The Celebrity Chatbot I Almost Brought To Life
And Why I Didn’t!
I spent 8–9 days building a bot, had 10–15 friends talk to it, created 10–15 promotional banners and Facebook covers, wrote a Twitter bot, and then I scrapped it all.
Do I regret it? Not at all.
If you have read any of my past stories on chatbots, you know that it would be an understatement to say chatbots fascinate me.
Well, chatbots and the opportunities they hold for businesses and individuals.
In the past month, I have made half a dozen chatbots, just to understand the different things that were possible with a bot. A huge shout out to the Chatfuel Team (especial thanks to Andrew Yaroshevsky, Dmitry Dumik) for giving me the tools to make that happen. Once I was done with those bots, I deleted all of them; they had fulfilled the prophecy and given meaning to their lives.
I know, I know. What a waste.
Fast Forward To The Second Week Of April
Around the 10th of April, I’m having my morning coffee, reading the news and random articles on my phone when I come across an interesting bit about a recently launched mobile app — 100MB (it has been renamed to “100mb cricket” since).
It was launched by one of the most recognised faces in India — the cricketer, Sachin Tendulkar. Apparently he went on a few TV shows to promote the app.
The first thought in my head — “Yeah. Now this app would have the kind of downloads that poor startups take months to achieve, and thousands of dollars.” And I was right. I don’t remember the exact number of downloads at the time, but here are a few things that I do remember.
- The app had a couple of thousand ratings
- Almost all of them were 5 star
- Every single review I went through (and I did go through maybe a dozen) — had two things in common: A. They had all given the app a 5-star rating, B. None of them had said a single good thing about the app.
Yeah, I am not kidding. Every single review had criticised the app, but the tenor of the criticism was humble, polite — bordering on the paranoia of offending someone.
The most surprising fact of all? I have no trouble in believing all of them were genuine reviews. Actually, I am quite sure of it.
Even today, as I check the app, here are the latest stats (All stats are from play store, haven’t checked the app store):
- 100,000+ downloads
- Average rating 4.8, with 10,480 ratings
- 9,699 of these 10,480 are a 5-star. That’s a whooping 92.5% (if you add 4-stars to this mix, the number shoots up to 96%)
If you are not from India, that last part may have left you baffled. So, this is what you need to understand.
- Cricket is religion, Sachin is a God. I don’t say that. It is etched in the holiest of scriptures.
- People would kill for Sachin if someone so much as looked at him funny.
Coming Back To The App
The app was fairly basic. Think of any celebrity bot. The kind of features you would expect out of them, this app had more or less all of them — whatever was relevant anyway.
Due to some reason the app also had a section dedicated to listening to music, playing games etc! Go figure!
I have no idea what the app makers were thinking. There was absolutely no contextual relevance.
Anyway. I digress.
Back To The Matter At Hand
So, by the time I had finished my coffee, these were some of the things going on in my head.
- People obviously love Sachin. So much so that they wouldn’t give even a single bad review to a bot that’s — pardon my french — utter crap!
- Why not give them a chatbot?
I put down the mug in the sink, sat on my desk, fired up the Chatfuel dashboard and a new bot was formed.
Sachin Bot — The Beginning
There were a few things I was crystal clear about.
The bot needs to do the following things:
- It should cover life of sachin outside of cricket
- It needs to highlight his life during his cricket playing days — the good, the bad, the ugly. In short, mix-bag of emotions. Think nostalgia for the fans.
- Life of Sachin now.
- Facts and trivia on Sachin.
- It should be conversational and not boring.
- When stuck, it shouldn’t say shit like — “I am sorry, I didn’t quite get that.”
Although the version that was ready by the end of day 1 said something along those lines. #Facepalm
I have added some screenshots of the bot conversations and some marketing banners that were created at the bottom of the story. I would have added more, but it has been more than two weeks, and they are buried deep under all this bot debris I have accumulating on my Mac.
The Path I Followed
- Googled “interesting facts and trivia about Sachin”. They went into a pool — each pool having its own block.
- Searched for interesting youtube videos of Sachin. They went into groups of ten, in different blocks — each block having two carousels. Only one carousel would be displayed at first, and once the user clicks on “Show more”, the other carousel would be displayed. The videos could be accessed via the persistent menu as well as by user messages — identified using bunch of keywords that went into the AI setup.
- Downloaded a bunch of mobile and desktop wallpapers for Sachin. Once again, accessible via both persistent menu as well as user messages.
- A basic flow was designed — for the user to have a conversation that would last 40–60 messages — counting both sent and received.
Bot messages that were broken down into 2–3 smaller messages but sent at the same time, were counted as one message. The intent was simple. This was the cold user flow. To keep the user within the system for a bare minimum time and then let the user explore things on its own. Whether the ideal number should have been 60 or 100, I don’t know. I was planning to crunch the data once the bot started having live interactions — which of course never happened.
The Starting Point
It was kept simple. Nothing fancy, nothing intrusive. Not distracting the users with a bunch of menus and carousel items.
- The bot would start with asking people a simple question — Do you remember when I moved on from international cricket? There were obviously only two options — either right or wrong. And from there the snowball starts to roll. By this time the bot had sent across one or two funny messages (including memes, gifs). The fun part was there to give it some resemblance to the cheerful, laughing image of Sachin that people associate with.
- I experimented with both things in this first question — user input as well as quick reply. Found out quick reply to be a better alternative, so quickly dumped the use input thing completely.
A few of my friends who actually got to play around with the bot had quite a lot of fun. A special shout out to Alok Pratap, who — due to some reason — came back to talk to Sachin every other day. But once I had stopped building up further on it, he eventually stopped.
The Curious Case Of Unfamiliar Messages
We would receive messages from the users that would be outside the scope of what the bot could understand. It was not a matter of possibility, but an undeniable fact. I would have loved to test out api.ai, but the mere thought of integrating anything that requires me to code gives me nightmares. So, if I can avoid it, I would avoid it.
Then I remembered the scene from a movie/tv-series — don’t remember which one was it at the moment. A female cop was going undercover, and there was an advice from her friend — if you get stuck and are asked a question you don’t know the answer to, sneeze! By the time you recover, they say gesundeit and have asked you if you are okay, they have been distracted long enough for them to forget what they were talking about.
Distraction!! That was my takeaway there.
So, I created 15–20 different blocks, fed one interesting trivia about Sachin in each one of them, and whenever the bot would receive a message that it couldn’t process, it would select one of these trivia and throw it back at the user.
Did you know I used to be a bully in school?
— Sachin bot
One of the messages also said “Uh oh. Looks like we’re running short on time today. Family time. Take care. Let’s talk later”.
It was fun!
The Marketing Plan
Obviously I wanted people to be using the bot — talking to it. Coming back again and again to talk to it. There were a few things that were already in the bot to address the “coming back again and again” part. And some things that were planned to go live a week after the first release. But before that second conversation would have happened, the first conversation needed to happen. And that is where marketing comes into the picture.
Now, I am a shallow man. If I can save marketing spend in any way, you can bet I will try those free alternatives first — at least for a few days, before I set up the add. But in this case, I had decided to spend some ad dollars on day 1 — which I later wasn’t sure of doing on day 1 itself, but that’s beside the point here.
There were three angles to it.
The Most Basic One
After some time of engagement, the bot would prompt the user to share the bot with his friends. A card would be shared with some eye-catching copy.
The Instant Autograph
A user could get an ‘autograph’ instantly — and the message would be completely custom. I would have used a simple script to do this. The bot would ask the user what would he/she want scribbled on the image, the input would be recorded as an attribute, sent across as a parameter and another message would be received via JSON api containing the generated image.
Phew! I am not a programmer. This shit was the toughest to implement, and that too when — (A) It is super easy, (B) I had a step by step guide.
Here are some sample autographs. (Please note at the bottom edge — it clearly mentions that this is a computer generated image. The idea was to make this text more prominent in the final release. We never got to that point.)
Overtaking Twitter
Okay, moment of truth. Honest confession time.
I am a Sachin fan myself. A huge huge one.
Think “Donald Trump” huge. Huuuuugggeeee!
So I knew April 24 is his birthday. And like any celeb’s birthday, there would be a swarm of users all over twitter wishing Sachin. A simple twitter bot was all that was needed.
#HappyBirthdaySachin
That was all I needed. The hashtag people will be wishing him with. I would get to know that by midnight itself, and then all I needed to do was fetch every single user tweeting this and send them all one of these messages:
- You may not get to talk to Sachin today, but you can talk to the Sachin bot (a fan made facebook chatbot)
- Get your message signed on a Sachin wallpaper. (Completely generated image. Not a real autograph)
- Hundreds of trivia and facts about Sachin you didn’t know. Talk to the Sachin bot today.
That’s it. Should work. Looked all good on paper. And then there were some backup strategies ready to roll out if all failed.
It All Seems Good, So Why Did It Never See The Light Of Day?
Uncle Ben!
With great power comes great responsibilities.
Just because I could have, doesn’t mean I should have. And that’s what stopped me.
Even in the beginning, when the idea started brewing in my head. When it was just a fun thing to do, and learn more about Chatfuel’s system, I obviously did not want to ruffle any feathers, or rub off of someone the wrong way. This was quite a different league that I would have been messing around with. And I had thought of a number of things from the get go:
- The bot shouldn’t — in any way — imply that it has any association with Sachin.
- It should very clearly mention that this is a fan generated bot, for fun and entertainment only.
- There should be no commercial angles to the bot — no affiliates, no direct sales. Nothing.
But still. A week before the bot was supposed to be ready and go live, at the request of one of my friends, I decided to give a call to an attorney.
The attorney agreed with me that I was using all necessary safeguards so as to ensure that there are no liabilities. He then went ahead to add.
We Can Make It Safer By:
Not using the name of Sachin, but calling it a general celebrity bot. A bot that answers things about Celebrities.
You are, after all, using the brand name, the brand image and the goodwill that comes associated with Sachin Tendulkar.
Now, this was a kink — for sure. Giving the bot a name that in no way relates to Sachin and when we don’t explicitly talk about the bot being directed to Sachin would mean I would need to rethink a lot of my marketing activities.
But it was a bump in the road — at best. Whatever the attorney’s suggestions were could have been implemented and integrated and yet the bot would have stayed its course. But then that point of me “using his goodwill” started bugging me. After all, he was right. And if I have to think so much on how to wade around the legalities associated with it, then maybe in an effort to ensure I was on the safe side of the law, I had trespassed on to the immoral side of the fence.
That’s it. That one thought was enough to scrap the project.
The Bot That Never Saw The Light Of Day.
SOME UNOFFICIAL AUTOGRAPHS
MARKETING BANNERS
These were some draft banners. The final banners contained a clear message that the bot is, in no way, associated with Sachin — in any capacity — directly, or indirectly.
Some Reviews Of The 100MB App
