needed.

If this confusion remains, we might have to change the shape of this product. Maybe next week.

https://boz.com/articles/communication-is-the-job 

working @ /neynar   🪐  casting @ /rish

founder /openventures | /proxyswap | tokenbot @clanker

not a problem, communication is the job :)

rish this is really on us, sorry that a lot of comms is falling back on you. will synch with Jack today to see what immediate changes we can make that will alleviate the issue.

The API could offer more ranking methods for developers, like adding OpenRank. Then devs can consider both, but this may introduce other problems, like latency..🙂

cypherpunk wannabe, engineer working on @frames /open-frames /opencast
https://stephancill.co.za create stories at https://stories.steer.fun

they're already there https://docs.neynar.com/docs/neynar-user-quality-score

Values-Aligned Communities | citizen of Black Rock City

Ooh good idea @manan @vrose.eth 

Let developers select what quartile(?) the threshold is

One way you could communicate it is just a enum mapping of ~5 levels of users that just map to scores, and let devs use the enum. 

(The enum can have a docstring with roughly how many users meet each level, or % of the network)

model will take a few iterations to stabilize

.92 is within top ~2k accounts on a network of 800k+ ...?

what is the discrepancy with the fact that I was at a 1 and now I’m somehow at a .92

dashboard is linked on this doc and in the cast I linked:

The raw feed. Good anoncasts get promoted to @anoncast by $ANON holders.

dashboard is linked on this doc and in the cast I linked: https://docs.neynar.com/docs/neynar-user-quality-score

On communication thought you might enjoy this article @royalaid.eth

"communication is the job" is a great saying

Maybe consider an additional API method that provides a percentile score e.g. 60th percentile = 0.N and encourage devs to use the percentile API (with guidance around what proportion of the population each percentile currently covers) so their app doesn't hardcode a value?

Hey 👋! Since the update, my Neynar score has dropped significantly. I haven’t been very active lately, but the gap with OpenRank is huge. What’s the reason for this drop? My previous score was much better 🥲