The gate that never skips: building the quarterly planning system
A public company's planning ran on gates that stopped nothing and a say-do ratio built on a 'say' nobody disciplined. I rebuilt commitment as a system: three monthly gates, one artifact lifecycle, and enforcement that doesn't need a police officer.
The setup
On paper, the planning machinery all existed. A six-phase development lifecycle. Ceremonies on the calendar. Checkpoints with names. From a distance it looked like a mature operation.
Up close, none of it had teeth. Checkpoints didn’t stop work. Compliance couldn’t be measured. Security design reviews were skipped almost universally, and nothing in the system was even capable of noticing. Gates were status meetings with no decision on the table, no decision-maker in the room, and no data behind the slides.
The cost landed on the people furthest from planning. Commitment was a surprise to the people who had to act on it. Go-to-customer teams learned what was shipping too late to plan launches around it. Leadership reported a say-do ratio to the board, but the “say” side of that ratio had never been disciplined, which made the number a story rather than a measurement.
This is a different problem from having no process. Having no process is honest. Having process theater is worse, because everyone gets the cost of ceremony and none of the protection.
Where I came in
I own product operations for the R&D org, and I had already built the unified development lifecycle underneath this work (that’s its own case study). This case study is about the layer above it: how a quarter’s worth of intent becomes a commitment the whole company can plan against. Nobody chartered that project. I made the case that planning quality was an assertion, not a measurement, and then I built the model that made it measurable. I designed it, documented it, socialized it, and ran it across product, engineering, design, security, and go-to-customer teams with no formal authority over any of them.
Two convictions shaped the design.
Discipline is a system output, not a character trait.
The old model made hygiene expensive, invisible, and optional, and it got exactly the data quality that combination buys. I refused to build anything that depended on people being better. The system had to make the disciplined path the cheap one.
The second conviction: a gate you can skip is not a gate. Whatever I built had to be always possible, because always possible is the only thing that makes always required credible.
What I built
The quarterly planning system. One planning quarter, three monthly gates, one artifact lifecycle. The gates give the quarter a heartbeat: a planning summit to set direction, a design freeze review to lock scope, and commitment week with a readout that go-to-customer teams can actually build launch plans on. Every initiative, in any zone of the portfolio, reports its state the same way.
The artifact lifecycle. Every required artifact carries an owner, the phase it’s initiated in, the phase it must be complete by, and the gate it blocks at. Dependencies are structural, not advisory. Artifacts move through five canonical states, and “Unknown” is deliberately a distinct state from “not met.” That single design choice surfaces data-quality debt instead of letting it hide inside optimistic defaults.
One hard rule as the single enforcement point. No complete requirements doc, no Ready to Implement. That’s it. Everything else in the model is visibility rather than policing. I chose a single hard rule over a rulebook because every additional hard rule is another place the model can be caught being ignored, and a model caught being ignored loses all of its rules at once.
Tiered gate profiles. The ceremony runs at three sizes, from the full multi-day summit down to a bare-minimum session that preserves the decision, the decision-maker, and the readout. The ceremony scales; the gate never skips. There is deliberately no tier zero, because skipping must never be a valid state. Compression is earned through maturity, never decreed by whoever is busiest that month.
The enforcement redesign. The old model ran on person-polices-peers, which meant enforcement was interpersonal, awkward, and therefore rare. I replaced it with ambient visibility computed from artifact data: readiness is calculated, and humans decide what to do about it. Nobody has to accuse anyone. The dashboard just says an initiative is a third of the way ready, and the conversation starts from a number instead of a confrontation.
Making it stick
A planning system lives or dies on whether leadership actually plans by it, so I treated executive attention as a design surface.
When an executive reviewer said the strategy material was too much, I delivered an operator’s one-pager the same day. That one-pager became a five-slide planning deck built in under ten minutes, and within days a seventeen-slide three-gate operator’s deck with dated ceremonies and explicit approval states, which became the org’s first planning calendar of record. Depth and accessibility turned out to be two products of one canon, not a tradeoff. The deep documentation exists for the people who run the model; the surfaces exist for the people who plan by it.
The model was also built to be inherited. It was documented specifically so an incoming program-office executive could pick it up and run it, which is a management artifact whether or not a management title is attached.
Impact and results
The model survived an org restructure and a leadership transition. Because it tracks work rather than org charts, a reorg changes who attends a gate, not what the gate asserts. The org’s duplicate tracking systems were retired against it.
The vocabulary became the org’s shared language for planning state. Gates, phases, and artifact states are how initiatives get discussed now, including by leadership, unprompted.
Security reviews went from afterthought to architecture. What was once skipped almost universally is now a tracked, gate-blocking artifact. The system notices, which is the precondition for everything else.
The say-do ratio got a trustworthy denominator. For the first time, the “say” side is disciplined through formal commitment, which turns a reported story into a measurement.
The tier architecture got exercised in production. Within weeks of publishing the second version, the first compressed-profile summit was proposed for the August cycle, scheduled around a major industry conference, using the tiered design exactly as specified: prioritization collapsed to one pre-work-driven day, the go-to-customer bridge intact, decision log and readout preserved. The design’s core claim, that the gate never skips because a leaner profile always exists, moved from theory to a scheduled event. [Proposed and pending approval; update after it runs.]
The second version shipped in June 2026 with AI built in. AI drafts and assesses; humans make every decision. That layer has its own case study.
Reflections
What I’d do again: design the exception path before anyone asks for one. The tiered profiles were designed before they were needed, which is why the first real compression request routed through the model instead of around it.
What I’d do differently: instrument the baseline earlier. The most damning fact in this story, security reviews skipped almost universally, was invisible for years precisely because nothing measured it. I now treat “establish the trustworthy baseline” as gate one of any operating model work.
Where it’s going: the model’s criteria are now encoded as data rather than prose, which is what makes the AI layer safe to build on top. That architecture is the next case study.