What is organisational design?
And when a business needs it.

The elements of a design

Five components make up any organisational design. Most redesigns fall short because they change the first and leave the other four as they were.

• Structure. How people are grouped, and who reports to whom. The part everyone means when they say reorganisation, and the least important of the five on its own.
• Roles. What each seat owns, and where one role ends and the next begins. Overlaps here create the conflicts that look like personality problems.
• Decision rights. Who decides, who is consulted, and who is merely informed. The element most often left implicit, and the one that causes most delay.
• Processes. How work actually moves between groups. A new structure laid over old processes reproduces the old behaviour inside new boxes.
• Capability and culture. Whether the people in the roles can do them, and whether the way the organisation behaves supports the design or fights it.

A chart is the output of the first element. A design is a decision about all five, which is why redesigns that begin and end with a chart change so little.

The symptoms that trigger a redesign

Organisations rarely redesign because the structure looks untidy. They redesign because something has begun to hurt, and the same handful of symptoms recur.

• Decisions take too long. Usually a decision rights problem rather than a structure problem, and frequently solvable while everyone stays in post.
• The same work happens twice. Two groups have overlapping remits and each has concluded the other is doing it badly.
• Everything escalates. A sign that authority sits higher than the information does, so the people who know have to ask the people who decide.
• Strategy stalls at execution. A plan that requires a capability the structure has nowhere to put it. New markets and new products produce this reliably.
• Senior people leave. Often the first measurable symptom, and usually read as a pay problem when the cause is scope.

Spans and layers

Two numbers describe the shape of any organisation, and both carry real costs at either extreme.

Span of control

How many people report to one manager. Narrow spans, around three to five, suit complex and interdependent work where a manager has to be genuinely involved. Wide spans, ten and above, suit standardised work with capable people, and they force managers to delegate, because the alternative runs out of hours.

Layers

How many levels sit between the chief executive and the front line. Each layer adds cost, dilutes accountability and slows information in both directions. Each also adds a step on a career path, which is the reason layers accumulate quietly in organisations that promote by creating a level rather than by widening a role.

The two numbers trade against each other, and the correct pair varies with the work. Widening spans to remove a layer succeeds only where the remaining managers can carry the load, and where they have been told which decisions they now own.

Decision rights are the cheapest thing to fix

Of the five elements, decision rights are the least expensive to change and the most frequently left alone. A structure change moves people and unsettles a company for two quarters. A decision rights change moves an authority, and it can be done in a fortnight.

The exercise is straightforward. Take the fifteen decisions that most often stall, and for each one name four things.

1. Who decides. One person, named by role. Where the honest answer is a committee, say which committee and what its quorum is.
2. Who must be consulted first. Distinct from who holds a view, and deliberately short.
3. Who is told afterwards. Usually the longest list, and the one people confuse with consultation.
4. What the limit is. The threshold in money, headcount or risk above which the decision moves up a level.

Published and held to, this resolves a surprising share of what looked like a structural problem. It is also the honest first step before any redesign: where clarifying authority fixes the symptom, the structure was probably adequate all along.

How a redesign actually runs

A structured redesign runs in five stages, and the discipline is in finishing each before starting the next.

1. Establish what the design has to deliver. Derived from the strategy, stated as outcomes rather than as principles. "Faster decisions on pricing" is usable; "greater agility" is decoration.
2. Diagnose the current state. Where work actually flows, where decisions actually get made, and where the friction sits. This rarely matches the chart.
3. Design the target. Structure, roles, decision rights and process together. Test it against the outcomes from stage one by walking a real decision through it.
4. Plan the transition. Who moves, who is at risk, what is communicated and in what order. This stage consumes more effort than the design itself.
5. Implement and hold. A new design reverts to the old behaviour inside a quarter if the decision rights are left to settle on their own.

Design the structure before the people, mostly

The orthodoxy is to design the organisation the strategy requires and then place people into it. It is right more often than the alternative, and it is applied more rigidly than it deserves.

Designing around existing people produces a structure that fits the current team and gives way when one of them leaves. Designing in pure abstraction produces a chart nobody in the building can staff, followed by a year of hiring against it and a quiet retreat.

• Design the target state from the strategy alone, leaving names for later.
• Then test it against the people you have, and record where the gaps sit.
• Where a gap is at the top of a function, fill it before designing the layer beneath. The person who will run it should shape the roles under them.
• Where a gap can be closed by development inside a year, plan it as development. Where it runs deeper than that, plan it as a hire with a lead time attached.

Why most redesigns disappoint

Four missteps account for most of it, and all four are decisions rather than accidents.

• Only the chart changed. Boxes moved while decision rights stayed where they were, so the same approvals happen in the same sequence with different titles on them.
• The transition was rushed. Announcing a structure before the roles beneath it are defined produces three months where nobody is confident what they own.
• Nobody owned the after. A design holds only while someone is watching whether it is being followed. Attention moves on, and the old pattern reasserts itself.
• It was designed around a person. A structure built to accommodate one individual becomes a problem the day they move, and everyone can see why.

The common thread is treating a redesign as an event rather than as a change that has to be held in place for two or three quarters before it becomes the way things are done.