| name | blind-unknown-unknowns |
| description | Use when scoping a project, entering an unfamiliar domain, writing a risk register, estimating novel work, or when a plan lists only risks the planner already thought of. Applies the knowledge matrix (known/unknown ร known/unknown) from 1960s aerospace risk practice, popularized via Rumsfeld 2002. |
Unknown Unknowns
Overview
The knowledge matrix (1960s aerospace/defense risk practice โ "unk-unks", documented from 1968, common at NASA; made famous by Rumsfeld's February 2002 briefing, who credited NASA's William Graham; the Johari Window of Luft & Ingham, 1955, is an analogous, independently developed 2ร2):
| We know it | We don't know it |
|---|
| We know we know/don't | Known knowns (facts) | Known unknowns (open questions) |
| We don't know weโฆ | Unknown knowns (unexamined assumptions, tacit knowledge) | Unknown unknowns (unmodeled surprises) |
Risk registers only ever contain known unknowns โ by definition. The two bottom-row cells are where projects die: unknown knowns (assumptions so deep nobody wrote them down) and unknown unknowns (categories you haven't imagined). Neither yields to direct listing; both yield to indirect hunting.
When to Use
- Risk lists that look complete (they're complete over the top row only)
- Novel work with no precedent in your own history
- Post-planning, pre-commitment ("what category of thing haven't we considered?")
- After a surprise: classify which cell it came from to fix the process, not just the instance
The Procedure โ hunting each cell
- Known unknowns: list open questions; assign each a resolution action and date. Cheap; do it fully.
- Unknown knowns (assumption excavation): complete the sentence "this plan works as long as ___" ten times. The last three are the load-bearing tacit ones. Also: explain the plan to an outsider โ everything you have to explain that "goes without saying" is an unknown known surfacing.
- Unknown unknowns (indirect only):
- Reference-class scan: what surprised OTHER people doing similar work? Their unknown unknowns can become your known ones for free (see blind-outside-view).
- Category sweep: walk a fixed taxonomy โ people, dependencies, legal, security, data, timing, money, environment โ asking "what haven't I even thought about here?" A checklist can't contain your surprise, but it can point your attention at unlit areas.
- Fresh eyes: someone outside the plan reviews it cold โ outsiders don't share your selection filter (see blind-johari-window).
- Buffer for the residual: whatever survives the hunt, budget explicitly for surprise (Flyvbjerg: uplifts, not optimism).
Common Mistakes
- Claiming a complete risk register. The correct closing line is "plus a residual we cannot enumerate, buffered by X."
- Trying to brainstorm unknown unknowns directly โ five more minutes of the same mind produces the same categories. Change the mind (outsider) or the data (reference class), not the effort.
- Ignoring unknown knowns because unknown unknowns sound scarier. Post-mortems attribute far more failures to "we all assumedโฆ" than to true black swans.
