How I built an engaging chatbot that instantly got 1k+ messages (and so can you!). Because you’re reading this I imagine you’re the type of person who believes chatbots have hidden potentials and have tried to make a bot yourself. I love that cause that means you’re my kind of person.
So last month I built a chatbot that I thought could be an interesting way to learn more about well, chatbots.
The bot which I named Mr. Bot (wait till I explain why it’s not called something else) was meant to serve one purpose: Be a conversational knowledge bank that chats with visitors about chatbot frameworks and platforms, advantages, game-changing use cases, the big players, latest developments in the bot ecosystem, you name it!
I did almost nothing to promote the bot (and I don’t have an influencer status to hook my followers into everything I do. Heck, I have zero social presence). No paid ads. No viral shout outs on social networks, not even an organically indexed URL to lead searchers to my bot.
Instead I maintained a low-profile: left a genuinely humble post on four bot groups on Facebook, let Mr. Bot sit in bottr.me’s bot library and waited! (Full disclosure: I work with Bottr but Mr. Bot is a personal side project).
Within the first 5 days, Mr. Bot received 900+ messages with every user spending an average of 4 and a half minutes interacting with it.
I don’t think this could have been possible in such a short span with say an interactive app or a blog post where a user comparatively spends only an average of 15 seconds or less.
Conception of Mr.Bot
Mr. Bot started out as a side experiment- still is. I wanted to build something that bot novices and enthusiasts could use to find quick, instant information.
Speaking from personal experience when I first got interested and involved with the chatbot ecosystem around 7–8 months back, I was looking for all kinds of information online- from the available bot platforms to ‘how to humanise a bot’s conversation style!’ In fact, I must have chatted with more than 300 bots myself!
However, somewhere in the middle of reading and grasping facts and futuristic visions, I realised that all of this amazing information wasn’t as easy to find as I’d imagined. Being all over the place and competing to rank on Google SERPs, interesting perspective and approaches looked lost in the clutter.
No justice, right? ⚖️
That’s when it struck me- wouldn’t it be cool if I could pull all of this amazing content under one roof so that people like me who’re looking to find the right information online can find it easily!
Plus at this stage of development, providing on-demand information and answers to FAQs are low-hanging fruits that chatbots could easily focus on to start off.
And who’s better suited to talk about bots than a bot itself?
What I got right
1. Building the bot’s knowledge bank from rich online resources
Getting straight down to the brass tacks, I had a task at hand: find the most common pain points that Mr.Bot could address. The idea was to pull in what people are actually searching for and I had a starting number in mind: 100.
The 100 questions quest: How I found what people were asking
a) Google Trends
No surprises here and I’m betting I’m not the only one who uses Trends as a go-to starting point when it comes to studying recent search trends a.k.a the major pain points.
So I did a straightforward trends search with the term ‘chatbot’ and subsequently picked queries and topics with the highest search volumes.
The results looked like this:
Instead of going with a keyword research tool, cause hey I wasn’t writing a blog post any more I decided to go to Quora: the one question bank to rule them all.
And Quora fit perfectly into my bot’s conversation plan. Here again, I quickly looked up ‘chatbot’, ‘best chatbots’, ‘create chatbot’ and all the other topics I picked up from Trends.
Results were indispensable to say the least. I found at least 300 questions, from there choosing the top 100 based on follower count was quite simple.
After my rendezvous with Trends and Quora, I had the top 100 most frequently asked questions related to chatbots. In fact, if you’re trying to create an informational bot, you could follow the same steps to uncover actual user pain points!
2. Adding a tinge of life and character
The biggest complaints chatbots get from us humans are usually along the lines of them either being stupid and useless or too robotic. As for the former, there’s no denying that everyone in the AI and bot space is working tirelessly to make bots smarter at understanding human intent(think development in NLP algorithms and neural network functions).
As for chatbots being ‘too robotic’ I knew I didn’t want Mr.Bot to get ruffed up by verbal abuse. I had to give it some flair- a tinge of personality even.
Mr.Bot’s character traits, what I thought would go with its informational self:
- witty bordering on cheeky
- optimistic with a humorous streak
Once that’s decided, naming your bot is just as important. Here’s a list I came up with before deciding on Mr.Bot (I’m open to changing it, so please feel free to put in suggestions):
- Wizard of Bots
- Botfather (taken by Telegram)
- BotGod (let’s face it, the name sounds more powerful than my bot actually is)
- Mr.Bot (inspired by Mr.Robot)
- Botasaur (sounded more like a bot event)
- BotGuru (no explanations there)
Golden Tip. I’ve found that drawing out a basic character sketch comes extremely handy when designing a chatbots’ conversational style. Something to refer to when you trying to give them a consistent personality.
Additionally, I used tons of images, sprinkled a few GIFs here and there too. Oh and don’t even get me started on my use of emojis.
When it comes to adding life to conversation, visual elements go a long way. They add appeal to content and generally stick more than words -exactly what I expected when I was adding those jpgs to Mr. Bot’s content cards.
3. I trained Mr.Bot to tell a story
Although there’s still room for improvement, I’d like to believe I trained Mr.Bot’s responses right to a certain extent.
For starters, I used a healthy mix of short and long responses. Thanks to Bottr’s elegant chat window and simplistic yet rich content cards, training Mr.Bot to tell a visually pleasing story was easy-peasy.
So while on one hand for questions like “what is a chatbot?” I wrote the responses to highlight the image on the other hand for questions like “show me how brands are using bots?” and “how to develop a bot from scratch?” I linked to URLs directly from the cards to lead users to handpicked blog posts and tutorials online (all the amazing content I was talking about earlier).
Further, Bottr’s in-built small talk capabilities helped Mr.Bot’s conversations flow easily without me having to train for every random thing a user might say!
4. I targeted the correct niche audience
In the real world, no matter how cool you think your own product/experiment/chatbot is, other people will generally not care. Unless, they too share the same enthusiasm you do about a particular interest area.
So that’s exactly what I did.
I shared Mr.Bot on four highly targeted facebook groups to be precise- ChatBots, Bots, Chatbot Directory and BOTS.
From there it was fairly obvious that a few people would go check out my bot and even have a conversation with it.
To be honest I was expecting around a 100–150 messages max. You can well imagine my surprise when I saw 500+ in the first 24 hours. That’s when I realise what started of as a brain fart might actually have some potential.
What I should have done much better
1. I could’ve built a bigger knowledge bank
In all fairness I didn’t expect Mr.Bot to receive a 1000+ messages in such a short duration, I was only expecting a 100–200 to start off. As the messages started pouring in, I realised I should have probably trained for a bigger sample set.
The silver lining however was that Mr.Bot maintained a list of all the questions I did not know how to respond to giving me room to train it further.
2. Promoted much more
I only shared Mr.Bot on four Facebook groups. I should have shared it with the intention to make it go viral. I could have gone to countless bot communities, discussion forums and threads. For that matter, I should have leveraged it on social networks. Plus, I should have gotten it optimised and indexed on Google Search from Day 1.
In effect, if I would have done these things Mr.Bot would have presumably received way more traction.
3. Given it more flair and conversational flow
The self-critical user part of myself could have done with more of humour. Plus, even though I did add links to quite a few good reads, I feel like I haven’t done justice to them all, YET!
What I Learnt:
- Chatbots could be a powerful content publishing channel. Thanks to rich micro cards which are both easier to read and write, chatbots could be an interesting way to increase content readability and actually get your stuff read by visitors.
- Chatbots facilitate faster and easier on-boarding. It is way easier to convince us humans to chat with a bot than to download an app/try out a software.
- Chatbots increase user engagement rates. Chatbots keep visitors engaged in conversation longer than landing pages/blog posts thus giving longer average session duration.
- Chatbots give rise to better retention. Well-trained chatbots even with their flaws see substantial amount of returning visitors. In fact, I myself saw 35% of the total visitors returning to chat with Mr.Bot which is better than what I have seen with any of my blog posts.
Now that I’m certain Mr.Bot could have some potential, I have a plan for the future:
- Get Mr.Bot search optimised and indexed.
- Build a bigger knowledge bank of at least a 1000 sample questions.
- Promote the bot well enough to reach everyone who is interested to know or learn about the chatbot space.
- I want to make use of Bottr’s live chat functionality. Although this is a long shot, I’d love to have thought leaders, evangelists, anyone with a unique perspective to host live chat AMA sessions with bot enthusiasts. In essence, be the voice behind Mr. Bot and make its identity truly unanimous!
All in all, “I’m getting there”- Mr.Bot’s words, not mine.
I’m open to ideas on how I could make Mr.Bot smarter or more interactive, please feel free to comment and encourage or discourage me.
If you made it this far/ enjoyed the read/ think Mr. Bot has potential, I would really appreciate a click on the Recommend Button 💜