Building bots is easy. Engaging users is not.
According to Botanalytics, 40 per cent of a bot’s users disengage after one interaction. With thousands of bots launching every month, what growth strategies distinguish successful from useless ones?
We spoke with the top performing chatbots to find out. Each of these bots has hundreds of thousands of users and thousands of daily actives. These are the best practices they use to thrive in an increasingly competitive bot market:
Poncho, 1–800-Flowers, and theScore all launched with Facebook’s Messenger bot platform. The media attention kickstarted their usage, but they sustained growth by promoting to existing customers.
If a purchaser on 1–800-Flowers’ website is already logged into Facebook, they are automatically opted in for notifications via the 1–800-Flowers Messenger bot. Similarly, the weather bot Poncho has a “core group of fans excited to try something new,” says Sam Mandel, CEO. The company promotes their weather bot via email newsletters and mobile apps. TheScore, a real-time sports media company, has three million fans on social media.
Cross-promoting is much easier than building new audiences. Top performing chatbots can acquire daily active users in the tens of thousands simply by advertising to existing fans. Fans already understand the underlying value of a brand or business, simplifying the transition to a chat interface.
What if you don’t already have millions of followers?
12 years ago, Steve Worswick built his conversational chatbot, Mitsuku, as a fun add-on to his personal music website. His bot soon became more popular than his songs. Worswick has since grown Mitsuku to more than five million users on web alone, plus many more on messaging platforms like Skype and Kik. The bot processes nearly a million interactions every day and has won awards for “the most human-like chatbot”.
“Putting a bot on Facebook is like advertising a blade of grass in the middle of a field,” warns Worswick. One of his favorite growth strategies is to avoid crowded bot directories and launch in unexpected places.
Mitsuku received ~100 visitors a day on Worswick’s website before the developer launched his bot on Mousebreaker, a popular gaming website featuring sports and arcade games. Mitsuku immediately stood out as a unique and artistic offering and became a hit with the website’s young, male demographic. Suddenly the bot saw hundreds of visitors and thousands of interactions every hour.
Launching on Mousebreaker was just the beginning. Users started posting videos of Mitsuku on YouTube. The bot received press attention, won chatbot competitions, and grew steadily to become one of the most popular chatbots on the internet.
Mitsuku stood out as an artistic and unique offering among arcade and sports games.
Despite the tech industry’s bot obsession, normal people usually don’t know how to interact with bots. Traditional advertising that pushes uninitiated users into bots have often been futile.
Shortly after their F8 launch,1–800-Flowers compared traditional paid Facebook ads that sent users into their Messenger bot or to their website. The website ads far outperformed the chatbot ads. Customers found the bot confusing and error-prone and made fewer transactions.
Swelly, a community polling bot, also experimented with paid ads and found them useless. Users assumed the bot had general intelligence and would ask Swelly random questions like “Are you single?” or “Tell me about the weather”.
Despite problems, Swelly grew from zero to over 800,000 users in three months. Swelly co-founder Peter Buchroithner paid YouTube influencers to make video endorsements of the bot. Buchroithner focused on fashion and beauty niches since Swelly is particularly useful for those categories.
Working with influencers cost only a few thousand dollars, yet these growth strategies acquired Swelly’s most engaged users. People who converted to the bot from influencer videos came in educated on Swelly’s functionality and excited to engage.
Limiting user options can lead to engagement boosts. 1–800-Flowers used to offer customers three delivery date options: Today, Tomorrow, or Choose A Date. The third option was removed after date input errors were found to cause the bulk of failed transactions.
1-800-Flowers saw so many errors from open-ended date input that now the bot restricts you to today or tomorrow for flower delivery.
TheScore implemented buttons to suggest popular sports teams to users and circumvent the need to type. “We’ve seen that users need to be given more prompts about how to get the information they need,” says Riaz Lalani, VP of Product at theScore.
Swelly initially required users to tap a “Next” button to move to the next poll. When the button was removed and polls automatically advanced, the average number of votes per session doubled.
To streamline usage and boost engagement, consider the following:
A user once chatted with Mitsuku for nine hours in one day. Most humans can’t even hold a conversation that long.
Engagement differs depending on bot category. As a general chit-chat bot, Mitsuku is expected to converse on a variety of topics. Users often ask her about recent sporting events, trending movies, or celebrity gossip. Worswick, Mitsuku’s developer, constantly builds in new domain expertise, including pop culture, anime, sports, politics, and other popular topics. The hard work pays off: 80 per cent of Mitsuku’s users return for another interaction. Many converse with the bot daily.
Utilitarian bots like Poncho require different engagement tactics. “Our retention is significantly better than a top tier iOS app,” says Mandel, CEO of Poncho. The bot’s high retention is driven by continuous testing and optimization of language, onboarding, and notifications. Poncho simplified language and minimised slang which engaged international users. The team also maximized the relevance and trustworthiness of their communication. “For people to let us turn on notifications, they have to trust that they can turn them off,” cautions Mandel.
Throughout Swelly’s development, Buchroithner and his team continuously reached out for feedback to make critical improvements. The result is over 800,000 unique users and 31,000 daily actives in three months since launch. The average Swelly user uses the bot three times a day and votes over 600 times total.
Facebook noticed these engagement numbers and made Swelly a featured bot in Messenger. Being featured drives thousands of new users per day for Swelly and is by far the single most effective of their growth strategies.
Poncho incorporates user feedback for smarter, personalised notifications. The team has to learn why specific users want to know the weather. For example, a user may wish to know if tomorrow is a good day to go running. Poncho can advise if she should run in the morning or afternoon.
Discovery is a constant challenge for all bots, even successful ones. Bots much constantly evolve their growth strategies in an increasingly competitive market. “We don’t know yet if bot stores are the answer,” ponders Mandel of Poncho, “The problem with mobile app stores is that paid acquisition is virtually the only way to acquire users. We hope that doesn’t happen with bots.”
Another industry-wide problem is user onboarding and education. Jon Mandell, VP of Multibrand Customer Marketing at 1–800-Flowers, believes these two major challenges will be solved when more people build bots and share their knowledge.
“We need everybody to jump on bots so customers learn how they work,” encourages Mandell. “Growth won’t just come from 1–800-Flowers. The community will thrive when bots become mainstream.”
Chatbot growth is challenging but possible. Companies like 1–800-Flowers, Poncho, and theScore capitalize on existing audiences. Bots like Mitsuku and Swelly employ effective growth strategies like topical awareness, niche advertising, and influencer marketing.
Learn their chatbot growth strategies and you might find success as well.
More like this
What is the state of measuring native advertising today? How do we work with quality measures? Will a ‘God Metric’ ever be a possibility?17th Mar 2017 Insight News
Content is a collection of data, an array of information and a bundle of details. As such content is flexible, it does not care on which platform it’s published or when it’s shared across social media and certainly content does not prefer one reader over the other.15th Mar 2017 Opinion
If you have Alexa, Amazon’s intelligent personal assistant, then you probably start your morning with a conversation with a chatbot. Because let’s admit it: chatbots are already everywhere. On the one hand, we love to hate them because they are, well, not human. On the other, we love to love them because they create an interactive and personalised experience. Brands have been quick to adopt this technology. Too many media companies lag behind experimenting in this field.16th Mar 2017 Insight News
FIPP newsletters allow you to keep up with industry trends, research, training and events across the worldFIND OUT MORE
Get global coverage of your launches, company news and innovationsFIND OUT MORE
What’s happening now, what’s coming next