I'm four editions into my App.net newsletter and it has definitely been an eye-opener into running a regular publication. Don't get me wrong, I'm enjoying the process of putting together each newsletter and finding content to include, but there is a few areas where I could improve the newsletter and the process of putting it together.

Better user recommendations

This is a new section which only appeared in the third edition. The idea is that I monitor the new accounts being created on App.net and poll for a few days before the newsletter goes out. I filter for possible spam accounts and only include accounts that have been marked as 'human'.

On the weekend before the newsletter goes out I can then analyse the most active accounts for that week and pick out some recommendations for the subscribers of the newsletter.

There's so many parts of this process that can be improved:

1. Discount feed accounts - A lot of 'human' accounts are using tools like Twitterfeed to automate posting to their account. These are not the type of accounts I want to recommend.
2. Filter for users that don't have a bio - There's not much use in recommending somebody if they don't have their bio filled in telling people about themselves.
3. Automate the user polling script - I manually kick off a couple of scripts to poll the App.net API. Ideally this should take care of itself and just run on an interval. Nothing to stop me doing that on my laptop, but I should probably schedule it to start during the day when I know my MBP won't be in hibernation.
4. Scripting the users list - I put together the list of recommended users in the format of Markdown (like the rest of the newsletter), although there shouldn't be anything to stop me grabbing my list of recommended users and exporting it as a snippet of Markdown that can be included in the newsletter.

Conversation recommendations

I'm still undecided about this one, but another idea for App.net members was a service that reads back through your timeline for a given time period and weights conversations based on the number of replies, stars and reposts it gets. On a daily basis, subscribers would then get an email showing the most active conversations over that given time period.

I want to include this in the Netterpress newsletter, but on a per user basis it makes more sense to run it as a separate service. The idea could still be used in the newsletter, but instead of reading back on a specific user's timeline, it could poll the trending conversations feed and use that to search for active conversations.

More automation

I've already talked about automation earlier, but where I also want to automate the newsletter is the finding of new content. It's a manual job right now. I have included hashtags in the first two editions of the newsletter that people can use to tag their posts if it contains news that could be included in the newsletter, but the response to this has been limited.

Going well

So far the newsletter has shown me that putting a newsletter together involves a lot of work. It has been hard work balancing this with freelance work, but with more subscribers I could start blocking off more time during the day to getting an automated news discovery and publication process put together. It's still early days for the newsletter though, and I'm only halfway through the trial run.