Collaboration and Cross-Team Agreements
4 min read
Lesson objectives
- The learner turns a cross-team delivery dispute into a working agreement and a coordination decision whose effect can be measured.
Please sign in to use all the tools
Once you sign in, the reading tools are yours: larger or smaller text, typeface, line spacing, focus mode, highlighting and bold, and notes saved to your account and gathered in My study.
Collaboration sounds like a general virtue, but daily work makes it a decision about who needs to know what and when, and what happens when views conflict. A collaborative team starts with a shared goal everyone can test, not an inspirational sentence on a wall. If the cycle goal is to let a customer book a confirmed appointment, the interface team and the scheduling team do not own separate successes. The work is not done when the interface sends a request unless the customer receives a correct confirmation. Put the completion boundary into a working agreement: who reviews the contract between services, when a change is announced, and how a blocker beyond one person's control is raised. The agreement has value only when adherence can be observed and the rule can change after evidence of a flaw.
Build the agreement from recent friction rather than asking whether everybody trusts each other. For each example, locate the handoff, delay or rework, and affected user. Ask participants for a behavior they can perform over the next two weeks: review a sample booking request before interface development, publish contract changes in a shared channel, or run an integration test with every increment. ‘Improve communication’ alone specifies no owner, cadence, or evidence of success. Nor should an agreement make the team wait for a manager's permission on every small decision. State the decisions the team owns, the risk limits that require escalation, and how reasons are recorded when views differ.
The coordination mechanism depends on team count and dependency type. When two teams exchange changes to one service contract, two representatives, brief synchronization when needed, and a visible decision record may suffice. When five teams depend on a common increment each week, a regular cross-team meeting focused on dependencies and impediments may be justified; it should not duplicate each person's daily meeting. The name Scrum of Scrums does not prove a structure fits. Measure whether decision waiting time and merge surprises fall. If the meeting becomes only a reporting channel to a manager, it has missed the coordination problem. For distributed teams, put the rule somewhere everyone can read and set a response window; reliance on an incidental conversation excludes absent colleagues.
Disagreement within a team is not automatically a failure. Ask a dissenter to describe the risk with evidence and give alternatives a bounded comparison. After a collective decision within safety constraints, members support the agreed experiment and record its review criterion rather than reopening the decision in every meeting. An objection revealing a privacy or safety breach is different: raise it through the agreed escalation boundary instead of hiding it under commitment. In a retrospective, inspect whether the agreement reduced rework and whether an essential skill is isolated in one person. Teaching a second member to run the integration test may remove more delay than adding another meeting. Collaboration then becomes an operating capability measured through delivery and learning, not a personal adjective.
Compare periods before and after an agreement using the same defect definition. If there were six incorrect confirmations in two cycles before a state dictionary and two in two cycles after, the count fell by four. Do not credit the dictionary alone without checking booking volume and test-data changes. Watch how long the interface team waits for a scheduling-team decision as well. Errors might fall while waiting rises because every minor change now needs a large meeting. If so, adjust delegation: let the teams decide changes that cannot affect a customer message, and escalate only semantic changes to product. Coordination should not become a new bottleneck. Finally, test the agreement with someone who did not write it: can they tell from the record alone what ‘confirmed’ means and when to escalate? If not, it still depends on individual memory.
Lesson summary
The learner turns a cross-team delivery dispute into a working agreement and a coordination decision whose effect can be measured.