Find your next learning step
Choosing one discriminating polymer experiment is only the beginning. Its result changes which protocol should run next, how many specimens remain and whether the mechanism map is ready to support a release or research decision. A sequential campaign treats batch design and stopping as linked decisions. A map is locked when further feasible evidence is unlikely to change the declared action within the agreed tolerance—not when every boundary looks perfectly certain.
Open the chapter contents · Polymer Science root · Gold Standard of Education · How X Works
Chapter contents
Choose the route closest to the decision, or read from Chapter 1. The chapters remain open; these grouped routes are a navigable index.
Route 1 · Define the decision
Route 2 · Work the ledgers
Route 3 · Transfer the method
Route 4 · Check limits and evidence
Route 5 · Ask and continue
CHAPTER 1 OF 12 · DEFINE THE DECISION
1. Define the map decision and lock tolerance
Name the boundary that matters, the action supported by each plausible map and the maximum unresolved probability or regret allowed at lock. A campaign without a decision tolerance can always justify one more test.
Separate an operational lock from a scientific claim of truth. The locked map may be adequate for the present release rule while remaining open to revision when materials, protocols or consequences change.
CHAPTER 2 OF 12 · DEFINE THE DECISION
2. Design batches that can adapt
A batch should be large enough to estimate protocol effects and small enough to learn before consuming the remaining specimens. Predefine candidate next designs so adaptation follows evidence rather than improvisation.
Keep calibration specimens and a final validation set outside the adaptive loop when possible. Reusing the same evidence for selection, stopping and confirmation can make the map look more resolved than it is.
CHAPTER 3 OF 12 · WORK THE LEDGERS
3. A sequential campaign ledger
A recycled polymer blend begins with two disputed mechanism boundaries and 30 preserved specimens. Each batch is chosen after the previous batch is interpreted.
| Stage | Experiment batch | Specimens used | Expected boundary resolution | Planned response |
|---|---|---|---|---|
| 0 | Baseline spectroscopy | 6 | 0.12 | Screen gross mismatch |
| 1 | Thermal cycling | 8 | 0.31 | Split ageing pathways if divergence is high |
| 2 | Interface microscopy | 6 | 0.44 | Resolve the remaining boundary |
| Reserve | Mechanical recovery | 6 | 0.20 | Use only if release action remains unstable |
Expected resolution is recalculated after every stage. The schedule is a decision tree, not a promise to run every listed protocol.
CHAPTER 4 OF 12 · WORK THE LEDGERS
4. Worked stop-and-lock rule
After microscopy, the team evaluates whether another recovery batch can still change the release classification. The rule combines boundary support, decision regret and validation performance.
| Gate | Lock requirement | Observed value | Result |
|---|---|---|---|
| Primary boundary support | At least 0.90 | 0.94 | Pass |
| Worst-map release regret | At most 0.08 | 0.05 | Pass |
| Held-out predictive check | Within declared band | Within band | Pass |
| Value of reserve batch | At most 0.03 | 0.02 | Pass |
If one gate failed, the response would be specified in advance: run the reserve batch, retain multiple maps or choose the action safe under ambiguity.
CHAPTER 5 OF 12 · WORK THE LEDGERS
5. Value continuation, not information alone
A further experiment may reduce entropy while leaving the release decision unchanged. Compare immediate terminal value with the expected value of continuing, including delay, specimen loss and the chance that a new result redirects later batches.
Simple thresholds are useful but can be myopic. A sequential design should consider the option value created by preserving specimens and keeping future protocols available.
CHAPTER 6 OF 12 · TRANSFER THE METHOD
6. Lock with versioning and an unlock rule
Store the locked map as a versioned object with data, protocol weights, uncertainty envelope, release rule and scope. New lots should not silently overwrite the historical decision.
Declare unlock triggers such as a predictive-check failure, a new processing route, a protocol calibration change or a material shift beyond the validation domain.
CHAPTER 7 OF 12 · TRANSFER THE METHOD
7. Changed-context transfer: barrier-film degradation
Transfer the campaign to a barrier film where humidity exposure, thermal cycling and microscopy compete for limited coupons. The decision is whether two degradation pathways can share one shelf-life model.
The transfer check keeps humidity and temperature blocks intact, prices coupon destruction and locks only within the validated climate range. A map that is stable at indoor humidity is not automatically locked for tropical outdoor exposure.
CHAPTER 8 OF 12 · TRANSFER THE METHOD
8. Connect selection, envelopes and influence
Begin with the discriminating-experiment guide. The consensus-envelope guide supplies boundary uncertainty after each batch.
Use the protocol-influence guide before locking a map whose apparent resolution depends on one high-leverage protocol.
CHAPTER 9 OF 12 · CHECK LIMITS AND EVIDENCE
9. Limits and common misconceptions
Stopping because a preferred map first crosses a threshold can exaggerate support unless the sequential rule is built into the analysis. Repeated looks are part of the design, not an invisible convenience.
Locking does not erase minority mechanisms or make future evidence irrelevant. It freezes a scoped operational decision together with its uncertainty and revision conditions.
CHAPTER 10 OF 12 · CHECK LIMITS AND EVIDENCE
10. Practical evidence checklist
Archive candidate maps, actions, tolerances, specimen inventory, adaptive policy, batch outcomes, protocol calibration, costs, delays, random seeds, stopping gates and every decision to continue or preserve material.
At lock, save the final map version, uncertainty envelope, held-out checks, residual alternatives, scope and unlock triggers. Record who can reopen the campaign and which evidence is needed.
CHAPTER 11 OF 12 · ASK AND CONTINUE
11. Frequently asked questions
Must every planned batch be run?
No. A sequential plan includes contingent batches. Stop when the declared decision is stable and continuation value no longer justifies cost and specimen use.
Can a map be locked with one unresolved boundary?
Yes, if every plausible position of that boundary supports the same safe action within tolerance. The unresolved region must remain visible.
What happens when a new protocol arrives?
Check whether it targets a locked boundary, changes calibration or expands the domain. Apply the predefined unlock rule instead of silently blending it into the old map.
CHAPTER 12 OF 12 · ASK AND CONTINUE
12. Sources and next reading
Cheng and Huan develop optimal stopping for sequential Bayesian experimental design. Filstroff and colleagues study targeted active learning for downstream decisions. Go and Isaac propose robust expected information gain under prior ambiguity. The polymer campaign ledgers are original teaching examples.
Continue through the Polymer Science root, the Gold Standard of Education and How X Works.
Top · Contents · Previous · How X Works