Beyond the Org Chart: Why Your SRE Team Needs a Membrane, Not a Beast

In 2023, my organization stopped working. Not gradually, but with the suddenness of a system struck by a series of uninterrupted changes.

We had just received several purchases, each bringing its own definition of urgency. Our engineers were drowning. TOIL—repetitive, manual, interruption-driven work that destroys engineering value—increased to 83.9%. We kept running, but nothing moved.

This fall was especially painful because it followed years of hard-earned progress. Each previous combination was absorbed faster than before—two years, then one, then six months. The frame was working. Then it wasn’t like that. We didn’t get there by deploying a new visibility stack or adopting a trendy scene framework.

We did it by rebuilding something that sits between our engineers and external chaos. It’s a concept that most SRE groups don’t openly mention.

I call it Membrane.

Org Chart Legend

Many organizations view governance as a safety net. They are wrong. Niklas Luhmann, a sociologist and systems theorist, rightly pointed out that organizations are not pyramids of power; they are communication systems defined by their boundaries.

In the elite world of SRE, the organization chart is a myth. Hierarchy tells you who reports to whom, but hierarchy tells you what the organization actually delivers—and therefore, what the organization really is. To survive, you have to stop building silos and start building networks.

The silo is a wall; it is impenetrable, creates problems, and promotes “not my problem” cultures. The membrane, however, is a relatively permeable filter. It separates important signals from overwhelming noise. Gatekeeping is not an administrative barrier designed to reduce population; it is a life support system. It protects builders from disruption while still being able to tap into real, proven needs.

The membrane is not a single gate. Systems maintain identity through borders—plural, each with its own measure. Another sound filter; others exchange people, manage the accountability of partners, or hold meetings. The following describes the first.

Your imaging board is like an X-Ray

In our context, we use this with virtual boards that take the triage criteria as the input device settings.

Your dashboard is not a productivity tool. It’s an x-ray of your membranes. A team whose entry board looks like a parking lot of suspended cards has a very strong membrane. The group whose pickup board looks like a firehose has no membrane at all. No team fails because of their ticketing tool. They fail because no one has taken responsibility for the mechanical settings of the filter—the criteria that determine what gets in, in what form, and to whom.

This is where we adopt the “Olivetti” idea: team performance cannot be measured by the output index alone. Adriano Olivetti understood that the team is a community to be cultivated, not a resource to be perfected. Preventing fanaticism is a moral obligation, and the membrane is the structure that makes that cultivation possible. By protecting the developer’s attention, we protect their reputation and their ability to do deep, meaningful work.

Breach 2023: A Study of Estimating

The membrane is a living thing that needs constant maintenance. Our crisis of 2023 happened due to unforeseen circumstances.

As we integrate new discoveries, we try to absorb new products and cultures—with their unwritten tribal knowledge and manual processes—without recalibrating our filters. The result was a breach of our operational integrity. We had to go back to maturity. The frustration was palpable: We had solved this before; What else were we solving it for?

The recovery took us to 2024 and into 2025. The membrane frame did not prevent the problem, but it allowed us to use it. We used the peak value of 83.9% of TOIL as the data input required to retune our filters. Under Google’s strict 5-point TOIL definition, we moved TOIL from 59.7% in 2024 to 44.7% in 2025 – back below the SRE health benchmark. We’ve compressed our P95 cycle time – the true pulse of the fast movement – from 294 snow days in 2020 to 57 days in 2025. It proved an important principle: an unbalanced membrane does not exist at all.

Frontier Engineering

The SRE industry has spent a decade developing the “inside” of membranes. We have excellent visibility, automated runbooks and flawless autopsies. The craft in that layer has grown.

But the border itself—emergent, reversible, deterministic—is often treated as a “soft” task. We dismiss it as “people stuff” or office politics. I found out that getting fired is incredibly expensive. Treating the boundary (or filter) as something less than a first-stage engineering problem is how teams drown.

I challenge you: Open your inbox tomorrow morning. Don’t look at it as a ticket list, but as a live x-ray of your lining. Ask yourself:

• What application did you approve this week that didn’t meet the criteria?
• What did we prevent that should have been an emergency escalation?
• Who paid the price for that measurement error, the engineer, or the petitioner?
• Do we protect systems or allow groups?

If the answer is “I don’t know,” you’ve found your next engineering project. Estimating is not an “extra” job; it is the only function that ensures the survival of your system.