A reorganisation is one of the most expensive things a leader can initiate. People are anxious for months. Productive work pauses while the new structure is debated, announced, and implemented. Some people leave. Some never get back to the productivity they had before. Despite all this, reorganisations are commonly initiated for reasons that have very little to do with whether the org structure is actually the problem. Data-driven org design is the discipline that distinguishes reorganisations worth doing from reorganisations that produce activity without improvement.
Diagnose Before Designing
The first question is not "what should the new structure look like?" The first question is "what specifically is the problem we are trying to solve?" Slow decision-making? Decisions getting made but not executed? Strategic priorities not getting attention? Functions duplicating work? Coordination cost overwhelming functional efficiency? Each of these has different structural implications. Reorganising without an accurate diagnosis produces movement without improvement.
The Data That Actually Helps
Span of control distributions — too narrow signals over-management, too wide signals manager overload. Decision flow data — where decisions actually get made versus where the org chart says they should. Time allocation data — what people are actually spending their hours on, ideally measured rather than self-reported. Organisational network analysis — how communication and collaboration actually flow, often quite different from the formal structure. Customer-facing outcome metrics — does the structure produce the outputs customers experience.
Span and Layers
Two metrics that come up in almost every org design exercise. Span of control (number of direct reports per manager) and number of layers between any individual contributor and the CEO. There is no universally right number for either, but there are typical ranges and warning signs. Spans below 4–5 frequently indicate either over-management or that managers are also functioning as ICs (which deserves explicit acknowledgement, not hidden assumption). Layer counts above 7–8 in mid-sized organisations usually indicate accumulated structural drift rather than functional necessity.
A pattern that data-driven analysis surfaces: the organisation has a stated strategy that emphasises customer focus, but the structure has the customer-facing functions reporting up through three layers of internal-facing management. The stated strategy and the structural reality contradict each other, and people respond to the structural reality. The fix is sometimes structural, sometimes a clearer accountability assignment within the existing structure. Either way, the diagnosis came from comparing intent to actual flow.
When Reorganisation Is Not the Right Tool
- The problem is leadership behaviour — restructure the leadership; the structure is fine
- The problem is unclear strategy — clarify the strategy first; structure follows strategy
- The problem is performance management — fix the performance management; restructuring redistributes the problem
- The problem is capability gap — invest in development or hire targeted; do not restructure the gap
- The problem is ambiguous priorities — clarify priorities; structure cannot solve prioritisation
When Reorganisation Is the Right Tool
Genuine structural problems do exist. A growing company outgrows its functional structure and needs business units. A merger creates duplicated functions that need consolidation. A strategic pivot requires capabilities the current structure cannot host. New regulatory requirements demand functions that did not exist before. In each of these cases, reorganisation is the appropriate response — and data-driven design produces a structure more likely to address the underlying need than instinct-driven design.
Verify After Implementation
A reorganisation is hypothesised to produce specific outcomes. After implementation, those outcomes need to be measured. Decision velocity, time to escalation, customer-facing metrics, employee engagement, voluntary turnover. The verification is not optional — without it, the organisation cannot tell whether the reorganisation worked, and the next reorganisation will be just as data-free as this one was.