STR.17:5 - Archetypal Grounding
STR.17:5.1 - Document preparation: four different readings of apparent progress
This is a constructed teaching case. The organizations, records, measurements, resource allowances and decisions below are supplied example conditions, not reported telemetry, a product comparison or evidence of effectiveness in practice.
A document-services firm performs human checking of invoices. It asks whether technological change could support a useful preparation service within six months while retaining human responsibility and recoverable evidence for each check. Current service is viable. A platform renewal is due in three weeks; the firm does not need to commit to a new service at that renewal.
The first map distinguishes image-to-field extraction, interpretation of checks, access to customer records and reviewer support. It groups connectors and supporting storage together as “workflow tools”. The firm compares two successive eight-week windows, using dated technical records from the same selected suppliers and independent test sources. Reporting dates are retained separately. The supplied controlled comparison uses the same fixed set of 100 labeled documents, the same resource allowance and recorded component versions; each component comparison changes the named component while holding the other inputs and versions fixed. This is evidence about that set, not a sample estimate for all customers.
| Place in the initial map | Supplied observations | Supported difference |
|---|---|---|
| Interpretation component | Public mentions increase. An apparently new benchmark uses easier documents. On the unchanged 100-document task, the old and new component have the same outcomes; the test does not measure every possible capability. | Attention increased. No change was detected on the relevant fixed task. The new benchmark cannot establish a gain on that task. |
| Extraction tool | With little new publicity, an updated extractor reduces documents with a field mismatch from 8 to 3 at the same direct processing allowance. The records retain outputs and versions; two replays reproduce them. Of the 3 remaining mismatches, 2 occur on documents the old version handled correctly. | A repeatable local improvement with regressions: 7 old failures are corrected and 2 new failures appear. Neither customer-wide reliability nor total human review effort has been established. |
| Customer-record connector | A supplier announces a new interface. Neither an applicable test nor access to the supplier’s supporting evidence is available to the firm. | Interface capability and integration burden remain unknown. The announcement supplies a lead, not an observed improvement or a measured failure. |
| Reviewer support | Examination of records shows that recreating some checks requires the versions of both extracted fields and accessed evidence. That dependence was hidden within “workflow tools”, and the same difficulty appears in records from the earlier window. | The map omitted an enabling function. Finding it now does not date its emergence to the later window. |
The extraction arithmetic matters. A change from 8 mismatching documents to 3 is a net reduction of 5 on the supplied set, but it does not mean the new version simply fixes 5 cases and preserves all others. The paired records expose the 2 regressions. A direction that removes human review would therefore require claims this evidence does not support. A preparation service with review remains a possible contribution, subject to its whole effort and traceability conditions.
The firm compares three explanations. One says the interpretation component has advanced enough to justify replacing the platform; its present support is attention and an incomparable benchmark, so it cannot carry that decision. A second says quieter extraction improvement may reduce preparation work; its local technical premise is supported, while total reviewer effort remains open. A third says the practical constraint has shifted to evidence reconstruction. Earlier records show that this constraint was already present: the inquiry has discovered a neglected condition, not demonstrated that technology moved into a newly created upper layer.
The revised map separates evidence capture, version association and later retrieval from record access. Existing observations keep their dates and are reclassified where their content supports it. The announced connector remains an unresolved possible supplier; its announcement date is not used as the birth date of those functions. A later qualified test of a new connector could establish a separate change, but there is none in the supplied record.
The returns now differ. STR.12 receives no supported technical reason to replace the interpretation platform at renewal; the current performance premise can continue within its existing limits. STR.5 receives a conditional direction: help current recipients prepare documents using the improved extraction path while keeping review and recoverable evidence. It also receives the two observed regressions and the missing whole-effort result.
The next worthwhile question is narrower than a live customer trial: does the improved extractor, used with the firm’s existing evidence-capture and review procedures, reduce total preparation effort on the permitted fixed corpus? The service owner can authorize work on that corpus but has not authorized customer-data access or a service change. The inquiry design budgets half a staff-day for setup, one for execution, half for reviewer interpretation and half for documenting and closing the comparison: 2.5 staff-days against a three-day inquiry allowance. Those days displace planned report preparation; the owner accepts that sacrifice for this bounded question. A favorable result would support developing the conditional direction; an unfavorable result would leave current service as the adequate continuation; unresolved version evidence would return the traceability gap.
A larger public-demo comparison is not commissioned: under either plausible result, the firm would still lack the whole-effort and evidence-capture answer governing this continuation. The inaccessible connector evidence is retained as unknown, with no invented value and no standing monitoring obligation. The owner chooses one return after the bounded inquiry and before renewal. None of these actions authorizes the proposed service or extrapolates the fixed-corpus result to customers.
STR.17:5.2 - Machining: a real local gain need not support a new offering
This second constructed case changes the technology, pace and receiving use. A machine shop is considering an offering for small batches of a harder alloy within eighteen months. Its adequate present arrangement outsources the difficult cutting and performs finishing locally. A supplier agreement will be reviewed in six weeks.
The map starts with required dimensions and delivery time, then distinguishes cutting inserts, machine control, cooling, dimensional inspection, material qualification and reliable material provision. An insert is an enabling tool for the shop’s service; its own manufacture would require a different map. Two successive annual observation windows fit the available tooling revisions and supplier qualification records better than the document case’s eight-week windows. The review in six weeks remains separate from both that history and the eighteen-month impact horizon.
The supplied records compare inserts on the same machine, feed setting and alloy from supplier A. The new insert makes 60 conforming pieces before the specified wear threshold, compared with 40 for the old one. That supports a local return on tool use under those conditions. Two recorded lots from supplier B do not reproduce the improvement, and inspection remains the limiting step for promised delivery time. No combined energy, labor or monetary saving has been established.
The technical advance is not dismissed because it fails the current business proposal. Its supported use is narrowed to the demonstrated material conditions. The contrary lots and unchanged inspection constraint defeat the inference that the shop can now make the broader offering at the required delivery time. The map also separates material qualification from generic purchasing: supplier identity alone did not explain why the effect differed or establish the conditions needed to reproduce it.
The competing continuations are an offering limited to qualified material, retaining the outsourced arrangement, or investigating whether inspection and material provision can jointly support the desired service. STR.5 receives the conditional narrower possibility. STR.12 receives no supported reason to replace current outsourcing at the imminent agreement review. Before proposing a production trial, the shop asks whether existing supplier qualification and inspection-capacity records can settle the conjunction. If no suitable material provision and inspection capacity can be secured before the choice, it stops that present inquiry and retains the adequate arrangement. A future material or inspection change can reopen the possibility.
This transfer changes the map, observation pace and meaning of resource return. It preserves the same operation: qualify the local effect, carry its material dependencies and contrary evidence, and return only the strategic implication that the whole receiving use can support.