Twitter says more than 50 percent of tweets identified as abusive are now being flagged by its automated moderation tools first and taken down before a user reports them. Improvements in its machine learning model for identifying problematic tweets enabled the faster moderation, Twitter said.

Cracking down on abusive tweets has been an ongoing goal of Twitter’s. The company regularly talks about how “one of our highest priorities is to improve the health of the public conversation on Twitter,” as its earnings letter says today. Ongoing abuse has made Twitter an unpleasant place to be for many longtime and highly active users, so its ability to cut down on bad actors is important to the site’s usage and user growth.

Twitter says its automated flagging has grown from 43 percent of all abusive tweets last quarter and 38 percent the quarter before that. To be clear: this doesn’t represent every single abusive tweet on Twitter, just the ones that Twitter has ultimately removed as abusive. A problematic tweet that Twitter either missed or ultimately chose not to remove wouldn’t count, while a tweet that was mistakenly removed as abuse would.

Those are important caveats, but Twitter would at least like us to think that these numbers represent an overall improvement in moderation. Twitter says that it’s also been sending more flagged tweets to human moderators for review and that reviews have been happening “more proactively and faster than before.” More abusive content was taken down in total, too.

As it’s said in the past, Twitter continues to acknowledge that it has more work to do on abuse and that its goal is to reduce “the burden on victims of abuse” and to take “action before abuse is reported.”

The moderation numbers came in Twitter’s latest earnings report, which also saw the company report adding 6 million new daily users during its third quarter. Twitter now has 145 million daily users, up from 124 million this time a year ago and 139 million last quarter.

User growth has been a big struggle for Twitter over the past couple of years, and today’s numbers at least suggest that it’s continuing to head in the right direction. They don’t tell the full story, though: until recently, Twitter disclosed how many monthly users it had, rather than daily users, and that number was shrinking for a year straight up until the beginning of 2019.

While Twitter’s daily usage numbers are a lot rosier, they’re still fairly small. Snapchat now has 210 million daily users, and as of the beginning of this year, Instagram had 500 million people just using Stories every day.

Despite the generally positive user numbers, Twitter’s stock is taking a serious hit on today’s earnings report, with share price down almost 20 percent before market open. Twitter turned a profit again, but the company only reached $44 million, narrowly missing its guidance and falling from $92 million the same quarter last year. Revenue of $824 million also fell near the bottom of Twitter’s projections.

Twitter said a number of things were at fault for the weaker-than-hoped revenue, including bugs that limited its ability to target ads and share data with advertisers. Bloomberg reporter Kurt Wagner said Twitter’s ability to target and measure ads fell significantly after the company stopped using data for advertising that users had not granted the company permission to use, as it had been doing in some cases for around a year. That mistake was disclosed in August.