Sooner Safer Happier: Antipatterns and Patterns for Business Agility

by Jonathan Smart

PATTERN 7.3 Architect and Organize for Flow

There are many ways to break dependencies, including, as above, moving from a monolith to independently testable and deployable components, service virtualization (to mimic the work having been done on a dependent service), feature toggling (so that the work can be shipped, toggled off), shared code ownership, and in some cases seeding developers onto other teams (alleviating human dependencies). Your choice will be unique to your context. Most important is the capability to inspect and adapt and continuously improve. Habitual behaviors and the support you have is important in building meta-level capability.

PATTERN 7.4 Smart People and Smart Teams with Robot Friends

Tools can help, but cultural change patterns such as unlearning and relearning, safe-to-learn experiments, inviting innovators per the Diffusion of Innovation curve, and generating social proof are needed to bring lasting success. The focus is people process tooling, in that order.

you don’t change culture or mindset before behavior; you change behavior first, typically in a small way—“act your way to a new way of thinking,” and the mindset will follow.28 You lead by example—not by mandating a process, but by sharing or suggesting a practice in the context of the people.

The trick for organizations is training and nurturing talent in testing specialist roles within a product development team. This quality assurance work should be focused on people using their creative mind, not performing repetitive tasks that a robot can and should undertake.

ANTIPATTERN 8.1 Information and Learning Silos

estimated that as much as 50% of information is lost in every handoff.4 That means that by the time the work has undergone just four handoffs, that recipient is getting just 6% of the knowledge associated with the work.

ANTIPATTERN 8.2 Outputs over Outcomes

ANTIPATTERN 8.3 The Bubble Effect

ANTIPATTERN 8.4 Applying a Deterministic Approach to an Emergent Domain

ANTIPATTERN 8.5 Weaponized Metrics

PATTERN 8.1 Optimize for Learning

the cycle of unlearn and relearn in three steps: unlearn, relearn, and break through.13

To optimize for learning, organizations must learn when knowledge is tacit versus explicit, and its implications on which practices to apply. They must identify and break silos, enable self-organizing teams, and establish channels for dialogue.

the Cynefin Framework. That framework can also be applied to knowledge as described in Figure 8.2. It provides a useful guide to how to optimize for learning depending on the domain. There is explicit knowledge on the right, which are the “ordered domains.” There is tacit knowledge on the left, which are the “unordered domains.”

In the Clear domain, workers follow the rules and apply best practices. This is the domain of explicit knowledge, where learnings can be shared through simple instructions, scripts, and guidelines—such as building a Lego model. The Complicated domain requires experts who analyze and respond. This is not child’s play. Work in this domain has been done many times before. There are known-unknowns. This is a domain of explicit knowledge where learnings can be documented and handed over even when they require plenty of detail. The Complex domain is where people probe and respond. Work in this domain is unique, or the context is unique. It is emergent and there are unknown-unknowns. Workers need to apply a probabilistic approach and test hypotheses in order to learn. This is a domain of tacit knowledge. As we saw in Antipattern 8.1, tacit knowledge cannot travel by handoffs without significant loss of information, of learnings. In this emergent domain, skill-based silos present a strong antipattern. Instead, optimized learning takes place with a bias to action, by doing and co-creating in order to avoid a single person dependency. The Chaotic domain has a blurry line between cause and effect. The rule here is to act first and hope something good happens. Any learnings generate tacit knowledge. Confusion is when you are not sure what domain you are in. In this case, attempt to create a smaller domain of work. The application of Cynefin in the context of information and knowledge ...

Value stream mapping can help identify the queues, handoffs, and information and learning silos that produce those blockages.

Representatives from each state of the workflow, from concept to production, should take part in the value stream mapping. Attendees can ignore their own cycles, whether they’re sprints, flow-based, agile, or waterfall, and focus only on the flow of the work and the steps needed to go from left to right to complete production.

The session installs transparency of silos, queues, impediments, and a number of improvement items to act on. It typically highlights bottlenecks, long wait times, and too much WIP. A typical outcome is the need to shift activities left, to build them in, collaborating early and often.

Glenda Eoyang, founder of The Human Systems Dynamics Institute, has developed a theory and practice of human systems dynamics. Her model for self-organizing systems has three dimensions: containers, difference, and exchange.

Containers: Long-lived, nested value streams with clear-cut strategic intent and shared strategic objectives, goals, business outcomes, shared beliefs, and mental models. Optimized for the fast flow of safe value and aligned to the customer. There should be high cohesion (do one thing well) and low coupling (minimize dependencies over time). Containers are interdependent services.

Difference: Diversity of skills, knowledge, experience, gender, and cultural background. This should be evident in multidisciplinary teams.

Exchange: Communication, feedback loops, and flow of inform...

4: Dialogue William Isaacs, co-founder of MIT’s Organizational Learning Center and the director of the Institute’s Dialogue Project, has described dialogue as the embrace of different points of view, literally the art of thinking together.19

PATTERN 8.2 Nested Learning with Built-In Feedback Loops

individual level, the team level, and the organizational level at the same time. The absence of any of these layers prevents the creation of a learning organization and keeps knowledge siloed inside learning bubbles. Focusing on outcomes with nested learning loops promotes organizational learning and continuous improvement.

Team events like daily coordinations, standups, weekly reviews, demos, retrospectives, and monthly show-and-tells are all team learning sessions.

Besides learning about outcomes, these sessions are the mechanism to learn about team processes, policies, ways of working, their effectiveness, and what to improve.

Creating a Dojo, a place for immersive learning,

There are many ways to share learning from individuals and teams across the organization as a whole. These include internal meet-ups, conferences, unconferences, webinars, brown-bag sessions, and show-and-tells. It is good to recognize and reward this behavior.

Unconferences are participant-driven sessions. Participants drive the agenda, bring their own ideas, and vote on the ideas that they discuss and develop. A popular style of unconference is Lean Coffee, where attendees write down their ideas on sticky notes, dot vote, and discuss the ideas with the most dots. The discussion is time-limited and participants can vote to continue discussing the same item for another period of time or move on to the next item with the most dots.

Open Space is another popular form of unconference. Like Lean Coffees, attendees bring their own ideas and topics to discuss. There is an empty timetable on the wall or flipchart with the timings (see Figure 8.6) and ther...

Four principles drive Open Space: whoever comes are the right people, whatever happens is the only thing that could have happened, when it starts is the right time to start, and when it’s over, it’s over. The rule is to go where you want and where you feel you can contribute and learn.

PATTERN 8.3 Communicate, Communicate, Communicate

Storytelling, case studies, and experience reports are all great for organizational learning. But unless we surface these stories and their insights, great things will remain in their bubbles and will struggle to bring benefits to other teams and to the organization as a whole.

So we introduced the Better Value Sooner Safer Happier Awards. Submissions were open to any teams across the organization regardless of size or domain, whether IT-related product development or not. Over two weeks, we had over eighty submissions, considerably more stories than we had collected previously over a much longer time span.

Teams needed to provide factual evidence, data, and quotes from customers and colleagues to show how they had improved on agile, BVSSH dimensions: •Better: Quality, such as a reduction in production incidents, faster mean time to recovery, and increased resilience. •Value: Measured by quarterly business outcomes, with leading and lagging indicators, such as market share, customer retention, referrals, revenue, carbon emissions, and diversity. •Sooner: A reduction in end-to-end lead time, an increase in throughput of items of value, and an improvement in flow efficiency. •Safer: Continuous compliance so that the team was agile, not fragile. Examples of improved Safety, with leading indicators such as proactive work done and lagging improvement such as a reduction in control or compliance-related issues. •Happier: Shown through an increase in NPS, retention, and referrals, as well as improved outcomes for citizens and for the climate.

Communities of Practice, or CoPs. These are internal to a company, voluntary, open to all, and Darwinian. If people meet, there is value. If they don’t, there isn’t. CoPs are not artificially kept alive. They may be aligned to a Center of Excellence, Guild, or Practice. The key is that they are open to all. This helps identify the natural champions, the innovators, the passionate people, the rebel alliance.

a successful CoP is made up of two co-chairs and a small group of core contributors (fewer than ten people). Chairs and core contributors ensure that there is a refined backlog of show-and-tells, both internal and external. There is an open invitation to everyone and anyone; those who turn up aren’t always who you expect, helping make the company smaller and increase shared learning. And it’s okay for people to be in listen-only mode. Use social media to help facilitate sharing and finding experts. A good cadence for a CoP is meeting at least once every two weeks, a regular heartbeat. This is very important. The meeting could be a Lunch-and-Learn or it could be a meetup at the end of day that goes into the evening with an optional social activity such as drinks and pizza. For global organizations, start out in one location, getting the CoP working primarily in person if possible and with the ability to join virtually. As it grows, have regional chapters to build community. In person meetings are preferable, in addition to dial-ins. Build networks and relationships. Have a roadmap of quarterly outcomes so there is an agreed North Star for each quarter aligned to longer-term organizational North Stars, ideally. Also, bearing in mind that it’s voluntary, anything done is a bonus. Each CoP should have a senior leader sponsor, for support, to act on ideas and to pass on top table challenges, concerns, or questions. Finally, have an overall Executive Committee sponsor who can he...

The loop begins by clearly calling out Align, to ensure that there is clear alignment to the higher level intent, goal, North Star, mission, or outcome hypothesis. This is the first step of the learning loop—it installs vertical transparency (as we saw it in Pattern 5.3), and it connects the otherwise disconnected bubbles of strategy and delivery teams. Determine your aligned hypothesis that you want to learn about here. Once aligned, it’s time to Sense the context, the customer, terrain, history, behavioral patterns, antipatterns, underlying mental models, values, and one’s own learning so far. It is fundamental to learn and understand the starting point, the unique context. As G.K. Chesterton is attributed as saying, “Don’t ever take a fence down until you know why it was put up.” Start where you are. Subscribe to learning from others, and sense the outside world to learn new potential patterns of success. Next Respond by designing one or more experiments in order to test the hypothesis. These could be new or they could be the amplification or dampening of previous experiments. Then, run the Experiment and Distill the results to generate insights and metrics. Discuss it with participants in the experiment; run a retrospective or debrief to identify learnings. Finally, Share by publishing the learning and updating doctrines and patterns. This enables others to pick it up when at the Sense step. This final Share step in the loop is essential, and I find it is rarely done.

PATTERN 8.4 Be Comfortable with Uncertainty

PATTERN 8.5 Measure for Learning

Better Value Sooner Safer Happier measures include: Better: Better is quality. Depending on your context, quality measures could include system outages, time to recovery, straight-through-processing exceptions, error rates, reconciliation breaks, rework, and so on. Value: Value is unique to your business and is measured via OKRs, specifically the KRs, which are leading and lagging value measures. This could include revenue, market share, profit margin, site visits, new customer attraction, customer retention, trading volume, diversity, and more. Sooner: Sooner is time to market. It’s lead time from starting work on an item of value to getting it into the hands of a customer. It’s throughput, which is the number of items of value over a given time period (which goes up as lead time comes down), and it’s flow efficiency, which is a measure of the percentage of time that work is waiting. This is one of the most important things to focus on. Look where the work isn’t. Alleviating impediments to flow enables lead time to reduce and throughput to rise. It increases “value-tivity”—the highest value in the shortest time.

Lead time distribution, as per Figure 8.9, is a useful visualization. It shows the number of days from when the work started on an item of value until it is in the hands of a customer (x-axis) versus the number of times that that same lead time occurred (y-axis). The curve typically shows a Weibull distribution, which is like a normal distribution skewed to the left. There is usually a long tail of long lead times. In this histogram view, it’s recommended to measure the 85th percentile. This means that with empirical evidence, any similarly sized item of value (ideally all are roughly the same size, small), there is an 85% probability that it will be done within that amount of time. Plot this chart regularly to see how its shape is changing. When the 85th percentile point moves to the leftover time, then you are delivering value Sooner to customers. (See Figure 8.9.)

Useful metric