We cannot measure regeneration. But regeneration produces fruit, and fruit leaves marks on communities that we can observe and track. This page states what each number on the platform reads, in the order you would meet them.
The Gospel Impact Metric is the movement in a tract’s Shalom composite between two readings: GIM = Shalom (latest era) − Shalom (previous era). For the current U.S. surface that is 2024 minus 2020. Positive means the ground is opening. Negative means it is hardening. Around zero means the place is holding where it was.
Impact happens over time, so the metric that carries the word is a subtraction. A neighborhood that has been hard for forty years and one that became hard last year are not the same field, and no standing score can tell them apart.
Three names, two numbers. Several other layers are built on something else entirely, and section 06 names each one.
The central question: where is Gospel fruit becoming visible in our neighborhoods — and where is its absence most urgent?
The chief evidence of conversion is habitual practice, not emotional experience.
Repeated across many lives, practice aggregates into patterns you can see at the scale of a neighborhood.
Shalom is not only a spiritual category. Its absence is visible.
Shortened life expectancy, unsafe streets and persistent poverty are diagnostic, not merely social.
Lydia’s conversion was invisible. Its ripple was not.
Her household was baptized and her home opened. That is the shape of what we watch for.
Regeneration is unmeasurable. Gospel fruit is not.
The church layer is the causal middle, and the one link GospelGraph cannot currently observe. What it measures is the far right of the chain, community effects watched over time. The pathway to them stays a theological claim.
We do not claim to measure salvation. Where the Gospel takes genuine root over time, we should expect observable changes in the conditions of a neighborhood.
Ministry gets measured at three levels. Each asks a different question. GospelGraph reads one of them, and it is worth being clear which.
You are holding one side of a comparison. Community need is on the page; what the church there is actually doing is not, and no layer here supplies it. A neighborhood portrait is worth having on its own terms. It just answers what is this place carrying rather than how are we doing, and only you can put the two together.
Every eligible U.S. census tract is scored across five domains, each 0–100 per era, each carrying its own national weight. Higher means greater structural shalom. The weights below were fitted to the data rather than chosen around a table.
The ratio has nothing to do with which one matters more. Each weight reflects how much independent signal its domain contributes to the composite, given what can actually be measured at tract scale. Economic indicators are dense and refresh annually; safety and cohesion indicators are sparser and noisier at this resolution. A weight describes how well we can measure a domain, not how much it matters.
Five steps take raw public indicators to a composite. The fifth is the one most indices skip.
Each indicator is converted to a consistent scale so different units — percentages, rates, dollar amounts — can be combined fairly.
Some indicators are good when high, others when low. Direction is flipped so that a higher domain score always means greater shalom.
Indicators roll up into one 0–100 score per domain per era. A domain with no scored constructs returns null with a reason, never a zero.
The five domain scores are combined using the national weights in Section 03 to produce one composite per era.
Tracts below the population or coverage gate get no composite at all. They carry a reason code instead, which travels with them through every surface that reads the file.
The refusal is most of what distinguishes this from an index that always has an answer. The last row below is the real GIM footprint, since a change needs two readings.
A domain score is a rank against the national distribution rather than an absolute quantity, so a 70 means “higher than 70% of the country.” ACS margins of error are real, and a rank near the middle can move several points from one year to the next without anything changing on the ground. Treat small gaps between two tracts as noise.
A Shalom composite is not a spiritual grade for a neighborhood. A high score does not mean God is at work there and not elsewhere. It tells you about observable conditions — nothing more, nothing less.
Gospel Shift takes every scored neighborhood in the country, sorts them by how far the Shalom composite moved between 2020 and 2024, and cuts that line-up into five groups. The labels sound like judgments — hardening, holding, opening — but they mark position, not condition. “Hardening” means a neighborhood moved worse than most others did. By itself it does not mean the place got worse.
It used to say exactly that, and it was wrong to. It read like something we had found out about America. It was arithmetic. The groups are cut at fixed places in the line-up: the bottom 30%, the next 20%, the middle 10%, and so on. Thirty percent of neighborhoods land in “hardening fast” because that is the name we gave the bottom 30% — not because we counted them there. Had every neighborhood in America held perfectly still for four years, this page would print the same five numbers.
The ranking is still worth having. If your neighborhood sits in the bottom 30%, it moved worse than seven in ten others, and that comparison is real. A tract reading −0.5 is not “slightly bad” — it is doing better than half the country. Read the position. Be careful with the word attached to it.
There is a second reason to hold the words loosely. The whole line-up sits left of zero, and part of that is the survey rather than the country. The 2016–2020 Census release — the base this subtraction measures from — was collected through the first year of the pandemic, and across the part of the model we can carry back to 2013 it sits about two points high. Subtract from an inflated base and everything tilts negative. That is a known weakness of comparing two snapshots, and it is most of the reason the eleven-year view exists.
Every Shalom score is built from a survey, and every survey carries error. Subtract one score from another and both errors come with you into the answer. For an ordinary neighborhood the four-year change is known to within about seven points either way, nineteen times in twenty. The entire holding band is one and a half points wide.
About three-fifths of what separates one neighborhood’s reading from another’s is real difference between those two places. The rest is the survey. That is enough to rank with. It is not enough to pronounce with, and that distinction is the whole of the advice on this page.
Comparing neighborhoods as they stand today. Gospel Shift. The line-up is fair, and a gap of several points between two tracts is telling you something.
Asking which way a place has been going. The eleven-year view. It is built for that question and carries about half the error of the number above — nearer a third of it when the two are measured across the identical set of constructs, since they are fitted on different shares of the model. Where it cannot establish a direction it says so rather than guessing.
Deciding something that will cost people years. Neither one on its own. Go and stand on the street.
None of this establishes a cause. Conditions are shaped by policy, markets, migration and investment, on timescales that swamp any single congregation. A rising GIM in a tract where your church has worked for a decade is an encouragement, and a good reason to go asking why. It is not evidence that the church produced it, and no layer here will be built to suggest otherwise.
GospelGraph carries more than one measurement, and they are never blended into each other. Each entry below names what it reads, and whether the Shalom composite is underneath it.
Structural Gospel-access gap, 0–100, higher meaning thinner evangelical access. Tiers: Watered, Sparse, Dry, Arid, Desert. Not Shalom-derived.
100 minus the latest Shalom composite. A standing level. Shalom-derived, and the basis for the GDI × need matrix.
The 2024 composite minus the 2020 composite. Shalom-derived. The map label and the metric name are the same number.
Whether the physical neighborhood supports repeated contact: third places 45%, intersections 30%, public rooms 25%. Not Shalom-derived.
Named structural patterns flagged per tract. They are descriptive tags for triage. Nothing scores them and no composite reads them.
Household-weighted composition from a catalog of 30 personalities and 12 context modifiers. It describes who lives on a block. It scores nothing and ranks nothing. See Section 07.
Every layer above describes a place: what it carries, which way it is moving, how thin Gospel access is, whether the streets support running into people. None of them describes a congregation. If a number on this platform seems to be telling you how your church is doing, it is being read wrong.
The scores tell you what a place is carrying. They do not tell you who is carrying it. That comes from market segmentation: a catalog of 30 neighborhood personalities and 12 context modifiers, matched to a tract by household composition rather than by its average.
A personality describes a household pattern — how people live, what stage they are in, what their week looks like. Multifamily First-Timers and Affluent Owner Family Suburbs can want very different things from a church on the same street. Modifiers sit on top as conditions the personality does not capture on its own, like displacement pressure or household crowding.
A personality is not a score, and no tract is prioritized because of the personalities in it. None of them is better or harder or more deserving than another. It answers “who is here” so the numbers in the rest of the report land on actual people rather than on an average.
The analysis produces real and useful information. It also has real and important limits. Both deserve plain statement.
The composite tells you what kind of place this is, and the delta tells you which way it is moving. Neither tells you what the church there is doing, because that is the part we are not currently measuring. Read a portrait as if it were a diagnosis and you will act on a question nobody here has answered.
These guardrails are not disclaimers. They are built into the theological design.
For a pastor or ministry leader, here is the practical workflow from start to finish.
Name the decisions you are trying to make before you look at any data. Vague questions produce noise.
Start with need pressure and the five domain scores: what is this place carrying, and which domain carries most of it? A composite will not tell you what to do on Tuesday. The domain ordering will.
Read GIM as a place in the national line-up rather than as a verdict: the five band shares are fixed by where the cuts fall, so they are not a count of how the country fared. A hard tract that is climbing the line-up and a comfortable one sliding down it call for different responses, and the standing score cannot tell them apart.
Before you carry a number into a board meeting, check Section 06 for what it reads and whether the Shalom composite is underneath it. Some layers on the map score nothing at all.
Bring the scores to site pastors, deacons and neighbors who know the ground. They can tell you what a tract is carrying; they cannot tell you why. Then set a rhythm: ACS indicators update annually on a rolling five-year average, so read trends over three to five years rather than reacting to one vintage.
Everything on this platform is built from three public sources. You can check any of them yourself, and none of them costs anything.
5-year estimates: poverty, income, unemployment, housing cost burden, vacancy, education, family structure, demographics. A major input to Shalom, and the base the neighborhood personalities are matched on.
Church locations and the physical fabric of a neighborhood — third places, intersections, public rooms — mapped by volunteers and joined to tract geography.
NTEE-coded religious nonprofits, deduplicated against the map points. With OpenStreetMap this feeds the Gospel Desert Index and the built-form layer, not Shalom.
Where a measure is unavailable we either substitute a named proxy and say which one, or leave the field null with its reason attached. We never invent a value, estimate one from its neighbors, or carry an older vintage forward to cover a gap in a newer one.
The Hood Shepherd built this framework — and applied it to every Census tract in the United States — because faithful incarnational ministry deserves the same rigor we would bring to any other serious endeavor. We count because counting is an act of attention, and attention is an act of love. But the numbers serve the Word. They serve the pastor who walks the block every morning, the deacon who sits with the family in eviction court, the elder who prays over the neighborhood by name.