What I Learned From Launching the Messenger Developer Platform

I was lucky to be a part of the Messenger team in the early days. As product manager, I worked on the Messenger Developer Platform, integrations with key partners (such as Uber, Lyft, KLM), and the transition from Facebook to a standalone Messenger app. It was an insane period of hypergrowth — we went from 200 million users to 1 billion users in 2 years. This is what I learned from helping drive the product for the developer platform during that time.

Why focus on engaging developers, and not just getting more people on the platform?

At Messenger, even with our impressive scale, we always lived in fear of irrelevance. We believed our product was a better way to communicate, but in a world with pre-installed SMS, we needed to give people a great reason to use Messenger instead. Similarly, in late 2014, there were a lot of great apps that took off, but each of them suffered from the same issue: it was extremely hard to get people to discover them, and even harder to build a habit of using them repeatedly.

By looking at the data, we were starting to see hints of people using other apps to share pictures and videos with their friends on Messenger, and over 1 billion messages were being sent every month to businesses on Facebook (i.e. Facebook pages). So, we embarked on the Silicon Valley right of passage: We created a platform. We had the opportunity to give developers and businesses access to the people using Messenger, while, in turn, providing useful tools and services to make Messenger more engaging for people.

Below I’ve shared 5 lessons I learned from launching the Messenger Developer Platform — all battle-tested, with the scars to prove it!

Start specific and expand

For any experience to be great, it requires an immense amount of detail, which is hard to achieve if you start with generic components. While the tools for Messenger are very open and flexible (to allow for experimentation in the early days of the ecosystem), the best experiences came from 1–2 use cases that the platform was uniquely suited to solve well. For example, we worked with KLM to build an end-to-end integration where people could seamlessly receive their boarding pass, flight updates and customer service directly through Messenger. This was a vastly better experience than searching your email for a 6-digit confirmation code and waiting on hold. But, this level of integration required the product and engineering team to fly to Amsterdam (twice), adapt our working style to work with a European airline in a time zone 9 hours ahead, and become (semi) experts on boarding pass layouts. On the surface, this seems like a large investment to make. But, there was nothing more valuable than sitting with the individual business leaders at KLM, learning what’s important to them, and understanding what it takes to get things done. Putting a “simple” Messenger button on a checkout page is, of course, not simple. With learnings from KLM, we were able to improve the platform for everyone, and provide a sample use case for others to follow.

Focus on growth

If a bot speaks, and no one is around to hear it, does it make a sound? Even with 1 billion users on Messenger, establishing it as a new habit with users is hard. Our first iteration of the Platform in 2015 required people to download “apps for Messenger”, to create content, and then share back into Messenger. It turned out this interrupted the natural flow of the conversation too much. From this, we found better ways to bring this content closer to people, including launching the “GIF” tab, powered by Giphy and other 3rd party apps, built right into Messenger. In 2016, bots were built right into Messenger, but still, the most successful experiences were distributed through existing channels. For example, Rogers (one of the largest telecom companies in Canada), used Messenger for their customer service, and simply replaced the “contact us” button on their website with “Messenger”. People didn’t have to think about using Messenger, but when they did, their NPS scores increased significantly. One of the key things holding the platform back was the limited organic distribution in its early days. Now, with the recent launch of the bot store, I’m excited to see where it takes brands and developers.

Pay attention to incentives

From the perspective of both building — and building on top of — a platform, get this right (and understand it), and everything else will take care of itself. Platforms will naturally gravitate towards their incentives. When structuring a new system, it’s imperative to deeply understand the incentives of all three parties — developers/businesses, consumers, and the platform itself — and design a system that maximizes value.

Platforms are nothing more than a trusted contract between 2+ parties to allow for specialization and shared resources. Any successful platform requires trust, and trust is best ensured through aligned incentives. When you’re building a platform, a huge amount of work will go into understanding the incentives and scenarios (both success and failure), and designing a lasting system, including product + policy + enforcement, that optimizes for the best outcome. Here are my three takeaways on structuring that system:

1. Have a shared North Star. For Messenger, our North Star was people having great experiences on Messenger with businesses & apps. This is a North Star that benefits everyone, and always wins/ends arguments. On the margin, there are policies and rules that would benefit one party more than another — but nothing can sacrifice that shared end goal.
2. Be Transparent. If everyone understands what you’re doing, how you get paid, and the rules of the road, everyone is happier. Specifically, businesses feel more comfortable investing resources on your platform if they clearly understand the rules of the road. Of course, in the early days of a platform it’s not always possible to define everything up front — some things need to evolve naturally. To counter some ambiguity in the early stages of a platform, investing in the community through blog posts, groups, events etc. can go along way to build trust.
3. Give, don’t take. To kick-start an ecosystem, platforms typically need to give away a lot of value to everyone involved. As the ecosystem matures, and more businesses / developers participate, rules often become more strict to allow for scale without harming the user experience. If a change is perceived as negative, this can have a huge negative impact on the perception of trust and transparency. So, in the early days of a platform, it’s important to (a) define what’s temporary vs. permanent and (b) where possible, start conservative and expand.

Culture of platform ! = consumer

At F8 in 2014, Mark famously changed the mantra “Move Fast and Break Things” to “Move Fast with Stable Infra”. This, more than anything, highlights the difference in culture between building consumer products and building platform products. With consumer products, you can try things quickly, if they don’t work, you learn, change, and try again. Messenger releases a new app every week, giving the team flexibility to take bets, build tools quickly for people, and fix mistakes. With Platforms, people are building their businesses on top of the decisions that are made — businesses require consistency, stability, and visibility multiple years in the future. This requires a much more deliberate process of building products.

Be your own customer

One of the most fun (and productive) parts about building the Messenger platform was our internal hackathons where we built our own apps for Messenger (2015) and Bots for Messenger (2016). In 2015, an internal team built “Sound Clips for Messenger”, which was a fun way to send “audio emojis”, created in a 3rd party app, and shared through Messenger. The sound design team at Facebook helped us create amazing sounds (including recording their own babies laughing and crying), and we built a simple app for people to share them on Messenger. This little side project not only taught us what it’s like to build on top of the platform, uncovering challenges like providing easy ways to test, but also was downloaded by millions of people, which was an awesome experience for the team. Similarly, in 2016, we all did an offsite and built our own bots — releasing creative energy and giving ourselves extremely valuable feedback. I was excited to see that the team continued that tradition at F8 this year and is planning to open source one of the first chat extensions in a group conversation.

I’d love to hear your feedback on this post, as well as your experiences in building platforms. In my new role at Greylock, if you are building a consumer tech company that involves platforms and messaging, I’d love to meet you (m.me/sethrosenberg or seth [at] greylock [dot] com).