Skip to content

Done-at-Eighty-Percent Demo

Demonstration review — instantiates Sufficiency-Bounded Work Containment

Shows the work to its customer while it's still visibly rough — around 80% — so their "this already does what I need" becomes the trusted signal to stop, before the expensive last polish.

A Done-at-Eighty-Percent Demo shows the work to its actual customer while it is still visibly unfinished — roughly 80% — precisely so the customer's "this already does what I need" can be surfaced early and made the trusted signal to stop. Its defining move is treating the remaining polish as a hypothesis to be tested, not a debt to be paid: the last increment is assumed to be gold-plating until the customer actively asks for it, which reverses the usual default from "keep refining unless told to stop" to "stop unless told to keep going." It does not define what "good enough" is or write the item's acceptance terms; it is the event that pulls the trigger — the moment the decision to stop short becomes legitimate because the person the work is for has said it is enough.

Example

An analyst is building a sales dashboard an executive requested. Rather than perfect every chart, colour, and filter first, she demos it at about 80% — the core metrics are live, a few edges are rough, and the fancy drill-downs aren't built yet. In the demo the executive sees that the top-line trend and the regional split already answer the question he actually had; the drill-downs and polish he'd vaguely gestured at turn out to be things he would never open. The demo pulls the trigger: the dashboard ships as-is. The last 20% — several days of polish and speculative features — is simply never built, because showing the rough version early revealed it was gold-plating rather than value.

How it works

  • Demo early and deliberately rough — the artifact is shown while clearly incomplete, so the conversation is about sufficiency, not final sign-off on a finished thing.
  • The customer's reaction is the trigger — "this already does what I need" is the signal to stop; the mechanism's entire job is to surface that signal early enough to act on it.
  • The unbuilt remainder is a hypothesis — the last increment counts as gold-plating until the customer explicitly asks for it, reversing the default from polish-by-default to stop-by-default.
  • Stopping is the success case — an early demo that ends the work is the mechanism working, not a corner being cut.

Tuning parameters

  • Demo point — how early "early" is (60%, 80%, 95%). Earlier catches more waste but risks showing something too raw to judge fairly.
  • Audience — the real user/customer versus a proxy. Only the person the work is for can credibly say "enough"; a proxy's approval doesn't pull the trigger.
  • Roughness shown — how unpolished the demo is allowed to look. Visible roughness invites honest "that's fine" reactions; a too-slick demo hides that the work is already sufficient and quietly invites more requests.
  • Default on silence — whether ambiguity means stop or continue. Defaulting to stop is what makes the demo contain work rather than license more of it.

When it helps, and when it misleads

Its strength is that it turns "when do we stop polishing?" from a guess into an observed reaction, attacking gold-plating — delivering beyond what the customer asked for — at the exact moment before the costliest increment is spent.[1] Its failure modes: demo too late and the polish is already sunk; demo to the wrong audience and the "enough" carries no authority; and it is easily run in reverse — staged as a box-ticking show-and-tell after the team has already decided to keep building, so it rubber-stamps rather than tests. It can also stop genuinely-needed work if the customer under-asks in the moment. The discipline that keeps it honest is to demo to the real customer, early enough that stopping is still cheap, and to treat the demo as a real decision point rather than a ceremony.

How it implements the components

  • stop_short_trigger — the customer's acceptance of the rough demo is the trusted signal to stop before the container is spent; the demo exists to surface that trigger early and make acting on it legitimate.

It pulls the trigger but does not define what "good enough" is (independent_sufficiency_criterion — Definition of Done) or the item's specific acceptance terms and record (elastic_activity_boundary / completion_evidence_packet — Acceptance Criteria Checklist). It consumes those bars and fires the stop against them.

Notes

The "80%" is a stance, not a measurement — the point is to demo before the expensive final polish, not to hit a precise percentage. Treating it as a literal metric ("we are at 79%, don't demo yet") reintroduces exactly the deferral the mechanism exists to cut. Demo when the work is far enough along to judge and early enough that stopping is still cheap.

References

[1] Gold-plating — delivering more than was asked for — is a recognised project-management anti-pattern: effort spent past the point of requested value. The early demo is a cheap test for whether the next increment is value or gold-plating, decided by the one person entitled to say. ↩