Building a startup within a startup

The more successful you become, the harder it is to apply a scrappy startup mindset to your product development.

An experienced team, a generous budget, a captive audience excited about your next release – none of these things guarantee the success of your next product and thinking they do can lead you to fall into a dangerous trap.

At the Dublin stop of our Inside Intercom World Tour, I discussed some lessons learned while developing a new product, and how our perceived advantages were what held us back.
Prefer a written account? You’ll find a written version of my talk below.

My story is about building a startup within a startup. The startup we’re talking about is Intercom, and the new startup we built within it is our product Acquire. Acquire lets you chat with visitors to your website in real-time and helps you turn those visitors into customers.

It was a big change for Intercom. Before Acquire, you could only talk to signed-in users of your app or website. When we introduced Acquire, we made it possible to talk to anyone that came to your site.

It was also quite a big change for me. When I joined Intercom two years ago, I had just finished college and had made a big career change. I had been a chef for about 8 or 9 years and was moving into a new job as a software engineer. Working on this project, just a few months out of college, was a daunting task.

The ingredients for success

We were feeling pretty confident and optimistic though. We already had a successful product. We had money, we had people with experience and we had a captive audience. We had 10,000 paying customers in over 90 countries and we knew many of them had been asking about extending Intercom’s capabilities.

In October 2015 we formally released Acquire after running a beta progam. It was a huge success.

In the first few months, our revenue was over $1 million. To put that in context, it took Intercom two years to reach that revenue milestone. I thought we’d nailed it.

Tackling churn

There was an issue though. Customers were leaving Acquire at about double the rate they were leaving our other products. Churn like this signals your product is going to die. It might not be tomorrow or the next day, but if you don’t address churn, it will at some point in the future.

We thought we knew how to fix things. Past experience had taught us that we could go back, research, talk to our customers and get a clear picture of what the problems were and fix them.

Acquire allows visitors to initiate a conversation from your website. The biggest problem we tackled was getting replies to visitors if they didn’t hang around for a response. If someone comes to your website, starts a conversation, doesn’t leave their email address and then moves on, you can’t reply to them.

We spent three months building a bot and it did absolutely nothing

We developed a solution in the form of a helpful email collector bot. When you first see it, it simply looks like another comment. It was conversational and it fitted with our ethos of making business personal.

But there was a lot happening under the hood. We had to consider tone of voice and conversation flow. We had to integrate it with our messenger and write a lot of complex server-side code in order to manage conversations and know when to start and end them (it can be pretty boring to get stuck in an endless loop with a bot).

With the product challenges, the engineering challenges and the high bar we set for ourselves when we’re building products, it took about three months for us to build this feature.

When we launched it, we were expecting to have emails coming in to our customers left, right and center. We thought our customers wouldn’t be able to deal with all the additional conversations they were having. But it had absolutely no effect. None at all. It was really disheartening.

A small change

At this point, the team took a little bit of time to make some visual tweaks to the email collector.

It was pretty small visual change, and it took about three hours. The performance gain we got from just this little change? 25%.

We had spent three months on a bot, and got absolutely no change. We spent three hours on some styling, and saw a 25% improvement.

Why didn’t we do this first? If we had been an early stage startup, we wouldn’t have had the luxury of spending two or three months building something and not getting any results. We would have needed to be pragmatic and move quickly.

You need to starve your new products

All our perceived advantages – products that were already successful, great people, money, experience, a captive audience – weren’t advantages at all. The constraints you get early on had been removed for us. We didn’t have a startup mindset. That’s what you need when you’re building a new product.

After this realisation we became much more vigilant when scoping features and projects to make sure we were building the simplest version of what we needed. We also put more emphasis on research and user testing in order to validate our assumptions.

I learned that you need to starve your new products. Being hungry makes you scrappy. Being scrappy makes you build better products quicker.