A large bank had rolled out a cookie-cutter scaling framework for 400 people across two sites. In one location, everything was going fine. In the other, the delivery rate had tanked, morale was low, key systems were at risk. What was going wrong?
We started with two days of Initial Assessment, to listen to people and understand their context. We learned that the teams in one location were working on greenfield applications that any of the teams could pick up. In the other location, there were dozens of legacy systems, with decades of technologies, data models, business rules and other quirks.
They had been arranged into ‘feature teams’, so the key people who understood these systems were now in arbitrary teams, each with its own backlog of work. For any work that needed specialist knowledge, the team would be blocked, waiting on a specialist in another team. Meanwhile that team would also be blocked, waiting on a different specialist, and so on. They had built an elaborate deadlock of key knowledge! No wonder nothing was getting done.
To rectify this, we identified the likely work for the next few quarters and used this to run three Demand-Led Planning sessions, in which programme teams of 60-80 people self-organized around the anticipated demand. This meant that the key people were in the right places so they could transfer critical knowledge to other team members.
The context
- 3,000-person core banking function in a global US investment bank
- 400 engineers across two primary locations
- Senior exec told us: ‘We have 3,000 people solving a 300-person problem’
- 40-50 legacy stovepipe systems in one location alone, using a variety of defunct technologies
- Lots of key person risk; only one or two people understood each legacy system
The impact
- Within 12 months, lead time had halved and throughput had doubled—measured by business change items (major features) delivered—across the portfolio
- Quality was not only maintained but improving, measured by number of Severity 1 & 2 production incidents (and time to resolution)
- Key person risk eliminated by ensuring each system had multiple people who could confidently work on it
- 12 months later, throughput had doubled again, and lead time halved again, with the same people, same tech, same systems (mostly), and almost no attrition, because people were having fun now!
The approach
We educated managers and senior execs about flow, systems thinking, intent-based leadership, Theory of Constraints.
Leaders challenged the teams to: ‘Halve lead time, double throughput, without reducing quality’.
Everyone started asking ‘What is lead time?!’.
Some teams found it easy to halve lead time, some impossible, but they all tried.
Some teams tried alternatives to Scrum, out-performed practically everyone else.
We introduced quarterly ‘big-room’ planning on both sites, and adopted fluid teams in one (‘moving the people to the work’), which unblocked most of the dependencies of scarce skills and product knowledge being in the wrong teams.
We introduced portfolio-wide delivery metrics using Cumulative Flow Diagrams, to identify where work was getting blocked, and what to do about it.
We built a league table that showed which people were learning about which key systems.
We created a ‘graceful exit’ for folks who did not want to come on the journey with us, either to move elsewhere in the bank or to provide résumé and interview training to exit well.