January 8th, 2019, as I walked into my office, one of the backend engineers told me that some of our servers are down and AppBrowzer App isn’t working. He wasn’t able to figure out the issue. I told him to check the logs in every server and see if he finds anything.
I sat on my desk thinking that this could be one of those crypto-jacking attacks that we had faced a few months ago. To rule that out, I logged into one of our main servers and checked. Thankfully, it wasn’t the case.
In my inbox, there was a flood of emails from our users with the screenshot of our App showing “Unable to connect to server”.
On investigating further, I noticed that we were getting close to 500 requests per second on our web server — which was a lot. I checked if all those requests are coming from same the IP address or a set of IP addresses to rule out any other type of denial of service (DoS) attack. It wasn’t the case.
40,000 Signups in less than 24 hours
Then I opened up Firebase console, and was surprised to see 7000 active users in the last 30 minutes!
I realized that our GPay like scratch cards based referral program that we had released(silently) two weeks ago has gone viral. I was amazed and worried at the same time as we were still testing the feature out and many bugs had to be fixed!
That day we got more than 40,000 signups, which was 10x of our normal daily sign-ups, and then it exploded the next day giving us 100,000 signups!
Sleepless Nights in our office
It took us four sleepless nights to make our backend capable to handle such massive traffic.
Hundreds of lines of code were rewritten, bugs were identified and fixed, servers were upgraded, and much more.
Below are a few pictures that were taken during that time. Everyone in the company contributed by going out of their way.
Journey to our first 300k users
Before we get into how we went from 300k to 1.5 million in 21 days, you should know the background story of our first 300k users.
In case you don’t know what’s AppBrowzer — it’s an all-in-one app where you can shop, recharge, book flight/cab, order food, browse the web, play games, read news — yes, all this in one app.
This concept was new in 2016. However, now you can see GPay, PhonePe, PayTM, Flipkart all morphing into an all-in-one (Super) app. Read more on that here.
Sunny and me, both knew very little about marketing when we started up. Sunny finished his MBA before Facebook even existed(he repeats this joke once a month, so I had to put it here).
This led to a lot of mistakes in marketing our app in the initial days. We hired the wrong people, spent a lot of money on Facebook, and Google Ads, hired useless marketing agencies, promoted on ad networks.
This went on for a year or so and we got 200k downloads(not users). There were many loyal users however we thought it wasn’t big enough. This was one of the biggest mistakes that we did – even 100 loyal users are good enough for you to start studying them and build for the product-market fit.
The decision to Launch a Referral Program
Towards the end of 2017, I decided to get into marketing. I watched videos, read books, case studies, blogs, joined courses — basically whatever I could.
I read about the importance of product-market fit, ASO, CAC, LTV, Retention Cohorts, and hell a lot of other concepts.
During this time our marketing team consisted of an intern and me — just the two of us.
Our goal was to acquire as many loyal users as possible within our budget. We realized that tapping on our existing user’s network was one of the best channels to acquire potentially loyal users.
We decided to launch a referral program and reward users for promoting the app to their friends and family. I searched for plug-and-play referral program tools to get up and running as soon as possible.
However, none could satisfy even half of our requirements. Also, these tools were very costly.
In the end, we had to develop it ourselves. In April 2018, after 1.5 months of planning and development, we launched our first ever referral program.
The Mistakes in the first version of AppBrowzer Referral Program
On 27th April 2018, we started the referral program as “contest of the month” which would end on 6th May 2018.
Users had 10 days to refer their friends and earn maximum points. We gave 50 points each to the referrer(the user who is inviting) and referee (new user) on successful referral.
Each point was worth ₹0.1, so we were giving away ₹5 to each party.
We had also created a leaderboard for ranking users based on their total points. Check out Mobile App Retention campaigns the screenshot above.
The result of this referral program was very disappointing. We got around 6–7k referral registrations through the contest.
Let’s look at the mistakes in the referral program that could have been easily avoided.
Mistake #1 — Poor Visibility
There was no mention of the referral program on the home screen. It was hidden in the navigation drawer with other menu options (screenshot below)
A referral program has to be visible for people to know that it exists. AppBrowzer Referral Program had poor visibility and hence poor participation.
I had earlier written on how Dropbox increased the visibility of their referral program by including it in their onboarding.
Mistake #2 — No explanation on what’s a successful referral
We assumed that users understand that their referrals had to confirm their OTP for them to win the referral reward.
A clear explanation of the referral process is a must for users to feel comfortable before they decide to refer.
Mistake #3 — No Referral Tracking Dashboard
Also, we didn’t have a referral tracking dashboard. I had already discussed here the importance of users being able to track their referrals.
It creates transparency for the users that their referrals are being recorded and also helps in tracking the status of each referral.
Mistake #4 — Useless Leaderboard
The leaderboard that we had provided little or no motivation for users to rank higher. This was gamification done wrong.
It’s important to understand that creating a good gamified experience is much more than simply implementing points and leaderboards.
Majority of the users who participated referred just 5 people to earn the minimum redeemable 250 points and stopped.
Mistake #5 — No fraud detection mechanism
You will always get users on the Internet who would try to find loopholes and exploit your system, especially when cash is involved.
We didn’t have any mechanism to detect these fraudsters. These were people who knew the way to reset GAID(unique id) that we were using to detect if the user is using the same device to sign up with multiple numbers.
I found this only after looking at the mismatch between download numbers in Playstore and total registrations recorded during the contest period. There were approximately 20% fewer downloads than registrations, which was a lot.
We also made a lot of smaller mistakes. Most of them and the mistakes mentioned above could have been easily avoided with thorough research. However, it’s easier said than done. There is a lot going on when you’re running a startup or working in a startup.
Reaching 300k and Dogfooding
Over the next 5 months, we kept on experimenting with changes in the AppBrowzer Referral Program and also different acquisition channels acquiring additional 100k downloads (Total 300k) before deciding for a complete overhaul of our referral program and rewards system.
That one decision marks the point when the seed of our latest product FLYY was planted. It was an internal product for AppBrowzer throughout 2019 before we decided to launch it as a separate product this year in March. We have incorporated all our learnings about the referral program, variable rewards, gamification in FLYY.
You can use FLYY’s acquisition module to get up and running with scratch cards based referral program in less than a week.
The discussion to overhaul our rewards system started in September 2019 when we were discussing how Tez (now GPay) used scratch cards based rewards system to grow from 0 to 55 million downloads in less than a year.
I read several books and materials on user psychology & gamification for more than a month while planning our new referral and rewards system.
In the next post, I will write about the planning process, how our referral program had gone viral taking AppBrowzer from 300k to 1.5 million users in just 21 days.