STR.12:5 - Archetypal Grounding
The following cases are constructed teaching cases, not reports of performed trials.
STR.12:5.1 - SensorCo stops the affected work, not the whole company
SensorCo’s Board has authorized four engineer-days of preparation while leaving the twelve-month choice between a viable device-only direction and service alternatives open. The preparation includes complete service costing and a feasible trial design. It authorizes neither customer-data use nor the eighteen-day larger preparation-and-trial configuration.
Suppose the completed costing adequately includes support, travel, incidents and the displaced device work, and makes the service alternative unattractive relative to device-only activity under the relevant conditions. The team recommends no trial and continued device-only activity for the present decision. The Board can close the larger inquiry on that basis. Four days already spent do not justify buying another experiment. A missing incident-cost component would instead leave this comparison unresolved; it would not be a negative cost result.
In a different continuation, suppose the required permissions and later trial allocation have actually been established. The agreed rule requires the service owner to suspend the dependent trial if the qualified forecast before the next month’s allocation exceeds eighty engineer-days while preserving the eight-day protected service reserve.
The capacity owner supplies a qualified forecast before allocation that exceeds eighty engineer-days with the eight-day reserve preserved. The service owner applies the agreed suspension rule to the affected trial; the Board reconsiders its scope, timing or rival. The forecast concerns next month’s planned shared workload. The service owner’s authority covers suspension of this trial, not redirection of the whole company. Independent device commitments and protected service continue under their existing authority.
If the same capacity issue is discovered during the original four-day preparation, there is no authorized trial to suspend. The result is an infeasible proposed allocation and a bounded return to the Board. Recovering the actual commitment changes the correct response.
STR.12:5.2 - A changed filter is not a changed observation
A working direction admits proposals that reduce rework this quarter, fit the available time, can be checked within two weeks and avoid irreversible migration. A proposal fits three conditions but fails the fourth.
Later, a participant suggests allowing irreversible migration where specified transition protection can be provided. That is a proposal to revise the filter, not evidence that the migration was always eligible. The responsible body compares the reason for the change and its consequences before adopting or rejecting it.
If the revised rule is adopted, the migration can be reconsidered for eligibility under that rule. Its actual rework contribution, time, checkability, protection and resource allocation still require their own support. If the proposal is rejected, the existing filter continues. Earlier observations do not need rewriting in either case.
STR.12:5.3 - A signal supports continuation
A professional has reserved two hours for an exploratory conversation. A general report predicts declining demand in another region, but the present conversation concerns an existing local client’s known question and makes no demand-dependent investment.
A qualified inspection finds no dependence on that forecast. The person can continue the conversation within its original scope. This is a bounded no-impact conclusion, not a claim that all future client-development choices are unaffected.
STR.12:5.4 - More paid customers need not mean a better conversion result
A service is considering further spending on a revised onboarding offer. Its dashboard shows paid-customer counts rising from twenty to forty-five. The analyst recovers the registration groups and actual first orders instead of treating that increase as the effect of onboarding. Assume that the permitted records establish each customer’s identity, entry and first-order date, with complete follow-up where stated.
| Registration group | Registered customers | Customers with a first paid order | Observation basis |
|---|---|---|---|
| January | 100 | 20 | Each customer’s first 28 days |
| February | 300 | 45 | Each customer’s first 28 days |
| March | 200 | 12 so far | Each customer’s first 7 days only |
The observed twenty-eight-day proportions are 20% and 15%. More customers paid in February, but a smaller proportion of that registration group paid within the defined period. March’s 6% after seven days is not its twenty-eight-day result. Without the older groups’ seven-day readings, it cannot supply that shorter-period comparison either.
Suppose February also used a different acquisition channel. These records do not isolate the onboarding change’s effect, prove that it is harmful, or establish the result for future customers. The analyst withdraws the claimed effect and returns the supported counts, proportions and uncertainty about that stronger inference. If an effect estimate is necessary for the next commitment, an adequate existing causal result or a worthwhile controlled comparison can supply it. The current descriptive correction can finish without another trial; the funding decision still needs its own comparison and authority.