Calculate Cumulative Unique Program Reach

Evidence and applicability: use only supplied evidence for source values, historical facts, quotations, completed checks and approvals. Cite source labels and row, page or section references. Distinguish observed facts, source claims, analyst calculations, assumptions and proposals. Keep missing evidence unknown or not evidenced; it does not prove that an activity did not happen. Do not invent values, verification, agreement requirements, authority or sign-off. Apply only rules actually supplied and applicable to the task. Descriptive reach and monitoring totals do not demonstrate outcomes or causal impact.

Use only material supplied in this conversation. Ask me for the actual reach tables, definitions and overlap evidence before calculating a unique total. Treat missing overlap as unknown rather than evidence of disjoint participation.

Before calculating a unique-person total, require the exact reach definition and eligibility rules, population/universe, reporting window, partner/site coverage, and actual safe source totals or tables with versions and references. Establish whether each total counts contacts, records, households or already deduplicated people. Ask what is known about repeats within and across partners and periods, whether overlap tables cover the same universe and window, and which corrections supersede which records. Prefer authorized aggregate overlap evidence. Do not request direct identifiers. Only use minimal pseudonymous linkage results if the user confirms they may be used in this environment; pseudonymous data can still be sensitive. Do not infer consent or authorization, design a new linkage system, or promise reliable deduplication from weak keys.

Calculate only the cumulative reach measures supported by the supplied evidence.

1. Make an aggregation specification: counted entity, eligible population, window and cutoff, source coverage, units, inclusion/exclusion rules, handling of missing reports and revisions, and permitted overlap evidence. Preserve source totals and references. Mark incompatible or disputed inputs rather than treating them as zero or silently reconciling them. Do not add original and replacement reports together.
2. Keep separate service contacts or attendance episodes, distinct people within each source/period, and distinct people in the union across the requested partners/periods. A record count or sum of within-period people is not cumulative unique people. Repeated contacts may be additive only when their definition and coverage support it; do not assume this proves any unique-person count.
3. Where verified comparable set totals and complete overlap evidence support a union, show the set definition, formula, substitutions, result, window and source references. Subtract an overlap only once under a valid accounting scheme. Pairwise overlaps alone may be insufficient for three or more sets because higher-order intersections matter. If using mutually exclusive membership groups, confirm their completeness and exclusivity before summing. Report inconsistencies such as impossible intersections instead of forcing a result.
4. If overlap is unknown, report cumulative unique reach as unknown and retain the supported component totals. You may provide conditional bounds only when their assumptions are supported and explicitly stated. For comparable exact within-set unique-person counts in the same eligible universe, the largest set count is a lower bound and their sum an upper bound on the union; cap by the eligible population only if that population is known and covers all sets. These are bounds, not a verified total or an estimate. Do not apply them to contacts, household counts, incompatible windows or incomplete within-set deduplication. Unknown or partial coverage may prevent a program-wide bound even when the observed sources support one. Use additional overlap constraints only if justified and explain how they tighten the interval.
5. Calculate a coverage percentage only if a valid denominator matches the numerator population and window; show both values and their sources. Do not sum percentages or average partner rates into program coverage. Keep mutually overlapping subgroup totals separate unless a justified union is available. Avoid identifying or small-cell outputs that could expose people.
6. Deliver a source inventory, supported calculation worksheet, overlap/unknowns ledger and concise reporting note distinguishing known contacts, within-period people, any supported union or conditional bounds, and remaining limitations. Include reproducible rules and targeted requests for missing aggregate evidence. A calculation does not verify upstream data, prove program effectiveness or authorize publication. Do not fill missing data with synthetic values.