Sensitivity Table¶
Record artifact — instantiates Sensitivity Analysis Protocol
Records one row per assumption — its range, outcome response, materiality verdict, and critical flag — so the whole analysis can be audited line by line.
A sensitivity table is the auditable record of the protocol: a structured grid with one row per assumption and columns for its range, the resulting outcome response, a materiality judgment, and a critical / not-critical flag — so the entire analysis can be reviewed, reproduced, and challenged one line at a time. Where a tornado chart compresses the same information into a picture for a glance, the table keeps it in full, sortable, footnoted detail for a reviewer. Its defining virtue is traceability, not legibility: it is the artifact an auditor opens when they need to know not just which inputs mattered but exactly what range each was tested over and who judged it material.
Example¶
An insurer's actuary has estimated loss reserves for a liability book and must defend the number to an internal audit committee and, eventually, a regulator. A headline reserve figure and a confident memo will not survive scrutiny, so the analysis is laid down as a sensitivity table. Each row is an assumption: the loss-development tail factor, claim-frequency trend, severity inflation, discount rate, and reinsurance recovery rate. For each, the table records the range tested (tail factor 1.05–1.20, severity inflation 3%–8%, and so on), the resulting change in booked reserve, a materiality verdict against a stated threshold of $5M, and a flag marking whether a plausible move in that assumption would breach the capital adequacy line. Two rows come back flagged critical — the tail factor and severity inflation — while the discount rate and reinsurance rate are recorded as immaterial. The table is what makes the reserve defensible: every assumption is named, bounded, and judged, and a reviewer can reproduce any row.
How it works¶
- One row per assumption. Enumerate the varied inputs as rows; nothing that moved the number is allowed to stay off the sheet.
- Fill the standard columns. For each, record the range tested, the resulting outcome response, the materiality verdict against the threshold, and a critical/not flag, plus a source note.
- Sort and filter. Order by materiality or by critical flag so the rows that matter surface to the top of the record.
- Hand it off as the deliverable. The completed table is the artifact reviewers audit and downstream mechanisms cite.
Tuning parameters¶
- Granularity — one row per raw input versus grouped assumption families; fine rows are precise but sprawling, grouped rows are readable but can hide which member drives the group.
- Column set — whether to add source, owner, last-reviewed, or confidence columns; richer columns aid audit but raise maintenance cost.
- Materiality expression — numeric (dollar swing) versus categorical (material / immaterial); numbers are precise, categories are faster to scan.
- Static vs. live — a frozen snapshot versus a sheet that recomputes as inputs change; live tables stay current but can drift from the version that was signed off.
When it helps, and when it misleads¶
Its strength is accountability: it forces every assumption to be named with an explicit range and a materiality verdict, and it makes the analysis reproducible and contestable rather than a black box. Its failure mode is that it records but does not validate — a neatly tabulated bad assumption looks exactly as authoritative as a sound one[n1] — and a fat table can enable parameter theater, dozens of trivial rows crowding out the one hard assumption nobody wanted to write down. The classic misuse is treating a fully-populated table as evidence that the conclusion is robust. The guarding discipline is to pair the table with an informal self-check on whether the material rows are actually defensible, so the record documents judgment rather than substituting for it.
How it implements the components¶
outcome_response— the core column: the recorded per-assumption result of moving that input across its range.materiality_threshold— a column that judges each response material or immaterial against a stated threshold.critical_assumption— the flag column that marks assumptions whose plausible movement would be decision-altering.
It records but does not reach the overall verdict — robustness_conclusion is drawn from the table by Scenario Variation or Probabilistic Sensitivity Simulation, not written inside it. And it does not turn the same rows into an at-a-glance picture; that visual parameter_prioritization_rule display is the Tornado Chart's.
Related¶
- Instantiates: Sensitivity Analysis Protocol — this is the protocol's auditable record.
- Consumes: One-way Sensitivity Analysis and Two-way or Multi-way Sensitivity Analysis — records the outcome responses they compute.
- Sibling mechanisms: One-way Sensitivity Analysis · Two-way or Multi-way Sensitivity Analysis · Scenario Variation · Threshold Analysis · Tornado Chart · Probabilistic Sensitivity Simulation · Assumption Stress-test Workshop
Editorial Notes¶
Form Classification¶
Form family: Representation, Specification & Plan
Rationale: Sensitivity Table operates by externalizes varied assumptions, tested ranges, outcome responses, and materiality judgments in a table. That concrete deployed or enacted form is Representation, Specification & Plan under the frozen taxonomy.
Nearest alternative: Record, Log & Register — Although Record, Log & Register can support this mechanism, the frozen evidence makes its operative form the act that externalizes varied assumptions, tested ranges, outcome responses, and materiality judgments in a table; the alternative is therefore secondary rather than defining.
Review outcome: Adjudicated after independent review; high confidence.
Origin Attribution¶
Primary origin: Operations Research
Origin pattern: Convergent development
Present-day reach: Universal
Rationale: A table of outcomes across input or assumption values is the canonical presentation of decision-model sensitivity analysis. NASA and NIST ground the operation in systematic input variation; statistics and mathematics supply estimation and structure.
Related originating lineages:
- Accounting & Auditing — Line-by-line traceability and critical flags turn analysis into review evidence.
- Data Science & Analytics — Data science, analytics, and operational monitoring supplies a parallel or contributing lineage for the mechanism's defining operation: records one row per assumption — its range, outcome response, materiality verdict, and critical flag — so the whole analysis can be audited line by line.
- Mathematics — Mathematical modeling, proof, and abstract-structure practice supplies a parallel or contributing lineage for the mechanism's defining operation: records one row per assumption — its range, outcome response, materiality verdict, and critical flag — so the whole analysis can be audited line by line.
- Organizational & Management Science — Decision registers use structured rows to assign attention and escalation.
- Statistics & Experimental Design — Ranges and outcome responses encode the underlying robustness experiment.
Review resolution: The blind reviewers disagree on primary lineage (operations_research versus statistics_experimental_design). Authoritative or primary research supports operations_research as the best historical origin: A table of outcomes across input or assumption values is the canonical presentation of decision-model sensitivity analysis. NASA and NIST ground the operation in systematic input variation; statistics and mathematics supply estimation and structure. The cited NIST, Guide for the Use of the International System of Units: Model Sensitivity; NASA, Sensitivity Analysis Overview directly supports the mechanism's defining operation. All independently supported contributing domains are retained without an arbitrary cap. origin_mode=convergent records the lineage relationship, while domain_reach=universal records later applicability separately from provenance.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Researched adjudication after independent review; high confidence.
Sources consulted:
- NIST, Guide for the Use of the International System of Units: Model Sensitivity
- NASA, Sensitivity Analysis Overview
Notes¶
[n1] "Garbage in, garbage out" — the principle that a faithful computation on unsound inputs yields an unsound result no matter how rigorous the processing. A sensitivity table is especially exposed to it because tabulating an assumption in tidy columns lends it an authority its underlying evidence may not warrant. ↩