Nelson Sarma

Leadership · Episode 02 of 6

Everyone has a backup. Nobody staffs the third layer.

Assigning designers to domains is what every design team does. The part that took me two years to get right is what happens when the backup is already underwater. The goal was never perfect coverage. It was making sure one person’s absence never becomes everyone else’s emergency.

Layers of cover
Primary, secondary, me
Crafts it spans
Five, not one
Voluntary exits
None in 2.5 years

Half of this is table stakes, and it is worth saying so before I describe it. Every design team embedded with cross-functional squads assigns designers to areas. Most of them also have some answer for who covers whom during leave. If that were all this was, it would not be worth an episode, and the pattern isn’t even ours. Engineering has run primary and secondary on-call rotations for decades.

What almost nobody does is staff the layer underneath.

Where a two-layer model breaks

The failure is invisible in a quiet month. The primary owner takes leave, the secondary picks up their domain, and it works, because there is slack in the week to absorb it.

Then a surge arrives. The secondary is already at capacity on their own primary work when the domain they back up lands on them as well. Cover, at that moment, means a second full workload. Nobody refuses it, because it has been framed as helping a colleague rather than as an assignment, which makes it hard to push back on.

What follows doesn’t get traced back to the coverage model. It shows up as a slow decline in quality, or as a good designer handing in their notice for reasons they describe as personal. The model that caused it is still on the wall, still looking sensible, because on paper every domain has two names against it.

A backup plan that quietly doubles someone’s workload isn’t a backup plan. It’s a delayed resignation.

The third layer is the framework

So the overflow comes to me. When a secondary owner is fully occupied with their own work, they are not expected to absorb a second load anyway; I take what doesn’t fit. The manager is the last line, not the escalation path.

That sounds like generosity and it isn’t. It is a constraint I have to plan around: it caps how much of my own week I can commit, and it means I keep enough hands-on capability to actually do the work rather than just route it. That cost is the reason most teams stop at two layers. It is also the only part of this framework I would argue is uncommon.

The two tracks underneath it

Primary ownership

Each designer owns a specific work type or product module outright, assigned on their core skills and where they want to grow. The second half matters more than it sounds, because it turns a resourcing chart into a development plan. That person is the standing design point of contact for their squad. They build real domain expertise, and the squad stops having to re-explain their product every time.

Secondary ownership

Each designer also holds a secondary track on a domain that isn’t theirs. When the primary owner is out, the secondary steps in already carrying baseline context: not a handover document, but familiarity kept warm by proximity.

A coverage table: five design domains down the left, each with a primary owner, a named secondary owner, and the manager as the final backstop. DOMAIN PRIMARY SECONDARY LAST LINE Product Social Web Presentation Illustration Owns it Covers it Me, when both are underwater
Every domain is covered three deep. The third layer is the one that usually goes missing, and it is the reason the second layer never has to do two jobs at once.

What a secondary can do across five crafts

This is the part that gets asked about, and the honest answer is less tidy than the matrix. When both people are in the same craft (two product designers), the secondary can do the work, and cover means what it sounds like. Across crafts it cannot, and pretending otherwise would just relocate the failure.

So cover is scoped rather than equivalent. What the secondary owns during an absence is continuity: hold the context, keep the squad unblocked, answer what can be answered, ship the work that is within reach, and triage the rest into what can wait and what cannot.

Concretely: the illustrator who backs up presentation can build a deck to the existing system, because the system is written down and the components exist. What they cannot do is invent the illustration direction that deck needs. That call either waits for the primary or comes to me, and deciding which is the secondary’s actual job, not producing the artwork.

Naming that limit stops the model overpromising. A squad that thinks it has a full replacement for two weeks is more disappointed than one that knows it has continuity and a clear escalation.

What it changed

The immediate effect was the end of the cognitive exhaustion that comes from bouncing between unrelated tasks all week. Designers got a domain deep enough to have opinions about, and squads got a face they recognised.

The measurable effect was continuity: design delivery across the business no longer stalls because one person is on leave.

Nobody has left in two and a half years, and I would be overclaiming if I said the matrix did that. Twelve people over that span is a small sample with a lot else going on in it: pay, projects, the market. What I can say precisely is that the specific thing that used to burn people out, being handed a second full workload with no warning and no way to decline it, has not happened since.

The matrix also turned out to be load-bearing for everything after it. It is the thing the reassignment route in Episode 03 reassigns against, and it keeps the velocity numbers in Episode 04 honest, because if nobody picks their own queue, nobody can pick an easy one.