- To
- Head of Operations, Monarch
- From
- AJ Dysick
- Date
- September 2026
- Re
- A quarterly planning cadence for a 150-person remote company, and where it breaks
At 150 people, the founders can no longer hold the plan in their heads, and the company is too small for a planning function. Whatever cadence exists has to be light enough for functional leaders to run it themselves and firm enough that a decision made in week 2 is still the decision in week 9. Below is the version I'd start with, then the six ways I've watched cadences like it fail and what fixes each one.
The cadence, in one quarter
-
Before the quarter, weeks –3 to –1
- Strategy note. Founders write two pages: what changed since last quarter, what we're betting on, what we're explicitly not doing. This is the only input that comes top-down.
- Objectives. Each functional leader drafts no more than three, each with an owner, a metric with a baseline, and a list of what they need from other teams.
- Dependency review. One meeting. Every cross-team ask is accepted with a named owner, negotiated down, or declined in writing. A dependency nobody accepted is a risk, not a plan.
- Commit. The plan is one page in the canonical doc. Not a deck.
-
During the quarter, weeks 1 to 12
- Weekly business review, 45 minutes. Metrics are updated before the meeting and read before the meeting; the meeting covers exceptions only. Every red item leaves with a decision, or a named decider and a date. If nothing needs a decision, it ends early.
- Monthly checkpoint. Each objective is green, yellow, or red, with one line on why. Yellow gets a plan. Red gets a decision.
- Re-plan window, week 7. The one sanctioned time to change commitments. Outside it, a change is a short written request that names what gets dropped to make room.
-
After the quarter, week 13
- Retro. What we committed, what shipped, and what we learned about our own forecasting. It feeds the next strategy note.
Four artifacts, two meetings. That's the whole system.
Where it breaks
| failure | fix |
|---|---|
| Objectives that are activities"Launch X" isn't an objective; "X moves metric Y from A to B" is. | A rule at the door: no metric and baseline, no entry into the plan. |
| Dependencies discovered in week 5The team that needed design capacity finds out design was never asked. | The dependency review before commit, and treating an unaccepted dependency as a risk with an owner rather than a hope. |
| Founders re-litigating in 1:1sThe plan changes on a Tuesday in a conversation nobody else was in. | The founder's authority is real; exercise it through the re-plan window or a written change request, so everyone runs the same plan. The mechanism protects the founder as much as the teams. |
| The weekly review becomes status theaterTwenty minutes of reading slides aloud and no decisions. | Metrics-first, exceptions-only, and a decision log. Track decision latency, the days from a red to a decision. If it's climbing, the meeting has stopped working. |
| Remote driftThe plan lives in a deck nobody opens after week 1. | One canonical doc the weekly review reads from directly. If it isn't in the doc, it isn't the plan. |
| Too many mechanismsEvery failure tempts a new meeting. | At 150 people, the cadence above is the ceiling. Add a mechanism only when a specific, repeated failure demands it, and retire any mechanism that stops producing decisions. |
How I'd know it's working
What this borrows. The weekly review and the exceptions-only rule come from Amazon's WBR, which I ran for a 350-person site. The change window comes from peak planning, where a late change to the plan costs real money and gets treated that way. The "one page, not a deck" rule comes from watching plans die in slide decks.
Where it comes from
Eight years, one through-line: build the operating system a business runs on, then run it.
- 2016 – 2020Vector MarketingTwo P&L-owning offices built from zero to profitable in four months. Designed the recruiting, training, and performance cadence for each.
- 2021 – 2025AmazonRan the mechanism stack for a 350-person site: WBRs, standard work, peak planning, OP1/OP2 inputs. Owned labor, capex, engagement, and supply budgets.
- 2025DP WorldNon-management analyst seat. $3.86M saved and 65% fewer inventory errors, moved through evidence alone.
- 2025 – nowPVH300-person e-commerce fulfillment shift. Cross-functional programs owned outside the reporting line: safety, training, labor migration.
The full resume has the numbers behind each of these.
What I built around Monarch
I run my finances on Monarch. I wanted to see further ahead than the app shows, so I built around it.
Compass
A personal finance command center on Next.js and Supabase, with Plaid for live account sync and Monarch exports feeding it. It turns scattered accounts into one picture (net worth, debt payoff) and adds the part I couldn't get anywhere else: a FIRE projection that takes what I actually spend and save, includes rental and business income, and answers when work becomes optional and what moves that date.
Integrations
Monarch is the ledger everything reconciles to. Compass sits in the middle. Sequence holds the rules that actually move money, and Google Sheets holds the amortization models. Claude runs across all of it.
Claude, through Chrome
Monarch has no public API, so Claude works it the way I would: through Chrome, reading the app directly and pulling exports, then handing the numbers to scheduled tasks that run through MCP connectors to Sequence and Google Sheets. Nothing gets re-keyed, and the weekly picture is current before I look at it.
- Read MonarchClaude opens the app in Chrome and pulls the export, the way I would.
- Scheduled task runsReconciles against Compass and the Sheets models.
- Write through MCPSequence rules and the amortization sheets get updated in place.
- Picture is readyThe weekly view is current before I open it. Nothing re-keyed.
What I wanted the product to do
Three things, and none of them are complaints about the core product. First, a forward view: Monarch tells me where I am; I wanted to know when work becomes optional, with rental and business income in the model, and what one change to spend or savings rate does to that date. That's the FIRE module in Compass. Second, moving money: Monarch sees the accounts but doesn't act on them, so the rules that split income and fund savings live in Sequence. Third, programmatic access: there's no public API, so everything I build runs on exports and on Claude working the web app through Chrome. If Monarch shipped any one of the three, I'd retire that part of my stack the same week.
Contact
- ajdysick@gmail.com
- linkedin.com/in/ajdysick
- Resume
- Dysick-2026_Resume_StrategicOps.pdf
- Site
- ajdysick.com