Bound a conversion share when some outcomes are unknown

An unknown outcome should not silently become a negative outcome. This lab shows the lowest and highest possible success share when a subset of eligible records has missing outcome information. The range is an accounting bound, not an estimate of where the truth is likely to fall.

Define the calculation before using it

Start with one eligible population split into observed successes, observed failures and unknown outcomes. The lower bound assigns every unknown to failure; the upper bound assigns every unknown to success. Keep the unknown count separate from a genuine zero-valued observation. This distinction matters before choosing an imputation model or comparing groups with different observation coverage.

Work through the synthetic example

For 1,000 records, 200 observed successes and 100 unknown outcomes, the success share lies between 20% and 30% under this binary accounting. The observed-known rate would be 22.22%, but that rate describes only the 900 known records. Reporting it as the full population’s rate adds an assumption about missingness.

Use the artifact and preserve its assumptions

Open the editable calculator to change the inputs and inspect the sensitivity view. The CSV records synthetic inputs and expected outputs; the JSON fixture keeps the equations available for reproduction. These calculations have been checked against the stated example. No measured campaign data is included.

The sensitivity rows vary only unknown by 20% below and above the entered value. They are scenarios, not confidence limits or a forecast distribution. A row outside the model’s constraints is labelled rather than turned into a plausible-looking result. Save the chosen inputs with the decision so another reader can distinguish a changed assumption from a changed formula.

Evidence and limits

Google’s export schema separates the reporting date, UTC timestamp and event parameters. The worksheet uses simplified aggregate inputs; it does not claim that an export already contains a correctly reconciled cohort. See Google Analytics BigQuery Export schema, especially “event_date, event_timestamp, event_value_in_usd and event_params fields”.

The bounds require a correct eligible population and mutually exclusive states. They do not identify why signal is missing or model privacy-related suppression, attribution eligibility or selection bias.

Background: UA metrics explained: CPI, ROAS, LTV and payback and Reading an MMP dashboard without fooling yourself. These existing articles provide context; the present calculation does not verify every archived claim.