How Braze forges timely consumer engagement at scale

Consumer brands want to give their customers highly personalized experiences using today's connected digital technologies. One of the specialist providers helping them do that at scale is Braze, which recently rebranded from its original name Appboy.

The company was founded back in 2011, after co-founders Jonathan Hyman and Bill Magnuson won that year's TechCrunch Disrupt hackathon contest in New York city. Braze is a pioneer of real-time, personalized messaging that help brands build engagement with consumers. This depends on a highly responsive infrastructure that's able to send each individual the right message at the right time, scaled up to serve tens of billions of messages to around a billion active contacts every month.

I met Hyman, who is CTO, during New Relic's FutureStack conference in New York this September, with the aim of finding out more about how it manages that high-volume infrastructure. But I was also intrigued to learn more about Braze's role at the intersection of marketing and product management in the companies it works with.

Mobile-first, cross-channel

The focus has broadened out since the early days of Appboy, which started out delivering messages to mobile apps. It's still a mobile-first world, Hyman tells me, but nowadays Braze delivers messages over email or web messaging as much as via push notifications or in-app — even email counts as a mobile channel, as most consumers open emails on their mobile device.

When I say mobile first I don't necessarily just mean mobile apps. I think of mobile as really more of a paradigm where customers can be engaging with their brands on their terms — I can interact with the brand in the store or on my iPhone or my iPad or on the Web site.

But ultimately consumers have high expectations that the content and messages they're going to get back are going to be highly personalized and relevant to them. And so that's the world that we're in. We help our brands actually achieve that ...

Essentially what we do is, we use technology to humanize a communication or connection between a brand and a customer at scale.

Braze has to straddle all of those different channels, although Hyman doesn't use the term omni-channel or multi-channel.

We like to say cross-channel, because consumers are operating on different platforms and channels. We help our customers build cohesive messaging campaigns that are on different mediums, on different channels. An email campaign also has web messaging and has push notifications or in-app messaging there. So we allow them to orchestrate all that together in one campaign.

Timely engagement

Infrastructure is important for Braze not only to support high volumes across these various channels, but also because timeliness is crucial. Hyman gives the example of pizza chain Domino's Pizza Tracker service. Braze powers the notifications that tell the customer when their pizza has gone in the oven, when it's boxed for delivery, and so on.

That's extremely personal, but the relevance there is the timeliness of it. If you get the push notification that it's in the oven when it got to you, that's a wasted, unmagical experience.

This kind of messaging blurs the line between pure marketing and the product experience itself. So while it may be driven by the marketing team, in today's world they're often working closely with product managers, says Hyman.

Is that a marketing campaign and marketing promotional material? Or is it part of the product experience? As a consumer I personally don't care because I'm just looking to have a great experience of the brand and that's what they delivered to me.

Or you make a purchase and then [the supplier] sends you an email about it a few hours later. Again is that marketing or is that part of the product experience? You can have it all be controlled by the marketing team, then they can start to optimize for conversions, and now you get to drive more product usage together.

Real-time operations

In these interactions, it's important for Braze to operate as near to real-time as possible. "It's one of the main things we focus on," says Hyman.

As event data comes in, that may trigger a message — we might want to send an e-mail to someone after they make a purchase thanking them for it. You want that email to be timely.

Or you could have events that come in that cancel messages. How many times have you gotten an e-mail about a product that you bought yesterday, but the e-mail was for you to buy the product? You're like, 'I just bought this yesterday.' That's a bad customer experience.

Keeping the infrastructure in peak condition is where New Relic comes in. The main uses are performance monitoring and incident handling, as well as testing the impact of anything from large-scale architectural changes to smaller bug fixes.

If you're doing anything that's performance-oriented, you need to have your performance data and be able to benchmark it, such that you can ensure that if you make changes to it, you're actually getting improvements.

New Relic allows us to have the data that we need in order to get performance insights, do incident handling, diagnose issues and challenges, and just get an at-a-glance, inside sense of what's going on with our product.

Performing at scale

The monitoring software also plays a part in automating infrastructure management so that Braze isn't caught short by sudden spikes in activity, for example when a customer unexpectedly starts running a campaign. This aspect of infrastructure management is a serverless process, although the bulk of Braze's operations are very much server based. "We scale up a lot of boxes," says Hyman.

It's an Amazon AWS Lambda process that every couple of minutes checks New Relic, checks our own internal monitoring that we have, and graphs around what do we think's going to happen in the future. So you get the current state from New Relic, the future anticipated state from us. It does some math on it and either adds servers or removes servers. That way we can be fully responsive.

We don't need to manually add anything anymore. It just can handle it itself. If it's two o'clock in the morning and one of our customers in Asia sends a large messaging campaign and we happen to have fewer servers up because we don't have as much US traffic at the time, it can just add it and then process it down.

Despite the scale of its operations, Braze is a fairly close-knit team, with a few dozen people in the engineering organization and around eight on the product side, handling product management and UI/UX design. A DevOps team runs the infrastructure, while other engineers build various parts of the stack.

They interface with DevOps almost like a service or partnership, depending on what team that they're on. Some teams may need to have a tight alliance with DevOps because it's just very intricate to what they're doing. Others can just engage with them kind of as a service like, 'Hey DevOps, I need this technology built.'

DevOps is central

Containerization and microservices are becoming a bigger part of the infrastructure. As that trend continues, each of the engineering teams will bring on an SRE (site reliability engineer) who can help with performance and optimization. But having a central DevOps function remains crucial for maintaining consistent quality of service, says Hyman.

For us it's important that we want to be incredibly rock solid in our uptime and reliability, given the customers that we have. Also it helps us maintain a tight grip on things like security and data privacy which are extremely important in our industry.

You don't end up having this team over here building their own stack and this team over here building their own stack — and they may not be talking to each other, or they may not be using best practices. We want to keep it there in centralized fashion, but we do want to have SREs that can make sure that we optimize in each of the groups.

We're looking at them more as people who are like embedded operations staff that may not necessarily be building infrastructure but are the ones that are performance-minded and focused on optimization, and making sure that what they're doing is going to play well with the rest of our stuff.

Machine learning and AI

Both Braze and its customers are increasingly taking advantage of machine learning and AI to augment the marketing automation toolbox. It's important to apply these capabilities where they can have the most impact, says Hyman.

At Braze we think that there are things that machines are good at and there are things that people are good at. There's also going to be a set of things that we can be great at for all of our customers, versus the things that our customers can be great at themselves.

Our intelligence suite lets you target better, lets you get answers faster and lets you figure out how to give the strongest brand experience in a great way. Some of the other things, we make sure we can we plug really well into content recommendation, if our customers have their own content recommendation engines.

We love that because we're all about building great experiences. Personalization in my mind just kind of a proxy to adding value. I think the reason it's great to get content that is tailored to me is because I'm going to find value in it. So if you send me emails with things that I'm likely to get to be interested in, that's valuable to me. So we want our customers to be thinking in that sense and to work with their recommendation engines.

My take

The Braze story is a powerful example of how strongly technology and marketing are intertwined today, and how both are becoming an integral part of the product experience, even in very physical businesses like takeaway pizza.