Since we moved away from chat a few days ago, we’ve had a lot of users come and ask us why we decided to bring chat to a bare minimum in our app.
While several of our users loved the chat feature, for a lot of our users chat was a cumbersome way of doing things (too many taps, easier if they did it themselves etc.).
After extensive debate internally, we took the call to phase out chat from our app and create the same level of experience using automation and simple user interfaces. The rest of the app remains the same – but it is simplified, fast and without any delays.
Since a lot of people believe that chat is the new universal UI (like we once passionately believed), I thought it might be useful to talk about our experience and learning.
Background
Back in early 2015, we thought that chat would become a universal UI because people were really using chat like crazy on their mobiles for P2P messaging. We thought we could build a personal assistant that helps them get things done over chat.
We built our chat process team over a period of 3 months from March to June 2015. By June 2015 – we were probably the biggest C2B (consumer to business) chat apps in the world. Probably bigger than all our Indian competitors and American counterparts like Operator and Magic combined. We were doing more than 70,000 chat sessions a day (not messages, chat sessions). The idea was that the manual chat would provide enough training data for building out a great bot.
Initially, we had a great response – customers hadn’t seen an experience like this before and loved the fact that there was a real human being on the other end helping them with their queries.
Learning on chat
The only problem with chat was that customers who tried it weren’t coming back. So we tried harder, optimised first response times, average response times, our knowledge base, canned responses and built better dashboards for monitoring all of these (all the while scaling the backend for the chat volume). That still didn’t work.
We thought that at some point users would get trained to use chat in the right way (like how people got trained to use Google). It now seems foolhardy to even think like that but we really thought that chat UI will truly work, if only we worked hard enough.
Next we thought – how about using a NLP-based chat bot and also experiment with reimagining chat (make it into a series of simple clicks with graphic UI elements).
The chat bot irritated the users no end (our data science team built a large set of training data, but coping with Hinglish, SMS lingo, vernacular language to write a general purpose chat bot becomes a 5 year science project very quickly). Breaking up across different channels didn’t make the task any easier.
Reimagining chat worked where the permutations and combinations were very limited. Example, we wrote a laundry bot where you didn’t have to type – just select options. It worked to some degree. Then we decided to implement it for ordering food but in user studies the users kept saying that it was very painful to use.
The hard truth (for us)
Finally somewhere in Sept 2015 it became obvious that the problem was chat (there were lots of learning along the way but this one was the biggest). The number of taps required to get anything done was just way too many, no matter what you do. People often confuse the debate with bots, AI etc. but does it really make sense to chat to get a cab when you can simply tap once and book (no matter what you think, our user data was very obvious on this. The answer is no.).
The customer would try it once out of curiosity but never come back. Other ways (simple UI) was just way faster and on the internet people will ALWAYS choose what is faster. Chat might be great for long tail use cases (that random philosophy book that you had always wanted but couldn’t find easily) but mobile businesses in India don’t get built on them.
This Dan Grover (WeChat product manager) article nailed it and before that it was Connie Chan’s (A16Z partner) tweetstorm.
We had learned this the hard way more than 6 months earlier.
Learning and changing
So in September 2015, we took stock and decided to introspect. To use a Charlie Mungerism, we were behaving like a man with a hammer (chat) and everything looked like a nail to us.
We decided to think with first principles and asked ourselves – What does a personal assistant do? What is the easiest/fastest way to do things? What is required to create a personal assistant for millions of Indians? What should be the principles based on which the product team should build? Etc. etc.
New launch and traction
We finally launched the redesigned app in January 2016 and we haven’t looked back since. It was based on a simple insight – don’t force fit chat, just build for the user.
We’ve been growing super fast with probably the best cross category transaction behavior of any app in India and it looks like we’ve gotten something right.
More than 20 API integrations are already live (and for most of our partners, we are the biggest partners in terms of transactions or traffic). Another 15-20 APIs will be live in the next 3 months.
It also proved to us that chat or voice didn’t have anything to do with building a personal assistant (people also confuse chat/bots with AI etc. We believe AI/Machine Learning/advanced algos are relevant even more so outside of building a chat bot – you’ll at least have actual usage and be able to train the system). Just focus on what is the easiest way to do things for the user and the rest will figure itself out.
The need for speed
We are on our way to creating the world’s fastest product and tech team. We launched recharge, bill payments, cabs, deals, news, journey cards – all in 5 months. In the next 3 months, we’ll be launching many more things. Some will wow our users, some will be duds. But that’s the only way to create an awesome personal assistance platform. Fail, learn, ship, fail, learn, ship…
Personal note
This tweet kind of sums it up.
While it is supposed to be tough, the toughest of all was the decision regarding our chat operations team. We have taken the decision to let go of most of our chat colleagues. We looked at the retention cohorts or uninstall cohorts for all our chat users but they were always much worse than others (for reasons explained above). No matter which way we looked at it, it made sense for us to take this decision.
3 month severance package and outplacement
We’ve given a 3-months’ severance package to everyone who was let go (i.e. our chat colleagues) and we will do our best to help them find new jobs. I’m working with our HR team on this and we have already roped in 12 startups and big companies for out-placing them into the right job profiles (chat ops, voice ops etc.).
I feel proud of the fact that I am working with such smart people in our team who are insanely passionate about our mission. Building a AI-powered personal assistance platform is exciting beyond words.
Ankur
Founder and CEO
Helpchat
Leave a Reply