Insomanic: Andy Young's Blog

Selective Tweets: 3 years, zero marketing and nearly 1M users later - now open source!

After 3 years, nearly 1 million registered users and zero marketing, I’ve just released the code that powers Selective Tweets as an open-source project - you can check it out on GitHub (it’s PHP). I hope this will release a new swathe of value to interested developers and the app’s future. Is Selective Tweets most popular Facebook app ever to be open sourced? If anyone knows of other examples please do share!

I created Selective Tweets back at the beginning of 2009 to allow Twitter users to cross-post only selected tweets to their Facebook pages. I had the idea while listening to a rant in the pub by my GroupSpaces business partner @langer how he didn’t want his Facebook stream polluted by people “oversharing” via use of Twitter’s default Facebook app that simply cross-posted all tweets. It occurred to me that many people used Facebook and Twitter to share in fundamentally different ways – in general I saw “Facebook-worthy” updates as a subset of “Twitter-worthy” ones. Immediately came the idea for selective cross-posting using the hashtag, and Selective Tweets was born.

The initial app took about 6 hours to build and I told about 5 people – back then Twitter was far from mainstream and that was roughly everyone I knew who actually used it. That was essentially the only “marketing” I ever did – from the outset the concept worked and a small number of people started using it. Seemingly there was enough desire (or peer pressure) for Twitter users not to continuinally post all their minor updates to Facebook, not just a desire from the rest of their non-Twitter-using friends to preserve their Facebook experience.

Learning about growth

That was pretty much it for the first couple of weeks - about 14 people heard about it and signed up in total. Things got a bit interesting in early Feb, when suddenly ~35 people signed up over two days - I think from a tweet or two. However things quieted off. Then on 10th Feb nearly 150 people arrived:

As it turned out, this influx of 150 users was finally sufficient seeding to get the word going - without any major changes to the app the rest of February looked like this: 

February accelerated into March and April as the total number of users grew to ~25k within 3 months:

And that was basically it - the app had achieved a self-sustaining loop and sufficient seeding that the next 2.5 years looked like this:

Finally, a we get a different view by looking at how many people visited the app to sign up or configure it over time - in this view you can see a couple of peaks followed by stabilising at a higher level than before the peak:

The app has been a great source of learning for me personally, certainly providing great first-hand insights into the realities of viral/word-of-mouth growth. Another day, in another post I’ll come back to the growth and my interpretations and takeaways in some more detail.

A business case?

Once usage took off, the challenge that appealed to me was to see how big it could get and learn what affected the growth. With this in mind I didn’t want to introduce barriers to entry in the form of charging. Separately I didn’t want to distract focus from my day (and night) job - my main startup GroupSpaces. I did take the opportunity to use the app as an experiment in donationware, adding a PayPal button and in 2009 experimenting with YCombinator social payments startup TipJoy.

Sadly TipJoy brought in almost nothing in pledges and precisely zero in hard $ before it - perhaps unsurprisingly - shut down later that year. On a brighter note the PayPal link continues to bring in $10-20/month, which tends to sit in my account until I get those emails about friends running marathons, climbing mountains and undertaking other insane acts in the name of charity.

Those who are business-inclined will calculate my Average Revenue Per User comes in around a cool $0.0005 – unlikely to win me any VC money based on revenue projections. A damning take on microdonation models? To be fair I could have tried to push monetisation a whole lot more – I’ve not tried Flattr and they seem to have survived the TipJoy fate. Not that I’ve ever needed to worry about monitising the app – nowadays it runs on a single Amazon EC2 Micro instance on the free tier, costing me a grand total of ~$4.50/month (due to over-quota I/O requests). A pretty convincing example of just how much the overhead costs of web apps have plummeted, methinks.

As the app nears 1m registered users it’s currently used by 150k unique users each month to send more than 2.6m updates from Twitter to Facebook – on average someone tweets using the hashtag #fb once every second. The hashtag has evolved a legacy of it’s own, inspiring similar use in subsequent years by the official apps from LinkedIn (#in/#li) and Yammer (#yam). Now the 5th most-used hashtag of all time, while it’s unlikely to overthrow #ff (for “follow friday”) anytime soon - #justinbieber is just a passing fad, right? ;)

The next step: open source

So what’s in it for me – why open source? To be honest, I’m going to discover in time – like many things this is a punt-and-see experiment. The code could be educational as a case study for those new to building apps on social APIs. I think it’s an interesting example of the success of simplistic product design and word-of-mouth spread (I’ll analyse this further in another post). In true open-source fashion the field is open for others to develop their own improvements and fixes and contribute back and/or host their own apps.

At this point I don’t feel particular “risk” in releasing the source – I don’t expect another app to gain significant traction in the same area anytime soon. I’ve purposefully kept Selective Tweets simple for mass appeal – now it’s achieved a significant established userbase and brand awareness. Things could get interesting if someone manages to tap a new viral channel - in which case we will all benefit from the learning. The MIT licence I’ve released the codebase under permits all forms of use/reuse so long as my original copyright and credit is left in place – at the end of the day it’s uncomplicated code, if anyone can make something else from it then that’ll be another win in my book.

I’m looking forward to see what comes in the next chapter for Selective Tweets, now it’s not (necessarily) just down to me!

If you enjoyed this post please feel free to engage in the comments below or find me on Twitter: @andyy.