dsp· by gregor hochmuth

The 3 Loops of Designing for Audience

In Audiences, Designing on 10 August 2008 with 18 comments

In the previous post Why Twitter hasn’t failed: The Power of Audience, I claim that Twitter has been so successful because it gives users a concrete model of who is listening to them — it gives them a sense of Audience.

Designing for this sense of Audience is a powerful tool to create cohesion and sense of utility among users of a service.

Twitter is a prime example for this kind of attachment: it survives countless outages and a slew of alternatives that could all pull users away… but don’t.

So how do you design for Audience?

Feedback Loops: The main ingredient

A system that wants to tap the power of Audience like Twitter needs 3 important feedback loops:

• Users building a model of their audience
• Audiences giving feedback to the people who post content
• Users giving feedback to the system about the people and content that are important to them.

Surprisingly few of today’s services actually close these loops effectively — many of them, including Facebook, have a number of lose ends.

#1 • Users building a model of their audience

A list of friends isn’t enough to create a sense of audience, it’s merely a pre-requisite. What users need the most is feedback (re-assurance) that their audience is 1) present and 2) listening to them.

Follow Me, Follow You. The loop starts with the notification that another user “is now following you” (on Twitter or Flickr or FriendFeed, which are all based on the same passive follower pattern). Note that Faceook’s model of a relationship is symmetric such that both users have to agree on being connected. (What are the implications?)

• Let users know who’s following them – so obvious that it shouldn’t need mentioning
• You’re already getting those countless emails. How could your system send fewer and still be effective? GetSatisfaction sends one update per day, Dopplr send one a week.
• Bonus points: Make a distinction between reciprocal followers, like Flickr: “Yay! grex has marked you as a contact too.” vs. “If you don’t know grex, grex is probably a fan of your photos …”

Being present: In Twitter, that’s is communicated by the other status updates. It signals who else used Twitter recently and who is presumably checking tweets, too.

• A system should clearly highlight the recent activity of other users.
• What’s important is the quick glance perception of quantity and frequency; the full content only matters on second glance when users are actually in “catch-up mode.”

Who’s listening: Retweets, @replies and references in future conversations all reinforce the user’s model of who’s listening on Twitter.

More generally speaking: Feedback that users receive from others is the primary source for building a model of Audience. It’s not the only source, but in this context, it’s the most relevant to consider.

For a blog or marketing campaign, visitor demographics and site analytics would be another relevant source for understanding who’s listening (and how frequently), but that’s beyond the scope here.

• The more, the better: Feedback between users. That’s the whole next section.

#2 • Audiences giving feedback to the people who post content

One of the most powerful ways to make your users feel good is to highlight the feedback and recognition they receive from their peers. It also turns out that this feedback is the best way to build and support a sense of Audience.

Providing different modalities of feedback is key here: Counting views and clicks, getting favorites and comments, etc. They all have different levels of effort and send different signals. Imagine a concert, in which you could only leave written notes at the door — no clapping, no booing.

In Twitter, it’s the re-tweets and @replies that reward the user with social recognition, especially when the @reply is not a “reply” but just a reference to the other person. Tweets that say congrats are often an example.

This is a loop, so it must be closed to work effectively: It’s not enough to provide the ability of giving feedback — the system must also provide the means to easily discover new feedback that users receive.

Which service understood this from the beginning? Flickr.

see this screenshot from 2005

Flickr’s Recent Activity tab was part of the interface since the early days and it answers a crucial question with just one click: What new feedback have I received?

Facebook, FriendFeed and other services all have feedback mechanisms, but none provide the range and easy access that Flickr supports:

• counting # of views
• favorites
• comments
• photo notes
• contributing tags
• inviting a photo/user to a group
• adding the user as a contact
• Your Stats
• and probably more

Facebook’s feedback mechanisms focus mostly on comments and gives no insight into the other activity that takes place around content: who and how many viewed my photos? who re-shared my link? who also become a fan of X? or joined my group Y? Facebook answers none of those rewarding questions. Bummer.

FriendFeed goes further and provides two modalities: comments and likes/favorites, but it doesn’t facilitate the discovery of new feedback. There’s no “You’ve got Feedback” link. FF now sends an email about new feedback (if you haven’t checked FF recently). But what if I “checked” FF and didn’t the see the feedback? Bummer.

#3 • Giving feedback to the system: Who’s interesting to me

The third important loop has users giving feedback to the system about the people and content that are interesting to them. Most of this data can be collected passively over time, by keeping track of page views, click-throughs and other interactions between two users.

To the dismay of some, this concept isn’t supported by Twitter… except for the Device Updates setting. Do you really want to receive SMS & IM alerts from everyone? No. Even on Twitter, some people are more important than others.

Leaving comments and favorites are other pro-active signals that users can give, but because they require extra effort, there will be fewer of them (far fewer than passive signals), which makes them less useful.

Once you have it, use this data to personalize and prioritize the content presented to users. It sounds obvious on paper, but in reality, personalization is often hard and computationally expensive. But you could start with simple steps, such as merely re-grouping and re-ordering items based on personal preferences. It would already be a strong improvement over the linear, time-sorted lists that prevail in many places today.

What closes the loop in this case?
Signaling to the user when personalization actually kicked in and what information was added, removed or re-arranged.

Not surprisingly, Amazon is a master of this feedback by letting customers know why certain recommendations are being made and giving them the ability to tweak those filters for the future:

Your Recommendations on Amazon

One more: The Ego Loop

There is actually a fourth loop that’s less apparent and it has even fewer real-world examples compared to the previous three: the system giving feedback to users about their consumption. Prime examples: Google Reader Trends and FriendFeed stats:

taken from entries on the Google Reader blog and FriendFeed blog

These statistics are often seen as gratuitous, a pastime for only the self-obsessed. But in reality, no data is more interesting to us than the data about ourselves.

Google Reader illustrates how this type of feedback can also be turned into actionable information: by combining a list of least read feeds with a conveniently located “unsubscribe” button, users can easily remove noise and adjust their stream. Users may also decide to change their habits in response to this statistical feedback — for example, someone reading 90% of articles in technology-related feeds may feel compelled to catch up on other things (or go outside).

Important: Guilt-free data only. The bathroom scale presents a daily opportunity for instant data about ourselves. You get the point. Don’t build bathroom scales, build achievement meters and guilt-free indicators.

Focus on rewarding information and on what users actually did, instead of what they missed or failed to do. Don’t say: “You read 2% of your messages”. Instead you could say, “Based on the messages you read, people find interesting are: Mike, Anna and Carl.”

3 + 1 + You

Feedback loops contribute greatly to a user’s sense of Audience within a service. I talked about four such loops here. What other loops exist? What else builds and supports a sense of Audience?