The triumph of product-centered versus user-centered priorities.
Sometimes all you need is a few buckets—you know, working categories named for observed patterns—to begin making sense of a broader thing.
This dynamic is emblematic of the yawning opportunity I see, well afield of notifications themselves, in designing for personalization more confidently and successfully going forward. (It’s why I started Bucket this year. More on that another time.)
UI as high-wire act: Notifications as friend and foe alike
Push alerts are the stalwarts of mobile, soldiering at the front lines of user experience: informing, nudging, or cajoling us to take notice or act on them. They are the natural evolution of the unbundled content experience gone orbital from our old containers. Today, they are widespread, nearly ubiquitous. The closer you look, the more places you find them: in an email, in or outside of apps, or in messaging interfaces, to name the most common settings.
No two brands may deploy notifications in precisely the same manner, but there are prevailing commonalities. How I find myself reacting to them, with seeming consistency, feels telling about the nature of what I call personalized product opportunity: how to implement personalized touches in a way that unquestionably advances a UX.
This is trickier said than done in the notoriously fickle terrain of push alerts, in whatever channel they come to us: user value comes down to preference, behavior and context. Your mileage and mine may certainly vary, both in matching products and across the range of different ones we use.
It feels important to acknowledge that the high-stakes, divisive reaction to notifications—not to mention community events like Betaworks’ Notifications Summit—has pushed thinking along fairly rapidly, even as others bemoan the current state. Yet for all this complexity—and for the budding thinking out there among product wonks (see an appendix below)—there are two kinds of notifications schemes I’ve come to see.
Defining raw-or-cooked design patterns
I call them the raw and the cooked (taking after Blake Eskin’s memorable coinage, for the prevailing and dissimilar modes of podcast production style, not to mention Claude Levi-Strauss’ original construction).
I believe the raw and cooked represent, respectively, the triumph of product-centered versus user-centered priorities.
What that means more plainly is this: a raw notifications scheme is distinguished by its functional, self-regarding and -furthering focus. It may be crude or sophisticated, but its agenda is rarely subtle. While notifications may not come in high volumes (they often do, with this category), they are conspicuously push messaging rather than a response to patterns of user behavior. Or if they do mirror past behavior, they continue insistently in this vein long after the first curiosity-click (LinkedIn’s birthdays notification seems like a good example, which I now believe provides a disable option when it is surfaced).
The cooked notification, by contrast, seems like part of a more considered recipe of overall interactions. First, the volume and type of notification is typically second-order complexity: rather than Quora’s somewhat pitiful “your friend X is on Quora” entreaty, Pinterest sends word of fresh “Pin twins” whose behavior or interests resemble your own.
Complexity, respectfulness and fair critique
None of this is to speak poorly of the more elegant and thoughtful formulations out there, such as Slack’s do-we-don’t-we-notify-you framework—which respects the role both customization and personalization should serve in better understanding context, which is the probably real rainmaker driving user behavior—or Digg’s, which attempts to bucket for the alert-value of news.
It’s particularly unfair to critique breakdowns or ruts in a given app’s notifications regime, especially without a view into their core recipe: the tacit and explicit events that trigger promotions, reminders and the like.
But that shouldn’t keep us from celebrating what clearly works better, particularly over time as a cooked recipe can deliver on consistent value, whereas a raw one becomes even more quickly over-familiar and frustrating with each “new” but rote occurrence.
A successful approach to notifications creates product differentiation at the very heart of the personalization premise: what I want, when I want it,—spinning up results also on the axis of relevance and, in some of the best and most supple cases of personalization, serendipity.
Anything subtracting from this formulation feels like a detriment to long-term success with notifications. Which is why promotions and usage-goosing hacks, when potent and tested, should be deployed nonetheless sparingly and strategically.
Quora and Pinterest: A tale of two notifications schemes
To demonstrate how the raw and cooked are not markers of attentive UX or a case of have-and-have-not product design budgets or proprietary platforms, consider Quora and Pinterest. Both present as well-funded startup unicorns squarely focused on different aspects of a knowledge or taste graph for their users.
Both deliver notifications (in activity feeds, pull-downs and as push messages or email) and do so at a level meeting or exceeding the mark in every sense functionally. They are both superbly engineered product experiences, and none of the critical interactivity with your questions or answers will go unnoticed and un-flagged by a notification in Quora, for instance.
Yet their execution sets them apart.
It feels as simple as the difference between engineering and design. With very low levels of general account activity or monthly active use, both products respond very differently. Quora seems to recognize my inertness and matches me with a general muteness. What it does trickle out, seemingly every 14 days, is a bunk pseudo-metric “your friend X is on Quora, follow them now”—despite them being longstanding users that I’ve not followed for years now. (This reminds of the notorious You May Also Know modules of LinkedIn and, earlier, Facebook, that reliably flagged ex-spouses, ax murderers, professional nemesis.)
Pinterest, by contrast, seems to have taken the deficit of account-based “news” to present me by shifting gears to reflect second-order patterns of potential interest, such as Pin twins and new Boards to follow.
The crucial distinction here is easy to overlook. Pinterest courts me with interests proximate to those I’ve already cited, through the rock-solid magic of a rich taxonomy. Quora goes quiet but for burping up stale data on people I’ve, more often than not, explicitly chosen not to follow. (It would be more commendable if these were new users, for instance, or friends-of-friends with matching interests, or even just anomalous excerpts of their once-winning newsletter digest.)
Cooked notifications privilege a designed experience over one that is merely the sheen of engineering alone.
It’s becoming increasingly apparent to me that there is a raw-or-cooked dividing line in how they represent different approaches to personalized experience. It’s a distinction with a difference, and I believe it’s beginning to matter: either as a point of utility, of differentiation, or of treating my time and attention with the utmost respect it’s due.
Yes, AI will unleash better notifications. But first, notifications need IA
Writing just days ago, Scott Belsky invoked the need for a platform-wide abstraction layer to tame the wildly disparate nature across apps of how notifications are managed. He has a point and I believe he is correct that more granular customization is coming, and that artificial intelligence will serve greatly to curb the cold start issues around determining relevant contexts for notifications. Initially, there will be global controls and sanity parameters, and then things will get more interesting from there.
But AI needs information architecture first and foremost, for that abstraction layer to have hooks to build intelligence. IA closes the knowledge gap by providing a map of potential relationships between users, contexts, and content.
Even rudimentary taxonomical hooks are crucial to the user profiling and content targeting of notifications, and going forward the heuristics of notifications will rely greatly upon the supple semantic richness underlying any notifications scheme, or logic in Belsky’s words.
To unpack that, let’s take a step back to look at personalization as a whole.
“Buckets” bring products closer to personalized product value
I believe these raw-and-cooked gradations may represent a more systematic divide in the design of personalization overall. One that all product and experience professionals need to be increasingly cognizant of as they push forward in seemingly disparate forms of personalized product opportunity, from voice to bots, and from recommenders to augmented reality.
The great hazard of personalized user experience, and no less of the technologies that support personalization, is when they become instrumental, dictating the parameters of one’s experience in much the same programmatic fashion we see in the crudest looping interactions, a logic-based inflexibility becoming dogmatic through overuse in our user interfaces. Instrumentalism is the classic instance of overfitting, and it carries more than a whiff of patronizing disparagement for its user audience.
To make an impact, consider where notifications can be selectively muted or tamped down by settings and context. Notifications we like seem so often a matter of less being more. Foremost, it’s a grave, utmost respect for the singular currency at stake, a user’s time and attention. I believe the more attributes one exposes and acts upon in these settings, the greater the chance a user will be clubbed, intentionally or not, into submission (or more likely, abandonment or uninstalling) by the weight of those notifications, a digital version of death-by-a-thousand-paper-cuts. Using many attributes in support of a recipe for a notification, however? Now you’re cooking.
This begins to explain what I mean about IA before AI.
A first step in this process of designing for notifications is bucketing the notifications being deployed for most users, and looking clearly at what drives value for the user foremost: not for the product, the brand or business. Design is the answer, a design pattern that furthers the desired heuristics and doesn’t foreclose on the capability of the user to further tune and extract more value from product interactions, electively. The worst notification schemes are crimes of omission, of preventing users from enabling or disabling more granularly (e.g., promotions versus functional alerts).
Granted, it’s pedantic to ask whether design or engineering will ride to the rescue of product teams battling the right approach to notifications: nothing is so simple. But it’s unquestionably true that the better our handle on the design patterns of successful notifications mechanics, and what makes them tick, the closer we’ll get to a satisfying peace—with products that prod us, or leave us far too much alone for their own good—and a more lasting solution for all of us shipping these experiences, too.