Data Methodology

Data sources

We aggregate data from trusted government and public health sources.

• CDC PLACES — health metrics. Census tract-level health data including chronic disease prevalence, mental health indicators, healthcare access, and social determinants of health. cdc.gov/places
• Census ACS — economic and demographic data. American Community Survey data including poverty rates, unemployment, housing costs, education levels, and family structure. census.gov/programs-surveys/acs
• TIGERweb — geographic boundaries. Census tract boundaries and geographic data from the US Census Bureau's TIGERweb service for accurate spatial analysis.
• Overture Maps — addresses and places. Open address and place data, used to locate churches and attach each one to the census tract its building sits in. overturemaps.org

Credits and licences

Open datasets this site is built on, how each one is licensed, and what it contributed.

Understanding severity thresholds

Rather than using arbitrary cutoffs, our thresholds are calibrated against national averages. A metric is flagged as “concerning” when it significantly exceeds what's typical for the United States as a whole.

Severity levels

• Low — Below or at national average
• Moderate — Slightly above national average
• Concerning — Notably above national average; prayer focus
• Critical — Significantly elevated; urgent need

Economic and family thresholds

Health metric thresholds

What we don't publish

We would rather say nothing than say something we can't stand behind.

No crime statistics, anywhere on this site

We hold crime data and it does not survive inspection. Three sources report at incompatible scales while all labelling their figures “per 100,000,” and one series reads 43,809 assaults per 100,000 — which would be 44% of residents. Until that can be reconciled it is blocked from every published page, and no threshold built on it would mean anything. There is no public safety metric on a church page, in prayer mode, or in the data a search engine reads.

No estimate too imprecise to state as fact

Census figures are survey estimates, not counts, and each one arrives with a margin of error that we check before showing anything. If the range around an estimate is wider than 40 percentage points, or a figure reads 0% or 100% on a sample of fewer than 50 households, we print no number at all — that silences roughly 47% of housing-cost-burden tracts and 20% of child-poverty tracts. Where a range is wide but usable we show the range, never the bare figure.

No filling in a gap that isn't ours to fill

When a measure is missing for an area we say so on the page rather than leaving a blank that reads as zero. Six social-needs measures are absent across eleven states because those states never fielded the survey module — no amount of work on our side can close that, so we tell you instead. We also never mix vintages: where a measure is published for different years in different states, each page reads the year that applies to it and says which year that is.

How community needs are selected

Our approach to presenting diverse, meaningful prayer needs.

1. Data aggregation — for each map view, we identify all census tracts visible on screen and calculate average values for each health, economic, and social-needs metric.
2. Threshold filtering — only metrics that cross the “concerning” threshold are considered for display. This ensures we highlight genuine needs, not just any data point.
3. Category balancing — we ensure diversity by selecting one need from each category (Family, Economic, Health, Mental Health, Access) before filling remaining slots by severity. This prevents any single category from dominating.
4. Daily rotation — prayer prompts rotate daily using a hash-based system, ensuring that returning visitors see fresh content while maintaining consistency during a single prayer session.