A test cycle is already half executed when a further area of the case library has to be covered. What are your options for getting those cases into the run, and what does each cost?
answer
- extend it, or start a second one
- the denominator grows, progress falls
- a scope that moved stops being comparable
- a live rule adds without asking
- record when the scope changed
basics
~20 sEither extend the running cycle, which grows its total, drops its completion figure and breaks comparability with earlier runs, or open a second cycle, which keeps both scopes clean at the cost of reconciling two reports.
solid answer
~50 sThree options exist and they trade the same thing. **Extend in place**: the cases join the running cycle, the denominator grows, and every progress figure moves backwards through no fault of the testers — cheap to do, but the cycle no longer means what it meant yesterday. **Open a second cycle**: the running one keeps a stable, comparable scope and the new work gets its own, at the cost of reconciling two reports at the end. **Let a live selection do it for you**: if the cycle re-resolves a saved rule, the new cases arrive without anyone deciding — the worst option, because nobody chose and nobody was told. My default is a second cycle once execution has meaningfully started, and extension only in the first hours when little has been recorded. Whichever you pick, record when and why the scope changed.
go deeper
Know that a cycle's contents can change after it starts, and that adding cases makes the completed share fall because the total grew, not because work was undone.
Explain the arithmetic of a mid-run addition and why a widened scope stops being comparable with the cycle it is meant to be measured against.
Choose between extending and opening a second cycle on the basis of how far execution has gone and what the cycle will be quoted for, and say how you record the change.
Own the convention for when a scope may move at all, so that cycle-to-cycle numbers across teams mean the same thing and late work is visible rather than absorbed.
## The situation Execution is under way. Results exist. Then something arrives that has to be covered — a late change lands, an area everyone assumed was out of scope turns out to be in it, or a batch of cases finishes authoring after the cycle was built. The membership of a running cycle is now wrong, and every way of fixing it has a cost. ## The three options | Option | What it costs | When it fits | |---|---|---| | Extend the running cycle | totals move, comparability with earlier cycles is lost | very early, little recorded | | Open a second cycle | two reports to reconcile at the end | execution meaningfully under way | | Let a live rule sweep them in | nobody decided, nobody was told | never deliberately | ### Extending in place The cases are added to the existing cycle. This is one action and it keeps everything in one place, which is genuinely valuable — one list, one report, one thing to look at. The cost is that the cycle's arithmetic changes underneath everyone. Adding items grows the total while the number already completed stays where it was, so any completion figure drops. Nobody did anything wrong and the chart says the team went backwards, which is a conversation you will have every time. More seriously, the cycle no longer has the same scope it had when it started, so comparing its outcome to the previous release's cycle silently compares two different populations. ### Opening a second cycle The new cases go into their own cycle beside the first. The first keeps a stable scope, so its numbers stay meaningful and comparable with history; the second describes exactly the late work, and is honest about being late. The cost is reconciliation. Somebody must add two reports together to answer "are we done", and a reader who sees only one of them draws the wrong conclusion. That cost is real but it is *paid consciously at the end*, whereas the extension cost is paid silently throughout. ### Letting a live rule do it If the cycle stores a rule that is re-applied rather than a fixed set, newly authored matching cases arrive on their own. This looks like a feature and behaves like an incident: the scope changed with no decision, no announcement, and no record of when. Treat an unexplained rise in a cycle's total as this until proven otherwise. ## What to do in practice 1. **Ask how far in you are.** In the first hours, with little recorded, extending is nearly free and clearly the simpler choice. Past the point where the numbers are being reported, the extension distorts something people are already reading. 2. **Ask whether the two sets belong together.** Late cases for the same feature belong with the original scope; a whole area nobody planned for is its own piece of work and reads better as its own cycle. 3. **Ask what the cycle will be used for.** If it will be quoted in a sign-off, keep its scope frozen and put the addition elsewhere. 4. **Record the change either way.** Note when the scope moved and why, in whatever field or comment the cycle carries. A total that changed without explanation is the thing that makes a report unusable months later. ## Keeping the late arrivals visible Whichever route you take, the added cases should stay distinguishable rather than dissolving into the set: - Tag or name them so "what did we add mid-run" is answerable after the fact. - Snapshot the case count and identifiers at the moment of the change, so the before and after are both reconstructible. - Say the total moved when you next report on the cycle, rather than letting a reader discover a percentage that fell. - Watch for the same addition arriving twice — once by hand and once through a rule still matching — which shows up as duplicate items over the same case. ## The judgment being tested An interviewer asking this is checking whether you understand that a cycle is a **measurement instrument as well as a worklist**. The naive answer optimises the worklist: put everything in one place, it is easier. The senior answer notices that changing an instrument's scale halfway through invalidates the readings taken before, and weighs that against the convenience. Either choice can be right; making it without noticing the cost is what is wrong.
- A running cycle's total went up overnight and nobody added anything by hand. What is your first hypothesis?That the cycle stores a rule that is re-applied rather than a fixed case set, and cases authored or re-tagged since creation now match it. Check the library's recent edits against the rule's conditions before suspecting anything else — a moving total is almost always a membership event rather than a defect.
- How do you keep the mid-run addition visible in the record after the cycle closes?Mark the added items distinctly — a tag or a naming convention — and note the case count before and after the change with the reason. Without that, the closed cycle shows one flat scope, and nobody can later tell which items were planned and which arrived halfway through.
saying these in an interview costs you the question
- Adds cases mid-run without mentioning that the total moved
- Compares a widened cycle against last release's numbers
- Treats a rising total as a display problem
- Says a second cycle is always the right answer
- Leaves late additions indistinguishable from the original scope