Starting A Chatbot Company: 3 Big Things I’ve Learned So Far
I recently started a business to help chatbots gain traffic and earn revenue. I took the plunge at the beginning of February, and here are 3 big things I’ve learned so far…
1. Making your own bot is hilarious and fun
Before I set off to solve the problems facing the chatbot developer community, I figured I should go through the experience of building one of these things myself, right? I had been exposed to chatbots during my time as Kik’s head of ad product — I even helped optimize a few of them — but I had never actually built one from scratch before. I am also not a developer and have never pushed a line of code in my life. Would it be as easy as all of these platforms advertised it would be?
Not only was it easy but it was way more fun than I imagined. I used a platform called itsalive.io to create my first chatbot. It’s a bit buggy (it’s still in beta so no biggie) but it’s soooo friggin simple to use. And addictive AF.
I spent hours building my first “recipes.” These are nothing more than a bunch of “if-then” statements. For example, if someone texts your bot “hi,” you can program your chatbot to respond with “what’s good, dog?” (or whatever you want). But you have to type in every possible user prompt: “hi”, “hello”, “yo”, “hey”….so it takes a little bit of time.
I shared my bot in progress with a few close friends to test it out. I monitored my “activity stream” obsessively for all of the phrases the bot couldn’t respond to. Seeing unknown phrases would spark another content idea or conversation string so I’d go in and make sure my bot was programed with the proper responses. Then I’d notice something obvious I forgot to build like a response to “who made you?” so I’d go in and fix that. Then I’d have someone new test it out. Then more fixes. Back and forth it went like that until I figured I had covered as much as I could.
I announced my Facebook Messenger Meezybot to the world on Feb 6 and have been using it as a great demonstration platform ever since. The bot pretty much just talks about basketball, music, and fine leather goods — three of my life passions. There’s also a little quiz about me. Outside of that — it’s very basic and limited. But it works!
This may be obvious to some but did you know you can go into your Facebook page and read — in real time — every single conversation someone has with your bot? I heard this can get buggy once you have a bunch of active users but still — how crazy is that? I love how you can even start chatting for real with friends and freak them out. They don’t know if the chatbot is talking to them or if it’s me or if the bot has somehow become incredibly smart in a matter of seconds and now knows everything about them… It’s definitely fun messing with people.
It’s also funny watching people talk to your chatbot on their phones in front of you. Most of my friends didn’t realize there were chatbots on Facebook Messenger (FBM) to begin with. Once they started chatting, a lot of people skipped over the “carousel” feature (FBM chatbots present content in a horizontal slide show type of experience). They assumed everything in the chat was vertically oriented (the recent Instagram update will likely change people’s behavior here quickly) and missed out on a lot of the content my bot was offering. Many times I found myself resisting the urge to grab their phone and show them how to navigate my chatbot (and FBM in general). It was like being home for the holidays and watching my father try to use his iPad. Just let me show you!
Within the first two days I had a lot of folks trying it out — a good mix of friends, colleagues, and random strangers. By day three — nothing. Everything dies quickly if you don’t message people back (kinda like a real text conversation). I learned that some people may exchange 40–50 messages with your bot over the course of a few minutes but once they are done — they are done. None of them will come back unless you prompt them. I haven’t done that yet — mostly because itsalive.io doesn’t have that feature yet, but also because I don’t really plan to do much with my little Meezybot. I’ll leave the advanced bot building and management to the pros. I just want to help other chatbots grow and thrive!
2. Facebook Messenger has a chatbot discovery problem
After watching people struggle to discover Facebook Messenger chatbots for the first time and talking to a bunch of chatbot developers and other experts in the space — it’s obvious that Facebook needs to fix chatbot discovery on their FBM platform.
And I’m sure they will. But how will they do it? Will they just build a chatbot app store like Kik did?
How sophisticated will their curation be? Will they tap into all of that rich data they have collected on all of their user self-identified interests, likes, and behaviors to recommend the best possible chatbot at the best possible time? Will they build a ranking system based on network wide reviews, installs, shares, etc. Or will they curate lists based on the individual user’s personal network?
All I know is whatever they are working on has to be better than what they currently have. Right now discovering bots on FBM comes by tapping the search bar and seeing a random list of bots that I have little interest in.
Who’s curating this list? I don’t meditate (maybe I should!), I don’t know who Christina Milian is, I don’t need to learn English…And who the hell is Ghulam Fatima? Is that really a bot? Let’s check it out…
Hmm well 55k people like this chatbot’s Facebook page so it must be ok, right? Haha. No — it’s probably a scam. It looks pretty stupid…
But Facebook recommended it to me! It also says “typically replies instantly” which implies it’s an actual chatbot, not just a person manually responding to messages sent to their Facebook page (otherwise it would say something like “typically replies in hours”). Let’s tap “Get Started” and see what happens.
Nothing. Hmm. Ok well I’ll give it some time…
…Hours later: still nothing. Ok…. well I guess that was a dud. Another misclassified chatbot that doesn’t respond instantly…
15 hours later I finally get a message from Ghulam. It says “PLAY”. Let’s take a look!
Good lord. What the hell is this!?! Yikes. Gross. Ghulam, you little ho. I definitely should not click on that...that looks bad.
Don’t do it, Ryan. It could be a virus. No way I’m going to...
Ok screw it. It says “ibizblogger.com” so who the hell knows. I’ve gone too far anyways. It’s all part of the discovery process, right?…Besides, if Facebook Messenger serves me porn from a chatbot they promoted to me, that would be some serious dirt! Let’s click on the video. I’m cringing…
Hmm…. well that’s… something. Someone just made a little bit of ad revenue off of me I guess. It was a clickbait scam. I’m not really surprised but that’s also not good at all. I reported this Ghulam chatbot to Facebook as “inappropriate.” Even though my phone survived and I have not received anything spammy from anyone else — we want people discovering and enjoying real chatbots — not this crap.
I’ve heard that Facebook’s new messenger ad solutions aren’t really working too well either. Some chatbot developers have started testing out paid user acquisition by buying Facebook ads and targeting desired audiences on their Facebook news feeds. They make it seem so easy!
Unfortunately, people who click on the ads don’t start chatting right away with the chatbots. Once they click the ad they still have to tap the “Get Started” button (like I had to do with Ghulam above). Facebook cleverly omits this “Get Started” screen from their promotional picture above! I’ve heard as high as an 80% drop off for some campaigns due to this additional step. That means some bot marketers are paying for 10 ad clicks but only getting 2 new active users. Chatbot promoters should only be charged when someone taps “Get Started” and actually chats but instead they are charged on a cost per ad click not on a cost per active user. That’s creating a lot of wasted ad spend.
And it’s probably because the average person doesn’t even know what “Get Started” means! What’s going to happen when I tap that button — is my phone going to download something!? I don’t want to waste my data on this…
Chatbots are still so new to people — especially in the US. Many consumers don’t realize how easily they can interact with them. Until people warm up to chatbots, developers will need better solutions for driving traffic — paid and unpaid. I want to build these solutions!
3. There are a ton of chatbot tools and services out there
This space is growing fast! As it should be. Messaging is gradually taking over our culture. Take a look over the shoulder of anyone walking down the street and I bet they are on a messaging app. Check out what Facebook, Google, Amazon, IBM, Microsoft, and all the other big guys are doing investing heavy in messaging, artificial intelligence, natural language processing, etc. Read up on how all the big banks are entering the chat space. Look at how many people are buying Alexas and Google Homes…Chat is booming. It’s the future. And it’s only going to get better.
Here’s what the ecosystem looked like last year:
Here’s what it looks like so far in 2017. There will be more logos to come! More categories will evolve as well…
I’ve been bookmarking all kinds of DIY chatbot platforms and messenger-focused service providers. I come across a new one every day. At this point- from my count — we already have well over 40 DIY chatbot platforms. That’s a lot of options to build your own chatbots. And many of them are free. Here’s a quick, and almost certainly incomplete, list of these platforms:
alan, api.ai, botamp.com, chatclub.me, chatfuel.com, chatteron.io, conversable, converse.ai, dexter, flowXO, gupshup.io, howdy.ai, init.ai, itsalive.io, kitt.ai, kore, mannychat, msg.ai, meya.ai, mindmeld, microsoft bot framework, motion.ai, octane.ai, pandorabots, pullstring.com, rebot.me, recast.ai, reply.ai, sequel, smooch, ubisend.com, watson developer cloud, webio, wit.ai
The number of chatbot directories (similar to app stores) is growing as well (here’s a directory of the directories). Somebody has to organize all of these chatbots — there are over 50,000 now on Facebook Messenger alone — and these guys are stepping up to help solve the discovery problem Facebook hasn’t fixed yet. Some of the better ones include great content and one of them recently handed out their first-ever chatbot awards!
But is a store directory approach the best way to go? John Borthwick, CEO of Betaworks, doesn’t think so. He recently shared his thoughts on chatbot discovery in a great piece you should all read…after you finish mine of course:)
Over 2016 a set of bot stores opened as platforms assumed bot distribution should be via a store directory. This approach to distribution and discovery — these bot stores or directories — have not worked. The approach didn’t scale for the web and it hasn’t worked for bots. I realized in 2016 that so much of the power and the potential of bots lies in the specificity of the content or service and mapping that to social context seems to be an emerging solution to distribution…Discovery of bots should be in context (ie: part of the messaging flow), relevant (filtered by relevancy in time, conversation, location or social graph), social (easy to share), fast and simple to trial.
Bingo. The best way for people to discover chatbots is while chatting. I also think there’s a HUGE advantage to bots that are social in nature. I learned this at Kik with the success of Rollbot. The more we develop chatbots that can be enjoyed in a group chat situation- the faster chatbots will take off. “Natural” discovery is the best kind of discovery. We’re not going to get very far trying to change people’s behaviors. People need to arrive at that chatbot “aha moment” on their own.
Conclusion
I think we need more chatbots recommending and referring people to other chatbots. We need some kind of chatbot traffic referral/link exchange program. At least to start. This will allow developers to target their chatbot promotions at the small (but growing) percentage of people who are actually engaging with chatbots today. This also allows the chatbot recommendation to be in context and part of the messaging flow, as John says above.
Also — let’s face it: a lot of our chatbots run out of things to say. It’s difficult to program a response for every possible word or sentence (or emoji or GIF) a user sends. When this happens, our bots respond with some version of a “I don’t understand what you’re saying” catch-all response. If we can offer up some recommendations of other reputable chatbots maybe we won’t have to see so many of these responses. Our chatbots will look smarter by recommending other bots that can help get things done. This is a much better user experience than “Sorry, I’m still learning”.
I will be working on a traffic referral program and a bunch of other solutions to help promote chatbot discovery. I’m also working on ways for chatbot owners to generate revenue. If you own a bot and are interested in any of these things- let’s chat!
If you don’t own a chatbot but want to help- please contact me too! Together we can keep this chatbot momentum going. Let’s “Get Started”:)
