This post has been a long time coming — Apologies to those that have been waiting for it.
2010 was early days of mobile in India, we had recently pivoted from being a automated curator of publisher content (@Taazza) to a local aggregator of information on mobile (@Localbeat) .
We were betting on the mobile market in India. Sub 250$ android phones were starting to get into the market and we expected the price points to get lower and for the mobile market to take off in the next 12-18 months — something that would eventually take 36-48 months while we ran out of runway!
Finding mobile users in India in 2010 was painful and unscalable.
We tried google ads but the conversion rates very pretty low. Getting press coverage without a significant announce was practically impossible.
We needed to find a way to expand our early users and do it soon.
Initial Experiments
I was an early user of Twitter and was using it at the time to connect and network with folks for business development — also keep up with what our competition was doing across the globe. The audience was fairly engaged and my tweets about @Localbeat would resonate with my audience. Especially ReTweets — which had a multiplier effect. In fact this is how we got our first 200 users.
Twitter broke the news about the aircraft landing in the hudson in 2009 — which lead to it taking off in various parts of the world including India -where it had grown from a few hundred thousands to a million or more.
We knew we had to find a way to engage this audience in India. One of the lesser known features was the meta data that was exposed with every Tweet — Which client did they Tweet from! And most of the clients were either platform specific or had a unique name based on the platform — “Twitter for Android”. This is exactly what we needed!
So in the early days of this experiment, I tried a few options
- Search for city names and send tweets to folks
- Search for a hashtag and find people to send a tweet to
The results were mixed and quality of the responses weren't great. Finally, I ended up using Twitter advanced search functionality and filter by client name and screen the stream manually to look for folks in certain cities and Tweet them.
“@[username] Hi! Looks like you are an avid user of Android and are from Bangalore. Here is an app that you would find useful”
A majority of the users would actually click the link. We needed to track this click so we ended up using Bit.ly (Found John Borthwick through Twitter). Our initial results were fantastic —
Folks were clicking our Tweets and more importantly downloading the app. More importantly they were recommending it to their friends on Twitter and rating the app high.
The Solution
We needed to automate this. At this point, Twitter had licensed their data to two different companies Gnip & Datasift. We didnt have a lot of money — We were bootstrapped. We ended up picking up Datasift because it seemed to have a querying engine built on top of Twitter stream. Datasift at the time allowed any one to try a few queries for 50 or credits.
The more I dug into Datasift the more excited the prospect was! One could filter the real time based on multiple criteria.
This meant that we could filter tweets from folks who were using a certain Twitter client and the location in the bio was in a certain city! Voila just what we were looking for.
Thankfully John & his team at Bit.ly had just launched bit.ly for custom domains. John was gracious enough to lets us try this. Bit.ly statistics were real-time and would let us see how folks were clicking our links.
All we needed to do now was to code this sucker and try it out. My amazing co-founder Rajat (@urajat) whipped up a quick script in ruby and we were off to the races. We had to tweak the script a couple of times for the following reasons
- Datasift had a few bugs that Nick & his team were working through.
- We had to play nice within the limits of the twitter API.
Exciting Results
Eventually, we worked out our kinks and the results were quite fruitful.
- 8 out 10 people clicked
- 2 out 10 people RT’ed (some of these folks were influential)
- 1 in 25 complained of us spamming them (In some cases this was a manual error as I was doing this by hand early on)
What started as a simple experiment was quite valuable to our business (regardless how it ended). Our initial loyal user base was built using these tools and built quickly.
I believe @nihalmehta eventually built something which turned out to be a LocalResponse. We didnt have the foresight to pivot from Localbeat to built this.
Twitter was invaluable as business development tool for me, we were able to connect with folks at Datasift & Bit.ly sitting in Bangalore.
Sincere thanks to @Borthwick & @nik and who provided us with the tools to put this together. It was one of the fun things that we did at Localbeat.
