A Dozen Ways to Order a Coffee | Shane Mac | Messaging Summit 2019

So my name's Shane Mac. I was a CEO of a company called Assist for about 5 years, and then we just got acquired by a company called Converse Social. So now my title is called Chief Automation Officer. I don't know what the fuck it means either. So we'll start there. The thing I want to talk about today is just 2 ideas. So it might be a little complex, I'm not sure, but if you're building in the conversational messaging space specifically around automation, automation, it can get pretty complex where it looks pretty simple. So I'm gonna show you two concepts that we really worked on for the last 3 years and kind of go through how we thought about it, but really where it all kind of messed up. And I hope that you won't go down the same path we did. We were the first company to launch on Facebook Messenger 3 years ago when it opened up with Zuckerberg's keynote for 1-800-Flowers. And then this, Yesterday we were, Sephora was on stage, and the use case I'm gonna show you is actually the stats came out yesterday, so they're updated, but the goal with Sephora was can we increase booking reservations on Facebook Messenger? And this was 2 years ago, as of yesterday it was 60% over their website, so now all the traffic is going to Messenger, not web and apps. The thing we've done though is how do we make a flow or an experience that's great that isn't a tree flow. So when you see here is it looks like you have 4 steps going through Messenger, but what we call random access navigation is the way we think about building automated conversations, which is the anti of a tree flow or like an IVR static flow, and we think of it really more about the ability for the consumer to go down an undefined path and change their mind at any time. Like the reason bots suck is 'cause when you change your mind at the 7th step 'cause you wanna change something from the 2nd step, it says I don't understand 'cause it didn't remember and then everything goes to hell and then it really sucks and you're like bots are terrible. That's about the last like 3 years of bots. But if you think about it like this, it's a little better. So you think of it as a circle instead of a line. Instead of A, B, C, D, E, F, G, think of it as a circle with a goal. The goal is book a class. You go through a little flow and it's like San Francisco. And then it's saying, okay, select the store type. So now we know it's Powell Station. And then it's saying, what type of class do you need? But the difference here is the customer's in control. So if they said, no Union Street, you need to be able to actually understand it's a location, update that intent or that entity, move forward while changing your mind, not go backwards. And it's a big design philosophy that we think actually drives a lot of great experiences. Because if you're building tree flows and scripts, no one wants them. So you end up with a bunch of bots that are chatty that no one wants to use, and then everyone says that bots really suck. And then if you go even further, you're to the point where one query, like, "Never mind, you have a personal one-on-one on Pal on Friday." Once you get to this level, you're able to do experiences that in one step can change everything. and you go from an 11-step flow to a 1-step flow. And the stat was on stage yesterday at F8, I think it's 60% of traffic that goes to Messenger converts better than the web, and 45% of those people actually, higher than web, show up to the stores. So the traffic and the conversion to in-store is a lot better. And we think of it more like a subway system. Some people want to go down every step, go 1, 2, 3, 4, 5. Some people want to go 1, 2, 3, back to 1, and then 4. Some people can say a query that can get them to the end of the funnel in one step. And so designing it in this way, we call random access navigation, and it's a circle, not a line. And I think it's a really important thing if you want to get into automation done well, and if you're going to get into the messaging space, uh, not building bad scripty tree flows. If you're gonna have automation, is a pretty powerful thing. The next one I'll skip 'cause I only have 7 minutes, so I'm trying to go fast. They were like, "We're really fucking behind, run." I was like, "All right." I got a handheld mic. They're like, "You don't have time for a lav." I'm kidding, I asked for it. So in 2018 July, we were at the Phillies Stadium And we decided that no one wants to download an app. Apple Business Chat had just opened up. And as one of the first partners of that, we were like, can you order beer in your seat? No one downloads the app every night, ever. It's like one order a night. And we put a QR code on the back of every single seat in the stadium. And very simply, just by scanning this— let's see if the video plays. You scan the code. This is inside of iMessage. So inside of your core messaging app on iPhone, they have a menu. You can pick a Bud Light. You can also say 2 Bud Lights and a Miller, 2 Yuenglings if you're in Philly. Type in your section. For certain seats, your section number was already built in, so you didn't have to do that. You could leave a tip, and then Apple Pay is actually built in. This was the moment when I actually— what we believed 5 years ago when we started the company, that I was like, this is the future. No passwords, no login, it's all encrypted, payments on file, and in 2 steps you could order a beer and someone's delivering your beer without payment. And we're talking about like the number of orders per section, I can't share the official number, but Magnitudes X's higher than ever before. And I think when you see something like this, it's the moments when you see— it's like payments, automation technology, language, all these things, and then Apple and Facebook, Google, Instagram DMs all opening up to allow you to do experiences that weren't possible before. The biggest piece about designing conversations for automation is to focus on the errors. And this is another piece that we spend a lot of time Because in the web and app world, you basically— or social world— you basically design something that's A, B, C, D, E, F, and it's like a flowchart with UX designers. Everyone's arguing, they're like, "My green button, it's below the fold," and everyone gets kind of like very aggressive about having an opinion. Whereas this space, the best part is that you can just sit back and learn from your customers what the errors are. So when it doesn't know what it's talking about, if you analyze the errors, you can actually get the intelligence. So an example would be two types of errors: understood not supported and misunderstood. Understood not supported in the context of a baseball game would be, "Do you sell jerseys?" Even though we didn't sell jerseys in the bot, we were able to design the language to say, "Hey, we hear you. We don't sell jerseys, but go get them at Section 14." And by designing errors in a way that are helpful, and making them— so they acknowledge that you heard the person actually makes people not hate the air. And then another example is, "Can you get cotton candy for my kids?" "No, but you can get your kid a beer." And at the top right you'll see we're logging every air. So at the end of the day, you're just collecting information on what people want. And in the past you didn't have machine-readable text from the consumer. You just had pixels and charts and heat maps and all this shit that like showed fallout charts and no one knew what it meant. So not understood is when you don't understand at all and you need to have some empathy and really tell them that you can't do that yet, but you hear them and you'll follow up when you can. And then at the end of the day, you end up with something that looks like this. There's no more arguing in like a boardroom or a product room or a strategy session on what you think we should do next. We should probably sell jerseys, then sell some hot dogs, Add some parking tickets. I don't know what end time means. Oh, it means when do you stop selling beer because it's the 7th inning of baseball. And then cotton candy, I guess. So expand the food types. And this is kind of the roadmap. And by using and leaning into the errors in messaging, you actually get to understand what everyone wants. And it seems like one of the most powerful platforms that If you think about it this way, don't be scared of it. Don't launch with some big thing. Launch with like something that just sells beer. And if you launch and you nail that use case, the CSAT on that experience was 4.8 out of 5. Everyone loved it. It was so easy. Every person that was the 0.2 just wanted some more. And then you use that to say, hey, we'll get you cotton candy, we'll sell some peanuts, get some jerseys, and do some hot dogs. And it's an infinite feedback loop forever. And then make sure that people can change their mind at any time and think of the design like a circle and all the things you need to get done to finish the goal, not like a straight line. That's it.