Real-time Messaging in the Enterprise: Here We Go Again

There was a good Wired article, published yesterday, that bemoaned the rapidly-growing plethora of communication applications centered around real-time chat. Its author lists consumer-oriented applications to demonstrate the situation:

“I bounce through a folder full of messaging apps. I talk to a few people on Hangouts, a few others on Facebook Messenger, exactly one person on WhatsApp. I Snapchat all those people, too. I use Twitter DMs, GroupMe, HipChat, Skype, even Instagram Direct a couple of times. Livetext, Yahoo’s new app, is fun; I’ve been using that. Oh, and there’s email. And iMessage. And, of course, good ol’ green-bubble text messaging.”

The same problem is beginning to develop within businesses as their employees self-adopt enterprise-first chat tools from startup vendors that have been in-market for a while, including Slack, Hipchat, Wrike, Flowdock and others. Oh, and let’s not forget that many employees use the consumer-grade applications mentioned in the Wired article to conduct business, even if it’s against company policy.
Of course, all of these newer chat tools compete with IT-approved enterprise real-time messaging offerings for employees’ attention and love. IBM Sametime, Microsoft’s Lync and Yammer, and Salesforce Chatter are just a few well-known examples of longer-lived, enterprise-grade messaging applications and services that support real-time exchanges. To further compound the clutter, we are also seeing new chat offerings, from established enterprise collaboration software vendors, that mimic their consumer-oriented cousins. Jive Chime and Microsoft Send are real-time chat apps that have been released in the last four months to support organizations’ increasingly mobile workforces.
There are a few problems created by this overwhelming collection of enterprise real-time messaging options. First, these applications are largely siloed from each other, so employees have to remember in which one a certain conversation occurred or know in which application they have the highest probability of gaining a specific coworker’s attention. Second, some can interoperate with other enterprise applications via RESTful APIs, while others require more costly, time-consuming integration efforts. Third, some messaging applications support information governance initiatives such as records retention and disposal whereas other offerings essentially assume that chats are throw-away conversations that do not need to be archived and managed.
There are so many other issues that they will be better dealt with in another post. But they are bound by one clear fact: we’ve made all of these mistakes with previous generations of enterprise messaging technology.

The BIG Problem: Why?

The biggest problem facing the newest wave of enterprise chat tools is an existential one. It is not clear why they are needed when existing real-time messaging tools satisfy the same use cases. I voiced this in the following mini-tweetstorm on the day that Microsoft Send was announced. (read from the bottom of the graphic to the top)
Larry's Enterprise Chat Tweetstorm
That’s right. You can hold my feet to the fire on that prediction. Enterprise real-time chat is destined to quickly fail as a market segment and technology with significant, positive business impact. Just like the combination of status update and activity stream features in enterprise social software failed to displace email, instant messaging and other, well-established forms of business communication.
Insufficient technology is not the cause of poor communication within organizations. We have had at our disposal more-than-adequate messaging technologies for decades now. The real reason that employees and their organizations continue to communicate poorly is human behavior. People generally don’t communicate unless they have something to gain by doing so. Power, influence, prestige, monetary value, etc.
Well-designed technology can make it easier and more pleasant for people to communicate, but it does very little to influence, much less actually change, their behaviors. So the latest enterprise real-time chat applications may offer improvements in user experience, but they won’t measurably increase communication frequency or effectiveness in most organizations unless their deployment is accompanied by change management efforts that include meaningful incentives to communicate.
I intend to track and chronicle the rise and fall of enterprise real-time chat as part of my research agenda at Gigaom Research. Stay tuned over the coming months as we watch this drama unfold.

Dissensus, not consensus, is the shorter but steeper path

A week ago, I read a Adam Bryant interview with Bob Pittman of Clear Channel, the radio network, in which Pittman makes a case for a role for dissent as a triangulation tool. He was talking about his early days as a new manager, and how he countered the natural suspicions of those older than him by actively encouraging people to present opposing views, and how that shaped his current approach:

Often in meetings, I will ask people when we’re discussing an idea, “What did the dissenter say?” The first time you do that, somebody might say, “Well, everybody’s on board.” Then I’ll say, “Well, you guys aren’t listening very well, because there’s always another point of view somewhere and you need to go back and find out what the dissenting point of view is.” I don’t want to hear someone say after we do something, “Oh, we should have done this.”

I want us to listen to these dissenters because they may intend to tell you why we can’t do something, but if you listen hard, what they’re really telling you is what you must do to get something done. It gets you out of your framework of the conventions of what you can and can’t do.

Pittman’s grass roots embrace of dissent is only the starting point of its utility, and its central role in the evolving laissez-faire management thinking behind the third way of work.

I want us to listen to these dissenters because they may intend to tell you why we can’t do something, but if you listen hard, what they’re really telling you is what you must do to get something done. – Bob PittmanNote that the heading ‘the shorter path’ is based on one of the paradoxes of computing, that algorithms that do more work at the outset generally wind-up being more efficient. Dissent is like that. Avoiding the comfortable traps of groupthink and the cognitive biases that herd us into consensus too quickly is hard work, but leads to better results, we will see.

Ulrich Klocke researched causes of poor group decision making in How to Improve Decision Making in Small Groups: Effects of Dissent and Training Interventions. First, groups may fail to process information effectively:  for example, not sharing information that may be relevant for various reasons, or giving some information only a cursory examination. Secondly, certain well-known cognitive biases can lead groups to poor results, particularly bias in favor of shared information (information known to many in the group), and a bias in favor of initial preferences (people get stuck on what initially occurs to them, and are hard to unstick).

As Klocke wrote about the sharedness bias —

Groups communicate predominantly about information, which all or most group members share before entering the discussion, and neglect unshared information, which only one or few members have initially.

[…] group members individually judge shared information as more important, relevant, accurate, and influential than unshared information. This bias seems to have two reasons: First, shared information can be confirmed by more than one group member. Second, individuals evaluate their own information as more valid than information from other members. Thus, unshared information, even if mentioned in the discussion, is not seriously considered by other group members and therefore has less impact on the final decision than shared information.

— and later, about the preference bias —

Even when all information necessary to identify the correct solution is exchanged during discussion, individual group members often stick to their initially preferred wrong solution. People bias their information processing to favor an initially preferred alternative. Other studies show the same phenomenon at the group level: Group decisions can often be predicted by the initial preferences of its members. If a majority favors a certain alternative before the discussion, the group seldom decides to chose another alternative. Thus, frequently, group discussions are superfluous, and groups would be better off using a decision shortcut like an immediate vote or an averaging procedure.

Klocke has catalogued a litany of cognitive barriers to effective group decision making, and — spoiler alert — the research he conducted supported these biases as being present in his experimental groups, as well as those he cites from the literature. [Note: All citations left out of these quotes.]

Frequently, group discussions are superfluous, and groups would be better off using a decision shortcut like an immediate vote or an averaging procedure. – Ulrich KlockePerhaps the worst news is that simply letting people know about these biases is not enough to counter their impact on group behavior. One thing, however, can positively change decision making: dissent.

The impacts of dissent are positive both at the group and individual level. As Klocke outlined,

[…] early field studies analyzed the effects of groupthink, a tendency for concurrence seeking that effectively suppresses the expression of dissent. They found evidence that groupthink can have detrimental effects on group decisions. […] These experiments showed that dissent (compared to consent) enhances decision-making quality, even when no group member favors the correct solution before the discussion. This effect was mediated predominantly by more systematic processing of information but also by less biased processing of information. Specifically, dissent led to the introduction and repetition of more information and to a more balanced discussion of shared and unshared and preference-consistent and inconsistent information.

With regard to individual behavior in the context of dissent:

There is evidence for more systematic processing by individuals after being exposed to divergent opinions. One factor that mobilizes systematic processing is surprise or a deviation from expectancy. Usually, divergent opinions are unexpected and therefore cause surprise and mobilize cognitive resources to explain the unexpected event. In addition, it has been demonstrated that dissent, especially when articulated by a consistent minority, promotes divergent thinking, a variable related to unbiased processing.

Klocke went on to test whether interventions in group activities — telling them groups about the various biases, but not actually introducing systematic efforts toward dissent — and discovered they don’t have much of an effect. Simply knowing that bias is likely does not counter bias.

On the other hand, the findings regarding dissent are powerful and demonstrate a maxim: the more important a decision, the broader the diversity of opinions that should be sought to apply to the decision, and the greater the attention to active and comprehensive dissent.

A premortem in a business setting comes at the beginning of a project rather than the end, so that the project can be improved rather than autopsied. – Gary KleinBob Pittman’s real-world experience is borne out by the experimental evidence. I think one of the best approaches to inserting dissent is to use project premortems which extends Pittman’s search for dissent at the outset. As described by Gary Klein, a premortem

is the hypothetical opposite of a postmortem. A postmortem in a medical setting allows health professionals and the family to learn what caused a patient’s death. Everyone benefits except, of course, the patient. A premortem in a business setting comes at the beginning of a project rather than the end, so that the project can be improved rather than autopsied. Unlike a typical critiquing session, in which project team members are asked what might go wrong, the premortem operates on the assumption that the “patient” has died, and so asks what did go wrong. The team members’ task is to generate plausible reasons for the project’s failure.

Klein seems to have grasped the biases that tend to push people toward groupthink, and his approach counters them. Sometime after everyone has been brought up to speed on the plan, a premortem is held.

  1. The facilitator starts by saying the project has failed ‘spectacularly’.
  2. Each person writes down — privately — every reason they can think of for that failure, and as Klein says, ‘especially the kinds of things they ordinarily wouldn’t mention as potential problems, for fear of being impolitic’.
  3. In a round robin, the facilitator gets one reason from each team member, who reads it aloud, and it is recorded, until all reasons are covered.
  4. Discussion can be as deep or as shallow as the facilitator needs.
  5. After the meeting, the facilitator and team leader can refine the list, and respond with an action plan to resolve issues, and then reconvene the team to make decisions on ways to revise the project plan or approach.

This intentionally sidesteps the sharedness and preference biases to at least a reasonable degree, and also can increase the deviation from the expected, given that a diverse set of viewpoints are brought into the premortem.

The biggest takeaways are these. We have deep cognitive biases that negatively impact group decision making, especially when group decision-making is not structured to counter those biases, and does not exploit the value of dissent and diversity. Techniques like premortems are one tool to help increase the leverage that dissent offers, and avoid the soft pitfalls of groupthink and unwarranted consensus.