Research & Evidence

Search for a method

Search titles, questions, territories and MSC identifiers.

    ← All methods
    METHOD DOSSIERMSC-P-003Evidence foundationsVerified scientific dossier

    How should uncertainty in a marketing result be expressed?

    Uncertainty must match the result: a confidence interval for a parameter, prediction interval for a future observation, or scenario distribution for a simulation.

    Scientific editorial team: Marketing Science Center

    Direct answer

    Compare estimates with their precision and assumptions.

    Uncertainty must match the result: a confidence interval for a parameter, prediction interval for a future observation, or scenario distribution for a simulation.

    NIST DATAPLOT FIT manualStan posterior_interval documentation

    01

    Operational summary

    Scientific question and scope

    How should uncertainty in a marketing result be expressed?

    Supported

    Compare estimates with their precision and assumptions.

    Forbidden

    Interpret an interval as a guarantee for an individual case.

    02

    Three reading levels

    1. 01

      Decision-maker — Connect the result to a declared decision, useful threshold and error cost.

    2. 02

      Practitioner — Fix population, unit, horizon, available variables and analysis rule before calculation.

    3. 03

      Analyst — Reproduce the calculation, quantify uncertainty and document diagnostics, failures and sensitivities.

    03

    Concrete marketing situation

    Uncertainty must match the result: a confidence interval for a parameter, prediction interval for a future observation, or scenario distribution for a simulation.

    04

    Scientific question and scope

    How should uncertainty in a marketing result be expressed?

    One declared object: parameter confidence interval, future-observation prediction interval, posterior credible interval, or simulated outcome distribution

    Required data

    Result × inferential target × data-generating process

    Population, unit of analysis, origin date and horizon must be declared in the deliverable. Without them, the estimand silently changes.

    05

    Why a simple analysis can fail

    • Interpret an interval as a guarantee for an individual case.
    • Match interval to target
    • Declare repeated-sampling or posterior interpretation
    • Check coverage or calibration

    06

    Method intuition

    The method does not automatically turn an association into evidence. It links a declared question to an estimand, specification, compatible data and a bounded interpretation rule.

    One declared object: parameter confidence interval, future-observation prediction interval, posterior credible interval, or simulated outcome distribution

    07

    Required data

    Scientific symbol dictionary

    estimate
    Point estimate of the declared marketing parameter. Unit: percentage point · Type: number · Role: input
    SE
    Standard error of that parameter estimate under the declared sampling model. Unit: percentage point · Type: positive number · Role: input
    z
    Two-sided standard-normal critical value for the declared confidence level. Unit: dimensionless · Type: positive number · Role: parameter
    CI
    Confidence interval for the parameter; it is neither a future-observation prediction interval nor a posterior interval. Unit: percentage point · Type: ordered pair · Role: output

    Exact sealed engine inputs

    estimate_pp
    Value: 3.2 · Unit: percentage_point · Type: number · Data status: synthetic
    standard_error_pp
    Value: 1.1 · Unit: percentage_point · Type: number · Data status: synthetic
    z_critical
    Value: 1.95996398454 · Unit: dimensionless · Type: number · Data status: parameter

    08

    Formal model

    Formal model

    CI = [estimate − z × SE; estimate + z × SE]

    Evidence claims: MSC-P003-C01 · MSC-P003-C02 · MSC-P003-C03 · MSC-P003-C04 · MSC-P003-C05 · MSC-P003-C06 · MSC-P003-C07 · MSC-P003-C08 · MSC-P003-C09 · MSC-P003-C10 · MSC-P003-C11 · MSC-P003-C12

    Scientific question and scope

    One declared object: parameter confidence interval, future-observation prediction interval, posterior credible interval, or simulated outcome distribution

    09

    Declared calculation

    1. 01

      Validate the page-specific slice, units and positive standard error before calculating.

    2. 02

      Compute the margin z × SE and subtract/add it to the estimate.

    3. 03

      Label the result as a parameter confidence interval and refuse prediction or posterior interpretations.

    10

    Numerical example or application case

    Worked example — synthetic data or declared parameters; no real observations

    Sealed inputs

    • estimate_pp=3.2 [percentage_point]
    • standard_error_pp=1.1 [percentage_point]
    • z_critical=1.95996398454 [dimensionless]

    Reproducible results

    • CI95=[1.044040; 5.355960] percentage points

    Verified dossier: claims are linked to passage-level sources, the Python/R calculation is reproduced, limits are explicit, and independent scientific review is complete.

    11

    Validity assumptions

    • Does the standard error correspond to the declared estimator and sampling design?
    • Is the nominal confidence level fixed before examining the result?
    • Is the inferential target a parameter rather than a future observation or posterior quantity?

    12

    Diagnostics and uncertainty

    Diagnostics and uncertainty

    Match interval to target · Declare repeated-sampling or posterior interpretation · Check coverage or calibration · Expose model and horizon

    Bounded diagnostic

    Parameter confidence interval; neither prediction interval nor posterior interval.

    Explicit uncertainty contract · quantified

    Method: Deterministic reference-engine calculation on the sealed input slice.

    Target: Sampling uncertainty of the declared parameter estimate.

    Engine evidence: metric.ci_lower_pp= · metric.ci_upper_pp=

    Interpretation: The 95% confidence interval is conditional on the supplied standard error and is neither a prediction nor a posterior interval.

    Stop when

    Interpret an interval as a guarantee for an individual case.

    13

    Result interpretation

    • Compare estimates with their precision and assumptions.
    • The result is conditional on the declared population, horizon, specification and diagnostics. It must not be extended to another decision without new justification.

    14

    Supported and forbidden conclusions

    Supported

    Compare estimates with their precision and assumptions.

    Forbidden

    Interpret an interval as a guarantee for an individual case.

    15

    Possible marketing decision

    1. 01

      Compare estimates with their precision and assumptions.

    2. 02

      The result is conditional on the declared population, horizon, specification and diagnostics. It must not be extended to another decision without new justification.

    16

    When to use — when to stop

    Use when

    Compare estimates with their precision and assumptions.

    Result × inferential target × data-generating process

    Stop when

    Interpret an interval as a guarantee for an individual case.

    Methodological alternatives

    • Use a model-specific prediction interval for a future observation.
    • Use posterior quantiles only under an explicitly declared Bayesian model.
    • Use simulation quantiles when uncertainty is propagated through a declared decision model.

    17

    Reproducibility contract

    Python and R are the executable references. SPSS and SAS remain secondary syntaxes until checked on the same data, specification and diagnostics.

    The engine selects only this page’s explicit slice and branch. Its hash, outputs, and diagnostics remain sealed in the atomic dossier; no generic fallback branch is allowed.

    Slice fingerprint: d4ab5c6b8263b73b956b8303aaf4e6f79531f3d09737b32fedba4f0a8d8602b8

    Formal model

    CI = [estimate − z × SE; estimate + z × SE]

    Verified dossier: claims are linked to passage-level sources, the Python/R calculation is reproduced, limits are explicit, and independent scientific review is complete.

    18

    Expected final deliverable

    1. 01

      How should uncertainty in a marketing result be expressed? — Result × inferential target × data-generating process

    2. 02

      estimate · SE · z · CI

    3. 03

      CI = [estimate − z × SE; estimate + z × SE]

    4. 04

      Match interval to target · Declare repeated-sampling or posterior interpretation · Check coverage or calibration · Expose model and horizon

    5. 05

      Compare estimates with their precision and assumptions. / Interpret an interval as a guarantee for an individual case.

    19

    Scientific sources and evidence status

    Verified dossier: claims are linked to passage-level sources, the Python/R calculation is reproduced, limits are explicit, and independent scientific review is complete.

    1. MSC-P003-C01 — Advance specification covers both the research question and the analysis plan before outcomes are observed.
    2. MSC-P003-C02 — The estimator and analysis rules must be precise enough for independent replication.
    3. MSC-P003-C03 — The estimand is the target, distinct from the estimator used to calculate it.
    4. MSC-P003-C04 — Business decisions require more than crossing a p-value threshold.
    5. MSC-P003-C05 — Statistical significance does not quantify effect magnitude.
    6. MSC-P003-C06 — A p-value is not treated as a complete evidence measure.
    7. MSC-P003-C07 — Prediction intervals and parameter uncertainty answer different questions.
    8. MSC-P003-C08 — Predicted-value uncertainty requires its own reported scale.
    9. MSC-P003-C09 — Prediction intervals concern a future observation rather than only a parameter.
    10. MSC-P003-C10 — Posterior intervals summarize uncertainty under a Bayesian model.
    11. MSC-P003-C11 — A central posterior interval is defined by posterior quantiles.
    12. MSC-P003-C12 — Bayesian posterior-interval interpretation is conditional on the data and model.
    1. NIST DATAPLOT FIT manual
    2. Stan posterior_interval documentation
    3. Nosek et al., 2018
    4. ICH E9(R1), 2020
    5. ASA, 2016

    Method connections

    Parent territoryStatistical decision methods: choose, quantify, validate

    Read next

    MSC-P-004Statistical significance or effect size: which result should be interpreted?→MSC-P-012How many observations does an experiment need?→MSC-P-022How do you estimate price elasticity and its uncertainty?→