Atoms vs. Sessions

Originally written in 2016 for the Quora design blog.

As I’ve written, the Quora design team thinks and works in terms of mechanics — the rules and objects of a product. The term is lifted directly from game design, a discipline so similar to software product design that I've wondered why we even distinguish them as much as we do. Nobody sees narrative and documentary filmmaking as wholly different fields, though I admit “non-fiction game design” doesn't have a great ring to it.

Accordingly, some of the most incisive perspectives on products I've encountered come directly from game design texts. The superb textbook Characteristics of Games is dense with insights that require little work to translate to our field. A great example is one of the book's foundational topics: length of playtime, which is broken into several concepts. There are two I find particularly useful:

Atom — The smallest complete unit of play, in the sense that the players feel they've “really played” some of the game (e.g., two possessions in football, or one level in Donkey Kong)

Session — A single continuous period of play (e.g., an evening of play)

The relationship between them is key:

An atom practically has to be shorter than a session: since the atom is the shortest satisfying unit of play, and a session is the amount of time you actually play, an atom longer than a session is not a pleasing experience. In fact, you really need the session length to be a multiple of the atomic length — that is, you want to end the session by completing an atom (if not an entire game). The shorter the atomic length, the easier it is to achieve that, and incidentally to be tempted into playing just a bit more (“let's play one more hand” in a card game is a lot more common than “let's go on one more raid” in an MMO). Computer games with save points that you can't reach before your session ends are one example of failure in this regard. Even if you can save anywhere, where, though, players will prefer to save at a point where they have reached a “logical stopping point” (finished a turn, killed a boss, completed a quest) — that is, finished an atom. In general, “good” atoms are ones that are fairly short, and fairly distinct, so that players can stop when they like and still feel they have had a satisfying experience.

Applied to social products, this framework reveals some useful insights about user experience.

Snapchat has possibly the shortest atoms of any major social app. An experienced Snapchat user can chew through dozens of snaps in sessions lasting under a minute, taking advantage of how quickly the eye processes images relative to text, combined with the app's rapid, one-tap processing interaction.

Twitter enforces short atoms at the mechanics level, with its 140 character limit. Tweets can be strung together, or link to longer-form pieces, but there's still a fairly strong guarantee for a user that opening Twitter will provide a satisfying experience given almost any session length.

Quora's atoms are naturally longer by comparison, given that many answers are multiple paragraphs of text. Longer content leads to more skimming instead of completing each piece of content as you read. Skimming by itself is not typically a “satisfying unit of play” so the risk of a user skimming for an entire session is of particular concern to us.

Let's look at how this framework can help in very practical ways by trying to deliberately shorten the length of a Quora atom.

When people come to Quora to read their feed, an atom is made up of a sequence of steps that only become a “satisfying unit” upon the completion of all of them: 1) deciding which feed item to look at, 2) deciding if the content is worth reading, 3) reading the content. One can imagine focusing on shortening each of these independently:

  1. For deciding where to look, we can rely more on images that help identify worthwhile content, say by increasing the size of the avatars of the users you follow to make them easier to recognize instantly.
  2. For deciding if the content is worth reading, we can experiment with different pieces of metadata to determine which factors help people make good reading decisions: the author's background vs. the number of people who upvoted the answer vs. how long it would take to read, etc.
  3. For reading the content, we can incentivize shorter-form writing, although this is in tension with our focus on answers with substance, detail, nuance, rationale, etc.

Crucially, we can discern between steps the user is taking that are necessary but don't provide satisfaction (1, 2) vs. steps that generate satisfaction directly (3). There are fewer trade-offs inherent in trying to shorten the former, making them natural opportunities for refinement.

Talking about the user experience in terms of their atoms and sessions like this allows for analysis that is more aligned with what designers care about the most: ensuring that users are having a satisfying experience whenever they use a product. Given that sessions in social apps can be just a few seconds long at a time, these ideas may be even more critical for us to understand than it is for game designers.