A team's iteration burn-down has matched the ideal line exactly for six windows. What would you suspect?

## What a chart on a wall is for An **information radiator** is a display put where the team already stands, so the state of the work answers itself without anyone having to ask. The name is deliberate: information leaks out of it passively, at a glance, whether or not you went looking. The opposite arrangement — a number you must open a system, find a page and run a query to see — radiates nothing, and by the time somebody retrieves it, the moment to act on it has passed. What makes a display a radiator rather than a report: - **Unmissable**, in the physical or virtual space the team actually occupies. - **Current**, updated as work moves rather than in preparation for a meeting. - **Cheap**, so it can be redrawn, argued with, or thrown away this week. - **Low-precision**, because its job is to raise a question, not to settle one. - **Addressed to the team**, which is exactly why it can afford to be ugly. The purpose is a conversation held **early**. A burn-down that goes flat on day two is doing its job when somebody standing next to it asks "what got added?" on day two, rather than at the end of the window when nothing can be done about it. ## How a chart becomes theatre Theatre begins when the audience changes. The moment the primary reader is someone outside the team who forms a judgement from the line, the chart stops describing the work and starts defending the team. Everything else follows from that: 1. **The measure becomes a target.** A number used to judge people stops being a good description of what it once measured, because the pressure acts on the number and the cheapest way to move a number is rarely to change the underlying reality. 2. **Updating becomes a ritual.** The chart is brought up to date the morning of the meeting, so it records what was reported rather than what happened. 3. **Definitions soften at the edges.** An item is marked complete when its author finishes coding, with integration counted in some later window. Work discovered mid-window goes on a side list instead of into the total. Both make the line prettier and the chart false. 4. **The chart is discussed as an object.** People begin talking about whether it *looks* good, which is the moment it has stopped being an instrument. | Radiator | Theatre | | --- | --- | | Read by the team, all week | Read by an audience, at a checkpoint | | Updated as work moves | Updated before the meeting | | Lumpy and sometimes ugly | Smooth and reassuring | | Prompts a question the same day | Prompts an explanation afterwards | | Cheap to redraw or discard | Defended, because it is reported | ## Why perfection is the tell Work does not arrive evenly. Items finish in lumps — nothing for three days, then four at once — because they are of different sizes and depend on one another. An honest burn-down is a staircase with uneven treads. A line that lies on the ideal reference for six consecutive windows describes a process nobody has ever observed, which leaves a short list of explanations, and the dull ones are far more likely than the flattering one: - items are being closed the moment coding finishes, before anything is integrated - discovered work is being kept off the window's total - the update is being reconciled to the reference rather than to the work - the window's contents are chosen so cautiously that nothing uncertain is ever inside An eleven-person team on a beekeeping-records product matched the reference line for six windows running and was praised for it. Asked how, an engineer explained that items were marked complete when their author stopped coding, and that anything found mid-window went onto a separate list. The chart was entirely accurate about what the team had recorded, and silent about what the team had delivered. Nobody was dishonest; the chart had simply acquired an audience, and the definitions drifted toward whatever kept the audience calm. ## Getting the instrument back - **Change the audience back.** Report outcomes upward — what shipped, what users did with it — and leave the chart to the team that has to act on it. - **Add discovered work in the window it was discovered.** This single rule restores most of the honesty, because it is the rule that gets bent first. - **Ask for the ugly chart.** A display that has never made anyone uncomfortable is not being read. - **Consider retiring it.** If a running increment can be shown instead, it is a better radiator than any chart drawn about it. ## What an interviewer is listening for Treating the chart as a social artefact rather than a data structure. Anyone can read a line; the differentiator is noticing who the line is *for*, predicting how it will be gamed under pressure, and having a concrete first move — usually changing the audience or the update rule, rather than changing the chart.