A quick explainer what changed and how brands can grow their Messenger audience with Messenger games.
Many of us remember the time when we received a handful of FarmVille invitations a day and, maybe for the first time Facebook was pretty annoying if you cared about friends rather more than your digital pigs.
Zynga, the game company behind the FarmVille craze, started after Facebook opened their platform for third parties in 2007.
They launched with Texas Hold’Em Poker, the first game on the social platform, and were the fastest growing third party on Facebook with 40 million users in April 2009.
Then came FarmVille. 10 million daily active users in 6 weeks, 80 million in 6 months.
Zynga became an upstart with a billion USD valuation and hundreds of millions in revenue through in-game purchases — gaming on Facebook was a global phenomenon.
In 2012 Facebook changed their mind and decided to walk away from gaming.
Mini-games started to hurt the user experience and Zynga became too big.
This was a tough decision, as Facebook made about a quarter of their revenue through gaming.
Facebook banned games into the dark corners of their platform and it became niche compared to the scale of Facebook.
Now Facebook is changing their mind again. Facebook Messenger games are placed front and centre and are starting to thrive like in the old days.
Facebook is displaying the active (!) players of a game every time you open the game. So it’s easy to spot that classic arcade games like “EverWing” (14 million users) “Basketball FVR” (8 million users) or “Super bowling” (3.6 million users) are dominating the platform.
For consumer brands, Messenger games are a real engagement and growth engine.
Here is why:
Brands which want to increase the engagement on social or increase the customer lifetime value (CLTV) have to build a Messenger bot audience.
To get an audience on Messenger the user has to opt-in, eg. through writing a message or clicking on an ad. Otherwise, the brand can’t message the user. Those are the customer acquisition costs (CAC) brands have to get Messenger user.
Messenger Games provide that opt-in, so every game user becomes automatically a Messenger bot user. 🎉
Red Bull – launched the Airdrop Game. Collect Red Bull organics with your balloon and avoid the rocks.
KLM – One of the leading brands in the messaging space launched „Flying in the sky.“ Flying with KLMs historic machines to collect coins and not run out of time.
The average CPC for a Facebook Messenger ad is currently a little above 1.20$.
With the traction that the first brand Messenger games have shown the CACs by using a Messenger game is around 3x lower than using ads. The best part about this: it’s a one-time investment so the return on investment is getting even bigger over time.
Especially for large brands which aim for a bigger audience that is a real opportunity to grow their audience on Messenger quickly.
All major tech companies already made their bets on chatbots — except Apple. Last weeks WWDC revealed how Apple is stepping into the conversational world. After the shift from web to mobile and apps now the next big change is coming: from apps to chatbots
As a chatbot creator, you have to choose the right feature for each interaction so that users will not get lost in too many options displayed. This article takes a closer look at some features by comparing their genuine qualities and technical limitations. If you are able to make best use of the qualities and see boundaries as a challenge you are ready to take up, you will be able to choose the best feature for every single interaction.
Content Display and Interaction
Before going into detail, let’s take a look at what we are dealing with. You will have to make some preliminary conceptual decisions before deciding on specific features, too.
Horizontal-scroll and list template are elements for content display. To chose correctly you need to define the main purpose of your bot. There are two possible goals: you either aim at a great display of content, followed by a nice interaction that converts to your website. Or the entertaining and meaningful interactions and conversation you created leave a happy and engaged user who you hope will return. Clarity on your intent helps your decisions on how to display your content best. It also influences your choice of interactions, be it due to technical reasons.
Don’t just think of where a conversation leads. Consider who starts the conversation and with what intent. Do you want to engage a user by sending the first message, displaying content and providing actionable buttons? Do users write the first message, turning to your chatbot for entertainment or in the urgent need of assistance?
These opposite directions and different intentions influence your choice of features, both for content display and interactions.
This article provides insights on bot-user-interactions such as buttons and quick replies. Interactions like these always define the link between single messages. For every new message and next link, you will have to choose, again. The right choice leads to happy users. They will engage in every new interaction you offer, thus use your chatbot a lot and/or follow links to external websites.
Two ways of visual content display: horizontal scroll or list template
The horizontal scroll or ‘h-scroll’
If chosen wrong or edited poorly, an h-scroll bares an overload of information that discourages users. Used in the right context and designed according to its benefits as well as the user’s needs, it is a great feature for storytelling.
An h-scroll consists of several structured messages in a row. Structured messages are message bubbles containing text and up to three buttons for interactions below. Image, title and subtitle can be added. Up to ten bubbles with up to three buttons per bubble let you display a lot of information in media and text. The buttons provide a number of choices: further interactions within the chatbot or a conversion to the website.
Because there is no overview at first glance, the user has to interact to see all bubbles.
Its benefits can also be a downside: h-scrolls can be too packed with content and options if poorly concepted. This can easily happen when the h-scroll is used in a conservative way: giving an overview on five similar news-articles or several products to buy, using interchangeable buttons in every bubble. It imitates a menu by giving choices that in a web-interface would be displayed as a menu bar. This is obviously a bad user experience, not making any use of the chatbots qualities as a channel.
So should you simply cut down on the number of bubbles and buttons? No. Instead take advantage of the h-scroll’s unique benefits. Doing so, more bubbles are not only okay, but desired and enjoyed by your user. The unique quality the h-scroll provides is its horizontal line of storytelling. It resembles the way people read in western countries, from left to right. Use this habit: put cartoons into the bubbles as a narrative, split one image into three or more to create a panorama view, tell a story!
The h-scroll is also perfect for step-by-step tutorials and advertising. Users attracted by the first bubble are very likely to scroll to the others. They will then choose from the several options the different bubbles and buttons provide to either leave the chatbot and visit your website or continue the experience in the next conversation.
You have already seen examples for this creative use in facebook carousel ads.
Use the h-scroll when starting a conversation if you intend to present something in wider scope and if you want to provide several directions to continue the conversation.
The List Template
If you want quick overviews and easy selections, the list template is your weapon of choice. You will have to meet its limitations in content display compared to the h-scroll, though: The list template lets you display up to four items vertically. Each item contains 160 characters of text an image and one button. You can display one item prominently with a cover image and up to three smaller items below. You may even add a different kind of interaction to the cover item’s button than to those of the others. If you do not intend to highlight one of your items, just display up to four items in the same size. These item’s buttons are likely to cause similar interactions, for example allowing a choice between the items.
Additional to those buttons you can provide another one for the whole list template, below the items. It can lead to a different conversation or action.
The list template displays all information at a glance. No further interaction is needed for a complete overview.
Its limited characters can sometimes be a hindrance. You can not display or explain as much as with an h-scroll. So take up the challenge to create concise text combined with significant pictures that add up to and enhance each other in meaning.
The choices users have in the list template are limited, compared to the several buttons provided in an h-scroll. This makes the list template a fit for quick decisions. Think of selecting the colour of an item to buy or the decision between the top news and three further topics in a news channel. The button below could link to other conversations, e.g. other products or different news topics.
Use the list template if you want the conversations to be speedy. If users turn to the chatbot with a need, it is your shortcut to serving that need. The faster users get to their goal, the happier they are. So the list template can be the suiting reaction to a question or other initiative by users that calls for multiple pieces of content. If asked or asking for specific content, determined sets of answers are helpful. Fewer, but lean options make it easy for users to choose fast. The results — next messages and interactions — follow up immediately.
Why display content in the first place? You want users to take actions. Whether h-scroll, list template or a plain text-, media-, or structured-message: two ways to define the interaction between those messages are buttons and quick replies. They are your alternative to the complex use of Natural Language Processing. So after choosing the way of displaying your content, you will now learn how to find the best feature to make your user interact by comparing button-interactions.
Buttons as well as quick replies are an actionable and fast way of interacting. Well chosen, they are helpful in making quick decisions and engage the user to play back and forth in a smooth dialogue. Each feature has it’s own benefits and limitations.
Combined with the wrong message or linking to confusing interactions, buttons will not be appreciated by your users. Learn what they can and can’t do and use them accordingly to script an interaction that feels natural.
Buttons trigger an interaction at the end of a structured message. This type of message contains up to three buttons, each linking to one interaction, e.g. the next message or a new conversation, subscribing or unsubscribing users, allowing user feedback, or containing an external URL. Each button consists of text describing its command and can be funnier with emojis, if that suits your chatbot.
Like the h-scroll, if used wrong, buttons can offer either too many different or boring repetitive choices. Several bubbles with different pictures but the same button-interactions below can be boring (how about using the list template instead?). Eight bubbles with eight times three buttons to choose from make users forget the first options while still browsing. Leaving all choices to the user is not a benefit, but a pain. A bot creator has to figure out the right preset of choices in every interaction. Better use three different interactions narrowing down the choices step by step than offering one with too many options and directions to choose from.
Also decide if you want the user to continue with the bot conversation or rather link to an external website.
Buttons and the different actions and several directions can widen the scope of a conversation. It may not always speed up, but users will enjoy their freedom of choices. With every click they make a decision towards further conversation or following a link out of your chatbot. Once made, this choice does not disappear, the button(s) can always be found in the conversation, right under the structured message they are part of. Users can return to this message and undo their choice by clicking a different button.
This distinguishes them from another class of buttons:
Quick replies can put your conversation on the fast lane. Their technical limitations do not allow reconsidering, so you have to choose wisely. Here is how: Quick replies are buttons prominently displayed above the composer, containing text, and emoji if suiting. They follow messages of plain text and/or media. Quick replies are perfect when asking for choices on up to five items. Users will less likely use the keyboard but just tap one of the quick reply buttons provided. Imagine a message asking for your favourite flavour in ice cream. If you then offer a quick reply with five choices, the user can easily tap on ‘strawberry’. Typing this word into the text field takes at least ten taps.
Once tapped, the buttons are no longer displayed. Users thus can not tap on old buttons unrelated to the last message they received. But there is no way back in the decision tree. This downside in using quick replies can also be seen as their benefit. ‘Burning bridges’ will automatically speed up the conversation. There is only one direction to go: fast forward. Use cases are comparable to the list view’s: providing quick selections if users ask for content. Users will then be grateful if you present them with 2 to 5 actionable quick reply-buttons to tap instead of to much info and many buttons in a structured message. The work you do in templating these choices suiting the user’s request will be appreciated. They make a chat meaningful and fun by leading to results. Even when using buttons, the conversation feels natural, with no delays or misunderstandings.
Quick replies are also a great way to grab user data. You can only use quick replies within the chatbot, they can not contain external links. So if you aim at linking to an external website, you will at some point have to switch to a button-interaction.
The choice on a feature made for an interaction is only valid for this one link. The next link to the next message means yet another decision. Choosing one feature for the first content display and interaction does not mean that it suits the next: you will have to reconsider. What sounds like a lot of conceptual work is actually a blessing: mixing features, thus ways of content display and interactions is what makes a meaningful chatbot conversation, because it addresses the users needs.
Push a message, mixing h-scrolls and buttons for broader scope in the beginning and then start narrowing down choices, collect user data and speed up the conversation using more list templates as well as messages with quick replies towards the end. Users enjoy your storytelling and several choices in the beginning while still being led to a result in the end, be that an entertaining chat or the visit of your website. If offering or asked for quick results, mix list templates and messages with quick replies. If needed at the end, use buttons for their benefit of linking externally when the list template and it’s limited buttons are not suited. There is unlimited ways of recombination. Choose the ones suiting your purpose best. Keep in mind the main outcome you aim at as well as your goal in every single message.
Users have tried bots for a while now. Some of them are fun and impressive. But people have also been dealing with bad technology and unsatisfactory prototypes. They tried to figure out why to use bots, even though identifying the purpose of a bot is the job of its creator.
What is a bad experience for some, is an opportunity for professional bot creators. Do not present your users with a boring bot they do not understand, they will stop using it. Scripting an engaging bot conversation that makes users come back for more is not as hard as it sounds. Making a great chatbot is thereby less dependent on the technicalities like conversation paths, keywords, and call-to-actions. The one thing that makes the difference in the end is whether your bot has a great personality. This personality consists of a purpose, matching characteristics and a suiting tonality. Crafting a personality is the most important part of chatbot creation. Taking the time to create one will pay off very quickly.
The first thing to have a closer look at is the way your bot ‚talks’.
How to Create Your Bot’s Unique Tonality
Conversations in instant messengers have been a private domain up to now. People chat with friends, family and colleagues. Handle these conversations with care. Do not try to imitate a human being because people will immediately notice! Provide the crucial benefit of a chat: being personal.
How can one be funny, interesting and personal in interactive scriptwriting? Not by scribbling down some lines, picking one and then write some similar. Don’t start scripting before taking the step of creating the bot’s personality, thus it’s tonality. Developing it incrementally and skipping the conceptual work is a fatal mistake.
Form Follows Function
There is a reason you want to create the greatest bot in the world: You want to promote a product, a brand, or a person. Let’s take a look at some examples on how a bot’s purpose is crucial to its personality and thus influences its form, the tonality it shows in conversations.
Rescue programmed a bot to call for help when you are not able or it is not safe to make a call. It scans all important information really fast, just like a police officer or doctor would do. Whether Rescue imagined this person while creating the bot, or if it happened by accident: The purpose of helping people very fast in a smooth and easy conversation calls for a focused, yet calm personality.
Clarity on your bot’s purpose helps a lot in finding a suiting bot personality. Outline about 5 benefits of your bot and prioritize them. The most important one has to be your main purpose, the reason you want to launch a bot in the first place.
If you are aiming at user engagement, learn from the best: Danish singer Aura’s bot represents Aura as an artist (bot-personality), thus uses her own tonality and her ways of communicating with her fans. They turn to the bot for entertainment as well as information on their favorite artist. Aura’s very personal tonality suits her bot’s purpose to provide her fans with the feeling of being close to and engaged with their star. This engagement leads to user retention.
Poncho managed to create a bot that makes weather reports less boring by crafting a funny personality. It provides forecasts and then enhances its informational content with funny media and remarks. This entertaining part lets users enjoy the chats. The extension in Poncho’s bot personality causes the switch to a more personal tonality, which leads to an engaging interaction and thus chatbot retention.
There is more to a personality than its purpose and tonality: characteristics.
How to Do It: Create Your Bot’s Personality!
Creating a full and concise bot personality is neither magic, guesswork nor luck, but work. This is how you proceed: Imagine you were a script-writer for a movie or play. As such you invent the bot personality by pinning down all its characteristics.
If your imagination does not provide the perfect persona instantly, these questions will help: Do you have a mascot (for a product or brand)? Does this mascot have a personality that might serve as inspiration? Is there a core value for your company in general or to its way of communication that could inspire the personality? Can you think of a person you know, a celebrity or a fictional character that your bot could resemble? Using one of those or even mixing different templates is your starting point.
Now you fill this person, animal, or whatever you chose, with attributes and facts: Age, heritage, friends, occupation, just make things up! Of course they have to make sense, correspond to your cause and be concise. When deciding on further features that are not distinct: Open a spectrum from, for example, ‘yoga-enthusiast and healthy eater’ to ‘couch potato and heavy drinker’ and place your bot personality somewhere in that range. Do it for several opposites.
There is no limit to your imagination but one: All the attributes and facts you make up have to make sense if you form them into one personality and fit to your brand. By the time you found a set of facts and qualities that really adds up, you can always ask yourself: What would my persona do? And your bot personality will give you the answer.
Your bot’s personality will now provide you with suitable ideas on the way this chatbot addresses the user: its tonality. Is it humble and nice or rather cheeky and sometimes a little off? Is its way of communicating structured and formal or rather funny and associative? You will also have to decide if the bot talks in the first person, singular or plural, or in the second.
Sound like a lot of work? The bot personality you are about to create based on this insight will make telling your story a lot easier. Your bot’s tonality will be far away from the boring machine described in the beginning of this article. Scripting your conversational bot will be as much fun as interacting with it!
Bearing this in mind, Google launched the Google News Lab in June 2015 . Two months later we founded Spectrm and our collaboration with the News Lab might be an indicator of how much Google and their relationship with publishers changed.
Here are three examples:
The core of the conflict between European publishers and Google is about how technology is changing journalism and its distribution.
Hacks/Hackers is addressing this very problem. A grass root movement where journalists are meeting developers to trade ideas. Started in 2009, Hacks/Hackers has spread all over the world with local communities.
One of their first bigger events took place in summer 2015 in Berlin with the support of the Google News Lab.
It was the event where we, without much expertise in the media business, got confident by talking to many journalists and entrepreneurs that we had the ability to solve a big problem for publisher.
Almost a year later Phillip invited us to join an event at Hacks/Hackers New York. It was our first time in New York and the event where we got connected to one of our major seed-investors.
Maybe the most important step in a startup’s life. 🎉
2. Digital News Initiative
The same day we arrived in New York, the Digital News Initiative pronounced us as their first funded project. The Digital News Initiative is a partnership between European publisher and Google and — yes — it is about money. Honestly, money for projects which would otherwise probably never get (enough) money to be successful.
Despite the current hype surrounding AI-startups it is actually an invidious affair.
It’s costly, it’s time consuming, full of experiments that often lead nowhere and work with way too little data. Being part of the Digital News Initiative by Google allowed Spectrm nonetheless to dig very deep into the field of artificial intelligence.The result is a unique technology in the field of natural language processing, chatbots and content recommendation.
We are pleased to launch this technology in the next few weeks.
3. Google News Summit
The Google News Summit took place in early December last year. Following the election and the subsequent discussion on fake news it was probably one of the most important media events of the year.
It took Facebook quite a while to acknowledge their huge impact on the news ecosystem and our society. Google instead learned the lesson and invited journalists, professors and startups to Mountain View to discuss all aspects of the current issues.
It’s not a coincidence that Google is acting fast and concerned with solving the problem rather than being part of the fake news discussion as Facebook is.
Sure, Googles efforts around journalism and media are not entirely altruistic. Google needs the open web and users will only search for content if the content is worth searching for. Without quality journalism a better search or distribution is worthless and vice versa.
That is the challenge for the whole ecosystem and partly Google itself.
Coming from the startup world, being collaborative and passionate about MVPs and experimentation is normal, but it’s very interesting to see the Google News Lab is using the same principals. Supporting early stage experimentation with the DNI and connecting the right people at Hacks/Hackers or the News Summit is a strategy that seems to be successful in building a new relationship with publishers.
The pleasant side effect of helping startups like us makes that even better. 🚀