Cross Platform Bots

In 2007, when Apple released the iPhone, Jobs made a controversial decision not to allow Flash on iOS. Given the long history between Adobe and Apple, not to mention the prevalence of Flash on the 2007-era internet, this was more than a little shocking. As a result of the flak, Jobs penned an essay where he wrote that the most important reason for not allowing Flash on the iPhone was because:

Flash is a cross platform development tool. It is not Adobe’s goal to help developers write the best iPhone, iPod and iPad apps. It is their goal to help developers write cross platform apps. … Our motivation is simple — we want to provide the most advanced and innovative platform to our developers, and we want them to stand directly on the shoulders of this platform and create the best apps the world has ever seen. — Steve Jobs

Today, in the face of a rapid onslaught of bots, designers are faced with ensuring a high quality experience across a myriad of messaging platforms. Some argue for consistency across platforms:

*Users should experience a similar, consistent interaction across platforms with your bot, and be able to share these services with multiple users on different platforms. For instance, if you’re on FB Messenger, you should be able to coordinate your shopping with users on Kik, WhatsApp, LINE & etc easily without leaving the Facebook interface. — Kip

And:

Does cross platform design become a thing of the past?? Adding functionality is as simple as adding a user to your chat — no longer a need to design for separate platforms. — The medium is the message

I see parallels between what happened with apps, and what is to come with bots. Jobs put it eloquently:

We’ve been there before, and intermediate layers between the platform and the developer ultimately produces sub-standard apps and hinders the progress of the platform.

Should bots be consistent across platforms, or not? To answer that, let’s see what the major differences between messaging platforms are. From there we can see if there’s some grand unified “theory of everything” for how to approach bot design across platforms.


Differences between messaging platforms

Differences between platforms

Currently, the messaging ecosystem is divided between a number of players. You can think of each of these like its own Operating System. Every Operating System has its own culture, its own language, its own best practices for facilitating communication.

WeChat

“WeChat is like a fully fledged cityscape where all the electrical and plumbing have been installed…developers can come in and build all kinds of unique and distinctive real estate that assists people as they go about their daily lives.”
We chat about WeChat, MLS Group

There are three broad areas where platforms differentiate themselves from each other: technical capabilities, audience, and communication styles.

1) Technical capabilities

Every platform supports a unique subset of technical capabilities.

The most basic, and certainly the most ubiquitous, is SMS. Almost everyone with a cell phone has SMS, and everyone knows how to text. That ubiquity is nothing to be scoffed at; it’s hard to find an easier onboarding process than sending a text.

A Chat Bot typing indicator

At the same time, (non-iMessage) SMS lacks some features that would be really nice to have. There’s no delivery receipts, no typing indicators. You can send text and pictures, but forget about anything like Facebook’s structured messages. And if your app depends on sending rapid fire messages that arrive in order, well, you’re in for a world of hurt.

Fat Cats Fat cats on Facebook

At the other end of the spectrum, Facebook Messenger supports typing indicators and read receipts, along with structured messages. Messenger also supports sending pictures, stickers, and audio. Messenger comes with its own limitations, however, like requiring bots to go through an approval process against Facebook’s Terms of Service.

The limitations of particular platforms lead to some interesting tradeoff decisions. If you support both SMS and Facebook Messenger, should you design for the lowest common denominator (SMS) and build your interactions in text-only? Or should you offer structured messages to your Facebook customers, and text-only interactions over SMS? What if you’re building a game, and speed is of the essence? Are you comfortable putting your SMS customers at a disadvantage? Do you force SMS customers to only play with other SMS customers?

2) Audience

Every platform brings with it a unique audience, with potentially unique expectations.

Top 3 From http://www.emarketer.com/

Platforms can be skewed by geography. For instance, WeChat dominates in China, with 700m users. In the US, Facebook Messenger tends to dominate with 60% of the market. Appealing to different parts of the world will drastically impact the design of your app; beyond the language differences, designers need to remain cognizant of the cultural implications their interactions could provoke.

Platforms are often skewed by age. 60% of Snapchat users are under 24; SMS almost certainly skews older. Different age groups will exhibit different expectations for how to engage with bots.

Race, gender, levels of education, or technical proficiency are all additional ways platforms can and will differ from one another. For instance, Slack is known for its highly technical early adopters, who are presumably more comfortable engaging with bots.

3) Communication Styles

Finally, the same user may adopt drastically different styles of communication across different platforms.

Slackbot Brandon gets it done

For instance, does the platform support group or solo conversations, or both? Bots on Slack generally interact with groups, while bots on Facebook Messenger are (for now) solo. SMS falls somewhere in between. These require fundamentally different strategies for design, and bots should know whether they’re speaking to a crowd or not.

Average message length and expected time-to-response are other ways platforms can differentiate themselves. My conversations over Facebook Messenger tend towards rapid fire communication, and I find the platform encourages this with instant delivery and read receipts. On the other hand, email leads to much longer response times, measured anywhere from minutes to days. Bots can and should mediate their verbosity and response times as appropriate.

Finally, the types of language used across each platform may also differ. Certain platforms are present only on mobile, which can lead to typos and a shorter, more to the point communication style. Or, a user might expect something closer to natural language.

Divide and Conquer

Surveying the landscape, it’s clear that providing a best-in-class experience for a particular platform requires a thorough understanding of that platform’s limitations and strengths. At the same time, there’s clearly a strong incentive, both from a branding perspective and from the a resource perspective, in providing consistent experiences across platforms.

Two conclusions stand out to me: 1) We need a way of separating content from presentation. 2) Tools used to build bots should allow and encourage designers to iterate, measure and test cheaply and easily.

Separating Content from Presentation

On the web, it’s common to decouple the business logic (the API) from the view logic (the UI). The server can speak JSON, and the clients can consume that JSON and display in whatever format is best for the platform.

A similar technique might come in handy with bots. We could think of the server as the content and the client as the personality.

Content

Content is the business logic and should be the same across platforms.

Whether you order Seamless via Messenger, via SMS, or via the Seamless app, you know you’re going to decide what to eat and then you’re going to order it and then you’re going to eat it. This should be consistent across platforms.

Personality

The way we interact with bots — that’s the part that can be unique to a particular platform. For instance, Poncho on Facebook Messenger is witty, and that’s part of its charm. However, consider that you’re interacting with Poncho over SMS. Is it appropriate to send so many messages? What if I’ve got a limited data plan? Or consider a bot on Slack. Do I want a chatty Cathy bot adding to the already considerable Firehose of Slack information? Probably not.

Bender

While we can strive for a consistent experience when it comes to the content, we should embrace diversity when it comes to the personality. Design and build for a specific platform.

Richer Tools for Design

On the web, designers have developed a rich set of conventions and tools for building best-in-breed experiences. Designers will need similar tools in the bot world to produce the same, things like:

  • A/B Tests
  • Detailed analytics
  • Prototyping and wireframing tools
  • Software Heuristics and best practices

The key is that iteration be quick and easy. The faster designers are able to iterate, the better the bots will be. I’m excited by how much innovation is happening in this space, and looking forward to further innovation; the faster and easier it becomes to create a bot, the faster we as designers will be able to find our way to best practices.


We’ve been there before, and intermediate layers between the platform and the developer ultimately produces sub-standard apps and hinders the progress of the platform. — Steve Jobs

Apple was able to explicitly forbid which technologies were allowed on the iPhone. When violations of messenger platform etiquette happen, they won’t be because of technology considerations, they’ll be because bots fail to implement design best practices. As a result, designers will be a key differentiator between a bot that succeeds and one that fails.


Kevin Scott

After ten years of experience building apps for brands like Venmo and GE Healthcare, I'm building bleeding-edge AI software. Join me as I go through the latest in machine learning, web development, and design.

Tensorflow.js and Machine Learning with Javascript