Using bots to build MVPs on Viber

(Ed. Note: the following guest post is provided by Itamar Mula, Global Developer Relations Manager at Viber)

Viber is an instant messaging and calling app with more than 900 million unique users in 193 countries. You can send free text messages, fun stickers, photos, videos and doodles and make free HD-quality phone and video calls.

Our product team constantly works to improve the overall experience of the app based on the lean-startup methodology. We regularly test our ideas, making sure they are viable. But instead of building the whole feature, we strive to test our vision with a quick prototype and iterate on feedback and results.

While we are prototyping fast, we didn’t want to release a sluggish UI experience, so we had to find a better alternative which will allow us to test our ideation process. And guess what? We found it in our bot platform.

Why bots?

Bots are quick to develop, and have a very simple loop, which enables us to test the most complex ideas in a very straightforward matter. We use text messages, videos, GIFs, carousel content and keyboards as our front-end interface.

Moreover, bots have a universal user interface and work across platforms (iOS/Android and Desktop), meaning there is no need for multiple developers/UX designers and product leads. This keeps our task force lean and mean!

Most important, bots enable us to improve our MVP without releasing a new update.

We love stickers bots

Our users love stickers. More than 300K stickers are shared every minute! This volume requires us to continually seek to improve our sticker search, serving the right sticker at the right time.

We decided to build a bot for ‘We Love Viber Stickers!’, the official account for Viber Stickers. With more than 167K followers, this account is the place for sticker lovers and a great place to test our new sticker search capabilities.


Technicalities

Our bot provides an intermediary between the user and the search API. Basically, the bot has three states:
  • On-boarding: Serving a welcome message to new users, explaining the purpose of the bot and how to use it.
  • Waiting for a query: The user tells us which sticker that she or he would love to use.
  • Serving a query result: If we find matching results, we return the top X results, where the we use some custom sorting features. If no matches are found, return a guiding string of different queries. 
Since it is a simple flow, we decided to host it in an AWS Lambda function. There is a single entry point, serializing the state to our ‘user_data’ field and basically iterating between query to result.

Measure and learn

So now that we are actually able to test our search algorithm, we’d love to measure our users reactions, and learn from the results.

It took us literally 10 minutes to integrate Chatbase, but it saved us days! We integrated the Chatbase REST API into our pipeline, but nowadays you can even do it even faster using an npm package.

For starters, Chatbase helped us easily monitor different metrics such as active users, sessions and retention. However, the real perk came with its ability to cluster problematic messages via machine learning, which saved us precious time that would otherwise be spent combing through the logs.

Results

After a few cycles we managed to increase query volume per user by 35% for our stickers bot by optimizing queries with high exit rates. I’m confident that we can enhance this even more and perhaps by setting intents in Chatbase, we can reveal some new truth.

We found the whole “MVP on bots” quest incredibly fulfilling. It enabled us to close the gap between vision to reality with minimal resource spend and short sprints, which is every developer’s dream.

About Chatbase 

Chatbase gives builders of conversational interfaces (or bots) sophisticated tools for creating better, and stickier, consumer experiences than ever before--leading to better conversion rates and retention. Chatbase is a cloud service that easily integrates with any bot platform and type, voice or text, and is free to use.

Among other features, Chatbase uniquely relies on Google’s machine learning capabilities to automate the identification of  bot problems and opportunities that would otherwise take a lot of time, leading to faster optimizations and better bot accuracy.

Chatbase is brought to you by Area 120, an incubator for early-stage products operated by Google.



Comments