How do you tell an exploratory test charter is too broad, and how do you split it?

## Why sizing is the hard part of chartering Writing a plausible-sounding charter is easy; writing one that a single uninterrupted stretch of testing can honestly finish is the skill. An oversized charter fails quietly: the tester works the whole time, produces notes, finds something — and cannot say what fraction of the target was touched. Over a few weeks that team has activity records but no coverage picture, and the areas nobody happened to wander into stay untested indefinitely. ## Signals that a charter is too broad - **You cannot state coverage at the end.** The clearest test. If the honest answer to "what did you cover?" is a shrug or a list of whatever happened to be interesting, the target was too big. - **The target names a subsystem, not a behaviour.** "Explore the appointment scheduler" is a heading; "explore waitlist promotion when a booked appointment is cancelled inside the notice window" is a target. - **An "and" joins unrelated areas.** "Explore booking and clinician rostering" is two charters that share a sentence. - **The resources slot is empty or generic.** "with the usual data" usually means the tester has not yet decided what varies, which means the scope is not yet decided either. - **The information goal is plural and vague.** "to discover issues" covers everything, so it bounds nothing. - **It keeps coming back.** A charter run in three sessions and still "partly done" was never a charter; it is a backlog heading pretending to be one. A concrete case: a four-person team chartered "Explore the scheduler for booking problems." The session ran 78 minutes and touched 3 of the 11 screens involved in booking. The notes were good, the finding was real, and the team still had no idea what the other 8 screens did — because nothing in the charter had ever committed to them. ## The axes you split along Splitting is not slicing the sentence in half; it is choosing a dimension that makes each piece independently explorable. - **By sub-area or workflow step** — booking, rescheduling, cancellation, waitlist promotion. - **By data condition** — clinics whose hours cross midnight; a 17-minute slot length that does not divide the session hour evenly; a patient with two active referrals. - **By user role** — what a receptionist, a clinician and a clinic administrator can each see and do against the same appointment. - **By quality attribute** — the same booking flow explored for accessibility, for behaviour under a slow network, or for what it discloses in error messages. - **By recent change** — the area a change landed in, explored against how it behaved before. Each split charter keeps the full three-slot shape. "Explore rescheduling with appointments already inside the 24-hour notice window, to discover waitlist promotions that skip an earlier claimant" is a session; the parent sentence it came from stays in the backlog as a heading, not as a runnable charter. ## The opposite failure A charter can be too *narrow*. If it is finished in twelve minutes with nothing left to vary, it was a check: the steps and the expected result were effectively known in advance, and the work would be cheaper and more reliable as an automated case. Two symptoms: the tester spends most of the timebox inventing extra scope that the charter does not mention, or the session notes read as a single pass/fail line. Merge such charters upward, or promote the fixed part into an automated check and charter the surrounding uncertainty instead. ## Sizing is a forecast, and forecasts get corrected You do not know the right size until you have run a few sessions on similar material, so treat the first estimate as provisional. When a session overruns its target, the useful response is to record what was actually covered, note the untouched remainder, and write the remainder up as its own charter — not to extend the session until the charter is "done". That habit is what converts a sizing mistake into a piece of planning information instead of an accounting hole. The reverse correction matters too: two charters that both finish early and keep bumping into each other's area should be merged, because the boundary between them was not real. ## What good sizing buys you Right-sized charters make exploratory work **countable** without making it scripted. Each finished charter is a statement of the form "this area, with these conditions, was explored for this risk", which composes into a coverage picture a team can reason about, argue with, and spot holes in. That is the entire practical case for chartering: not the sentence itself, but the honest ledger of targets that the sentence makes possible.