When a scaling company stalls, the reflex is always the same: sharpen the objectives. Tighter OKRs, better cascaded, reviewed more often. Every function gets its key results, every key result gets an owner, the dashboards light up green.
And the company stays stalled.
Hundreds of people working like exceptionally efficient bees — disciplined, fast, hitting their numbers — while the thing that blocks the company sits untouched. Not because anyone is lazy. Because the block belongs to no single bee. Structural tensions live between functions, and OKRs live inside them. A cascade assigns every objective to exactly one owner; the instrument's own geometry hides the disease.
What does such a tension actually look like? Always the same shape: two readings that must both be true for the plan to be real — and aren't. The plan needs 30 installations a week; the field runs 11, at 10 hours per install where the unit economics assume 6. The CTO, asked alone, rates the platform ready for 10x; the COO, asked alone, rates the same system ready for 2x. Each number, alone, looks unremarkable. The diagnosis is in the distance between them — and the distance appears on nobody's dashboard, because its two legs sit in two different functions.
Our diagnostic reads a company through 65 such tensions, and sorts every one by the only classification that matters: where the break sits. External — gated by approval regimes or a price race you cannot win; these can only be watched or escaped, never worked. Design — built into the machine itself, like a project business priced as a product business. Current performance — the one layer where operational excellence genuinely helps, which is exactly why companies treat every problem as if it lived there. Under load — invisible today, firing only when volume arrives.
The layer determines the fix, and the sequence determines everything: external before design, design before performance, performance before load. A company clearing them out of order clears nothing — and an OKR program, pointed at the performance layer while a design fault is live, produces the cruelest outcome of all: every key result achieved, every tension worse.
That is why the answer to a stall is not another action plan. Nobody follows action plans, and it wouldn't matter if they did — pressing the buttons harder does not change the machine. The answer is a sequence of structural resolutions, each spanning the two executives whose functions it binds, each with defined evidence that the tension no longer measures. We call it a Gate Masterplan. The bees will keep flying either way; the only question is whether anyone has fixed the hive.
The full essay on Under Load goes further: the four tension types with worked numbers, and one composite stall traced through five tensions and four layers — including what the OKR response did to it.
Read "The Bees Are Fine" on Under Load
Published July 2026 · Robert Kellner