People Strategy case study

In a Matrix, Draw the HR Org Chart Last

7operating models put to a vote
243people activities, each with one owner
5role-mapping workshops on site, in one week
10KPIs to show whether the model works

The situation

The question behind the question.

In an operations directorate, work runs through processes and projects, across sites an ocean apart. People management ran through the line. A technician could answer to a site, a process and a project, and none of the 3 owned their development. Its new people strategy asked for people excellence in operations, at a moment when the workforce was set to grow from about 170 to more than 200. The question was not whether to change. It was which model, and who would own what once it changed.

As it was asked

“Who should manage our people in a matrix?”

As we reframed it

Choose the model in the open, among every credible option, then decide ownership activity by activity, before anyone draws a box on an org chart.

How we worked

How the work unfolded.

01

Lay out the options

We started from the full range of designs, not from the current one. 7 operating models, drawn from organisation design patterns and from our experience of designing operating models inside and outside HR: people management embedded in a process, anchored in a function or a site, 3 matrix combinations, or segmented by type of workforce. 8 senior leaders were interviewed on the same 14 questions, and each answer was mapped to its closest model. 4 leaned one way, 3 another, 1 a third.

02

Choose in the open

Then the choice, made in the room. In a design workshop, leaders assessed every model on their own first, then voted yes, maybe or no on each. 5 models fell; 2 survived. Neither was enough alone, so we combined them: one people leader for every staff member, with process and project leaders giving input on performance. The steering committee approved it and recorded why each of the 6 alternatives was ruled out.

03

Give every activity an owner

Then the detail, where matrices usually fail. In one week on site, 5 workshops walked the organisation’s own processes and policies as use cases, from hiring to performance reviews. Every people activity got its roles on a 5-letter grid, and exactly one accountable owner. 243 activities in 5 clusters, agreed by consensus in the core team. 5 design principles held the line, among them: grouping follows design, never precedes it.

04

Make it run

The design became 3 packs. Roles: 4 role descriptions, each sized in hours and FTE, each with a section on what the role is not. Governance: decision rights, a 4-level escalation path and an annual operating rhythm. Metrics: 10 KPIs and 5 leading indicators, each kept only if it showed whether the model works, had an owner who could move it, and drew on data from systems live at go-live. A fourth pack, beyond the brief, sequenced the hires.

The instrument

One more letter, one owner

R, A, C, I. The classic grid: who does the work, who answers for it, who is consulted, who is informed.

O. The organisation’s own addition: oversight and support. It owns the process without owning the outcome, the place a matrix usually hides its arguments.

One A. 243 activities, each with a single accountable role, and the next level up named for every escalation.

The org chart is drawn last, from the grid, not the other way round.

What moved

What changed, and what didn't.

243people activities with one accountable role, approved with the model
Evidence Decision taken Adopted in practice Measured impact
What moved

The steering committee approved the model and the 3 packs, and the Director of Operations validated them in writing, calling the governance clear and functional. 243 activities now have one accountable role. Hiring was planned in 3 waves through 2027, for a go-live on 1 January 2028.

What didn't move

The staffing pack was noted, not approved: whether the transition counts as a restructuring under the organisation’s rules still needed a legal answer. How staff are allocated to people leaders was left to the next phase. The effectiveness KPIs need a pulse survey the organisation does not yet run, and the people-leader profile is rare enough to slow the first hires.

What we'd say to a peer

In a matrix, the org chart is the last thing to draw. Put every credible model on the table and let leaders vote before anyone defends a box. Then go down to the activity: one accountable owner each, and a written line on what every role is not. That is where a matrix holds or fails.

Let's talk Discover People Strategy Read more cases
Method note:

Own approach in 4 steps; 8 interviews; 7 operating models voted; RACIO on 243 activities; roles, governance and metrics packs; 3 KPI tests.

WeAreHuman@Work

One piece of evidence on sustainable work, every Sunday. Free, and one click to leave.

Get the weekly edition