A $10000 BTC Bounty for an open source tool to export your Twitter followers

The community has crowdfunded a $10000 bounty, payable in BTC, for an open source tool that helps people export their followers from Twitter to Substack, Ghost, Locals, or other user-controlled platforms.

Submit your bounty entry here
Read on for detailed problem description and possible solutions

The goal is to take a Twitter account with N followers and produce a list of N emails (or phone numbers) for those followers with their explicit consent to contact them on another platform, and without bothering them very much or at all. The ideal solution shouldn't take too much time or money on anyone's part.

There are many possible approaches to this problem, and the best one will win the first $10k committed. With any remaining monies, we'll fund testing and/or second and third place prizes.

We care more about the results than the method. We are not going to be prescriptive about the best approach to take, though we do suggest two possible ideas below, the "mass DM" and "affiliate link" approaches.

How do we determine the "best" approach? We are thinking about scoring the different approaches by allowing different accounts to try out the top submissions on a subset of 1000 followers, and then rating them by conversion percentage, cost-per-conversion, ease-of-use, and happiness-of-converted-user.

Because a tool like this is desired by many people, a good solution to this bounty could become a product or even a startup. You'd want to first get it to work robustly for Twitter influencers, probably via white glove service for the first 100 or so accounts. And then make it work for other social platforms. Companies like Buffer have done surprisingly well ($20M+ ARR!) with similar seemingly simple products that fill a hole in the social networking landscape.

Problem: exporting followers from Twitter

Twitter is an incredible platform for learning and finding people of like mind. I admire much of what Jack Dorsey and his team have accomplished.

However, it has several downsides for users with large follower bases, as it lacks tools for:

distribution: you can't contact all of your followers without permission from Twitter
customization: you can't customize the experience for your followers or offer them added benefits
monetization: you don't make money as a Twitter influencer, and don't get a storefront to sell things
moderation: you can't maintain a civil tone among your community
information: you don't have individual-level search and analytics on your follower base
prioritization: you can't up- or down-regulate what appears in your feed or that of your followers
brand management: you don't control your own brand or domain name, as it's twitter.com/example rather than example.com

For these reasons and more, many people are moving to platforms like Substack, Ghost, Locals, or the like. In these platforms, an influencer has full root privileges over their community and can give a richer experience. You don't need to buy Twitter ads to reach your followers, or trust Twitter to deliver your content to your followers' feeds.

The issue, however, is that it's not trivial to export a large follower base from Twitter!

Note that the problem here is not the export of one's own profile data. That is relatively easy and Twitter already supports that. The issue is not the access to your own data, the issue is the ability to contact your followers (eg via email or phone) without Twitter's additional consent.

Solution

As noted above, the ideal is for an account with N followers to get a list of N emails (or phone numbers) for those followers with their explicit consent to contact them on another platform, and without bothering them very much or at all. In practice, you won't get all N, but you may get a subset. Also, the ideal solution shouldn't take too much time or money on anyone's part.

In theory, Twitter could add a feature where users could opt in to allowing their emails to be viewed by a few, some, or all of the accounts they followed, perhaps in return for a micropayment of some kind. For example, an account could pay $X for the verified email of each follower with an account age over 3 months and >100 real followers themselves.

But in the absence of a built-in feature like this, there are several overlapping ways to export your userbase to example.com/subscribe.

Put example.com/subscribe in your name, bio, photo, and profile
Create a pinned tweet directing your followers to example.com/subscribe
Regularly post a tweet with example.com/subscribe
Publicly tweet to each follower with example.com/subscribe
Set up a separate new account that does #1-4, to keep the public mentions out of your timeline
Parse the bios of each follower account for public contact info (eg if they have their email in their bio), and email them out-of-band with example.com/subscribe
Set up a welcome message with example.com/subscribe for each new follower
Privately DM each follower once with example.com/subscribe
Generate a per-user affiliate referral link, and pay follower to refer others

None of these are perfect. You might want to use several of them at the same time. But let's discuss the last two in some detail.

The Mass DM Approach

The concept here is to send a private DM to each of your followers with a link (and possibly some incentive) to sign up for a new platform, like Substack.

The simplest way to implement the mass DM approach might be a command line app that takes as input a Twitter API key and a message to send each user. It would give you a preview of all of your followers, addsome options to rank them by importance, and let you try it out by sending test messages to a few accounts before opening it up to message 1000 accounts per day. It stores state so you know who you messaged in the past, such that you don't inadvertently recontact them.

You can imagine a fancier local Mac App that puts a nice GUI on top of the command line engine described above. Or an even fancier hosted version with individually attributable conversion links, so you know which usernames converted to which emails. The hosted version would also allow you to keep running the 1000 DMs per day in the background, so it could be a SaaS service.

The Affiliate Link Approach

This is a very different approach. The idea here is to make export from social networks as organically viral as the original import was.

One way of doing that is to generate an affiliate link for each follower, such that the one who refers the most/best users to your site each day receives $100 in crypto.

Why "most/best"? Because you want some measure of quality, not just N spam emails. Why crypto? Because it's easy to automate small, fast, international payouts.

A major advantage of this approach is that it can generalize to other social networks. It's also much harder to shut down than mass DM, and it aligns the crowd with you economically.

Here's one possible flow, in which all emails are recorded at a single site that tracks everything and then re-exported into Substack, Ghost, Locals, Mailchimp, or other applications. Note that this will work for free email lists, but the upsell to paid subscriptions as per Substack would then have to be done separately as this implementation doesn't also include a checkout flow.

Feel free to modify this.

Phase 1: Set up referral URL

Suppose we define three parties: the developer, the influencer who wants to export their followers, and each follower themselves.

The developer sets up a site at (say) socialexporter.com
It says EXPORT YOUR FOLLOWERS
An influencer lands on this page and logs in with their Twitter account
The influencer gets a public page generated for them at socialexporter.com/username
The influencer can admin this page at socialexporter.com/username/settings
This public page has some copy asking a user to sign up and explaining where their email will be used (eg Substack, Mailchimp, etc)
It also has a small form where a follower can enter their email address
After this email address is entered, a second small form is displayed where a follower can optionally generate an affiliate URL (more on that below)
And a link appears to socialexporter.com/username/leaderboard, which shows a leaderboard of which followers referred which emails (all zero at the start)
Separately, when logged in, the influencer sees a different private page with all their followers (kind of like Social Blade)
The private page also has a Bitcoin address to deposit funds
The influencer sends (say) $1000 to that address to fill up the referral budget
And the influencer sets a daily budget of (say) $100 for payouts, and can edit the copy on the private page

At this point the influencer has written copy encouraging users to sign up with their email, as well as allocated a budget to incentivize this action.

Phase 2: Post referral URL

Now the influencer incentivizes their audience to refer people.

The influencer tweets out socialexporter.com/username along with a message
"Hey, I'm setting up my mailing list. You can sign up here and get paid to refer others: socialexporter.com/username"
A Twitter user sees this tweet and visits socialexporter.com/username
At first they just see copy encouraging them to sign up to the username email list and where it will be used (eg Substack, Mailchimp, etc)
After they enter their email, they then see a second small form where they can optionally generate an affiliate URL
The site may require them to be a follower of twitter.com/username for this affiliate link to work, not just a random user
It's a bit.ly-like URL generator where they can type in their Twitter username to generate an affiliate link
So a follower named foobar would get an affiliate URL like socialexporter.com/username?referer=foobar
A user who landed on the site through this affiliate URL and then entered in a quality email would give foobar credit on the socialexporter.com leaderboard
Anyone can view this leaderboard at socialexporter.com/username/leaderboard to see themselves alongside the top referers for the day for this username, ranked by the number of quality emails they've referred, similar to pioneer.app
Quality can be determined by something like kickbox.com or something more sophisticated, to filter out fake signups

Phase 3: Payment

Every 24 hours, the socialexporter.com site issues payments to the top referers.

First, all emails are analyzed for quality using a tool like kickbox.com or more sophisticated custom code
You might require an email validation, for example, or similar criteria to consider an email to be "high quality"
Once you have the number of quality emails referred by each user over the last 24 hours, payouts are calculated
In one model, 50% of the payouts are proportional and 50% are a prize to the best referrer
So if there were three referers who referred 30, 20, and 10 quality emails respectively, along with a daily prize of $100, the first referer would get $50 + (30/60)($50), the second would get (20/60)($50), and the third would get (10/60)($50).
These payouts would be sent to the Bitcoin addresses on file every 24 hours
If transaction fees are too high, payments can be batched, sent via an off-chain method like Coinbase, or sent via another cryptocurrency
In this fashion one can make fast, international, automated, small payments to users for decentralized work

Phase 4: Tokenization

Note that the affiliate link approach, unlike the mass DM approach, is more general (as it works across social networks) and doesn't make use of the Twitter API. However, it does cost some money to operate. A variation of this method would allow the influencer to compensate users in a new cryptocurrency issued for the purpose (such as Reuben's Bramanathan's personal token).

Note that this whole user story is just a sketch of an incentivized social referral model for viral export. It can be modified in many ways.

Bounty: $10000 bounty for open source Twitter export tool

Here are threads with context on the bounty:

Initial thread
First followup, with increased bounty
Second followup, with affiliate link approach
Third followup, with updated writeup and scoring method

Please follow @balajis for future updates.

Bounty Scoring

As noted above, we are thinking about scoring the different approaches by doing two passes.

First pass: identify promising submissions versus those which are obviously buggy by looking at the code, polish, visuals, etc.

Second pass: work with volunteer accounts to try out the top submissions on a subset of ~1000 followers and rate each submission by conversion percentage, cost-per-conversion, ease-of-use, and happiness-of-converted-users.

Let's define these terms.

conversion percentage: out of N followers, the percentage that submit their emails (eg 10%)
cost-per-conversion: the amount of money paid to convert emails divided by the total number of emails collected by the end of the campaign
ease-of-use: self-reported subjective measure by the influencer who set up the campaign
happiness-of-converted-users: self-reported subjective measure by the users who were pinged to submit their emails

Assume that the only emails that count are those that are above a quality bar, as measured by kickbox.com or a similar set of quality filters. It goes without saying (but we're saying it) that all users must consent to submitting their emails, ideally via email verification or some other form of demonstrated consent (such as clicking a link and manually submitting their email).

We're open to feedback on this scoring system, but it appears to capture the spirit of "export Twitter followers to an email list" without being prescriptive on the exact method.

Bounty Commits

Enough other people were interested in this that the total bounty funds are now >$10000 as of Monday June 15, 2020.

As more money comes in beyond $10,000, we will allocate the first $10,000 to the winner. We'll then allocate remaining monies to folks who test out the app and/or possibly to second and third place submissions.

Bounty Logistics

In terms of mechanics, I (@balajis) will decide on the winner by June 28, 2020, and then the BTC address of the winner will be provided to the various people who have publicly supported the bounty.

It will be incumbent upon each of these people to pay what they have committed to the winner. They can optionally publicly post an on-chain confirmation.

FAQ

How to submit an entry for the bounty?

Go to this issue and make a comment with a link to your project. Ideally it should have a gif or something that shows how it works.

What about other approaches?

As noted above, the ideal is that if you have N followers, you get a list of N emails (or phone numbers) for those followers with their explicit consent to contact them on another platform, and without bothering them very much or at all, or requiring too much time or money on the part of the influencer.

If you can achieve this goal another way besides the mass DM or affiliate link approach, that's fine. We'll score them as per the bounty scoring section by running them on a subset of followers.

What about Twitter API and DM limits?

These are a critical constraint. If you have a solution that works in theory, but that breaks on accounts with many followers, or gets an account banned or shadowbanned for spam, or has some other serious negative side effect, then you don't have a solution.

What about existing tools?

Here are a few that are relevant:

Example of using Python API to extract followers: get_followers_bios.py
Example of getting followers without official API: Twint
Command line tool for interacting with Twitter: t

There are tons of open source tools for working with Twitter out there on GitHub and the broader internet. Feel free to use them rather than reinventing the wheel.

Mass DM specific questions

As noted, there are multiple possible approaches to this problem including the "mass DM" and "affiliate link" approaches, as well as something creative you might come up with. If you take the mass DM approach, here are some answers to FAQs.

If I do the mass DM approach, can it be command line only?

That's the simplest version, but ideally there should be a simple Mac app and a hosted version as well.

The command line version would be for developers.
The Mac app would be for folks who aren't developers, but don't want to give their Twitter credentials out to a new website. They'd prefer to paste in an API key locally.
The hosted version will probably be the most widely used, where someone logs in with Twitter OAuth and sets up this mass DM to run in the background, with analytics on who has been contacted, who signed up, and so on.

However, because there are many fly-by-night Twitter apps, and because of the nature of what this app is doing (mass DM of followers), the influencer will want to carefully babysit it and the presence of the open source versions will increase trust in the hosted version.

Why would ranking followers be important for the mass DM approach?

Because you can only send 1000 DMs per day, you will want to prioritize your followers such that you export the most important ones first. These might be the ones who have the most followers themselves, or who have some attribute in their bio (like a #Bitcoin hashtag), or that have some other criteria. In practice this means that you'll likely want a follower table, where the first column is the Twitter username and subsequent columns are metadata on that username (along with a timestamp for when that metadata was last collected, as it can become stale as people update their profiles).

Can the mass DM tool run for multiple days?

Yes, Twitter's default API has a 1000-daily-DM limit. So your tool might need to run for multiple days.

The open source version would run locally and have some state, like a SQLite database or a flatfile. It would use that state to record what people it had DM’d and when, along with what was said and whether they responded. You should also allow re-running of the tool periodically as new followers arise, without requiring the influencer to think about whether they are inadvertently messaging people they already messaged.

Can the mass DM tool require something above the basic API?

Maybe, so long as that API access isn't too expensive or time consuming to get for the typical Twitter influencer with (say) 10,000 followers.

The DM API for customer service seems like it might be quite relevant.

Next steps

The problem of exporting your following applies to all social media, and has been discussed for the better part of a decade, but it now feels like it's gotten to a boil. There is a critical mass of people who are moving to platforms like Substack, Ghost, and Locals that offer greater control over your audience (and, with it, monetization for the influencer and benefits for the audience).

If we can solve it for Twitter, we can probably generalize this to work for partial exodus of influencers from other platforms like Facebook and Instagram.

Please note that Twitter and company will be fine and won't go away any time soon. This is a cave-and-commons approach. The large global social networks will likely persist for some time as "commons" where you recruit members, but these "caves" off to the side with individual influencer-led communities will thrive.

Finally, as noted above, companies like Buffer that solve seemingly simple problems like this have gotten to $20M+ in ARR. If you can build and maintain a tool that reliably exports a large percentage of follower emails from Twitter and other social networks, with the consent of the users and at a reasonable price, you have something that a large number of influencers will likely pay for.

balajis / twitter-export

readme