Church Metrics That Matter: The Nine Beyond Sunday Headcount

The church metrics that matter measure how often people come, not how many came: attendance frequency per household, first-time guest return, group participation, serving share, giving participation. A headcount tells you the room was full. It cannot tell you who left it.

Key points

• Nine metrics cover almost every leadership question, and all nine are computed per person or per household rather than per Sunday.
• Lifeway Research’s State of Discipleship study, a survey of 2,620 US Protestant pastors conducted in September 2024, found that while 71% believe there are ways to measure discipleship, only 30% say their church has specific methods for doing it.
• Attendance erosion is now a frequency problem: Gallup put weekly or near-weekly attendance at 31% in 2025 while monthly attendance held steady at 10%.
• The National Congregations Study (2018-19) has the median share of regular adult participants serving in a leadership role at 20% in the congregation the average attendee sits in, unchanged across four waves since 1998, and 30% from the average congregation’s perspective. One of the few real benchmarks in this list.
• Most of these metrics have no national benchmark at all, first-time guest return rate included, so the comparison that matters is your own church twelve months ago.

Ask a church leader how many came on Sunday and you get a number immediately. Ask which nine numbers they would take to a board meeting and the list usually stops at three: attendance, giving, and whatever crisis is current.

That isn’t a discipline problem. It’s a measurement problem. Attendance and giving are the two things a church counts as a by-product of doing them, and everything else has to be deliberately produced.

Why doesn’t a Sunday headcount tell you anything?

Because it’s a total, and totals cancel. Eleven families drift out over three months, a dozen new ones start coming, and the number holds. Nothing in it tells you eleven families are gone. The arithmetic quietly nets them off, which is the whole reason finding out who stopped coming is a spreadsheet project rather than a report.

There’s a second problem, and it’s newer. The people who used to be your weekly attenders haven’t left; they come less often. Gallup found that in 2025, 31% of Americans said they attended weekly or nearly weekly while attendance about once a month “has been relatively steady” at 10%. The erosion is concentrated at the weekly end. A church whose regulars each drop from four Sundays a month to three loses a quarter of its headcount without losing a single family.

A third problem is that Sunday is no longer one channel. Pew’s 2023–24 Religious Landscape Study found that of the 40% of US adults who participate at least monthly, 16% do so both in person and online, 17% only in person, and 8% only online. The Hartford Institute’s attender survey puts it more sharply: 19% “move between online and in-person attendance”. Those are the people your headcount misclassifies from one week to the next.

You can watch this happen in one dataset. In the 2025 Faith Communities Today survey of 7,453 congregations, the median change in in-person attendance over five years was 0%. Flat. The median change in combined in-person and virtual attendance over the same five years was 11% growth. Same congregations. Two different stories, decided entirely by which number you put on the slide.

The same survey asks how many unique individuals participate at least monthly, and attaches a note worth reading twice: “this number is typically higher than average weekly worship attendance.” Median 76 monthly participants against a median in-person attendance of 70. Your roll and your room are not the same size, and only one of them is being counted.

The nine church metrics that matter

Nine is not a magic number. It’s what fits on one page and still answers the questions a board actually asks. Every one of them is computed per person or per household, and every one of them can be turned back into a list of names. That’s the test for keeping a metric at all.

The right-hand column is the honest part. Planning Center records all nine of these faithfully. It reports almost none of them as a rate, because a rate needs a numerator from one product and a denominator from another, and Planning Center’s products don’t compare notes.

Rhythm: how often, not how many

Attendance frequency by household

The single most useful number in this list, and the one no church has. Take thirteen weeks, count the distinct Sundays each household appeared, divide by thirteen, then do the same for the thirteen weeks before. The difference is the metric.

A household at 0.75 that drops to 0.25 has not missed a Sunday in any way a report would flag. They’ve changed what your church is to them. Nothing built into Planning Center compares a person’s frequency between two periods, and the attendance reports post walks through every report that gets close and stops short. The Activity tab in Check-Ins gives you exactly this view for one person at a time, twelve months, with a CSV export. One person at a time is the constraint.

Read it by household, not by head. Four people leaving as a family is one phone call. The catch is that the Household ID never survives an export and exists only inside an import file, so the collapsing is manual work you should plan for rather than discover.

Regular participants

Count the distinct people with any check-in in the last 30 days. It will be larger than your weekly attendance and smaller than your database, and it is the correct denominator for most of the rates further down this page.

The Check-Ins event chart graphs the number of unique people checked in per date, and exports those per-date counts as CSV. That export does not get you the number: someone who came three Sundays sits in three rows, so summing them overstates the distinct people in the window. Either read the Home attendance widget, which deduplicates over its own timeframe, or export person-level check-ins from a custom report and count distinct Person IDs. That also gets you the list of who those people are. Worth remembering that Headcounts records attendance without names, so any room you count with a clicker contributes to your headcount and nothing else on this page. Planning Center’s own guidance is to use Check-Ins or Groups if you need to track individuals.

First-time guest return rate

First-timers in a month who came back within 28 days of their own first visit, over first-timers that month. A guest count measures one Sunday’s welcome; the return rate measures whether anything happened afterwards.

Check-Ins gets you the numerator’s raw material: a custom report with the Scope set to first timers only returns everyone who attended that event for the first time. The return half is a People list with two conditions: first check-in inside a month-long window, and a second check-in between that window’s start and four weeks after its end, run once on that date, because a list can’t bound a window to each guest’s own first visit and “any check-in since” counts returns from any time later. First-time guest tracking works through the bounds; the two-condition mechanic is the method the Check-Ins analytics post sets out, including the trap where label-only visitors never reach your People database at all. First-time guest tracking takes the follow-up workflow further than this section can.

Participation: what share of the church is in something

Group participation rate

Adults on an active group roster who are also regular participants, over regularly participating adults. That’s the overlap, matched on Person ID, rather than the whole roster. Someone attending for a year with no group can disappear without anyone having a reason to call; someone in a group is known by eight people who will notice within a fortnight.

In the 2025 FACT survey, congregational leaders estimated that 34% of their regularly participating adults are part of small groups. That is a mean of leaders’ own estimates across congregations, so treat it as a rough marker rather than a target.

Groups holds the numerator well. A group’s Attendance tab shows each member’s attendance percentage relative to the number of events in the chosen timeframe, and the Reports page tracks attendance across multiple groups with average percentage attended, typical attendance and event counts, exported as a CSV emailed to you. The denominator is in People, which is why this is a two-export rate. Small group metrics goes group by group.

Share of adults serving

Regular participants who also served in the last 90 days, over regularly participating adults. That’s the two lists intersected on Person ID, not the Services roster over your adults. This is the metric with the best available benchmark, and it comes in two flavours that are easy to confuse.

The National Congregations Study summary tables report the median percent of regular adult participants “serving in leadership role in past year” at 20.0% in the congregation the average attendee sits in. That figure is 20.0% in all four waves, 1998 through 2018–19, with no statistically significant trend. Twenty-one years, flat. In the median congregation, which is far smaller, the same row reads 30%.

General volunteering is a wider net and a higher number: the 2025 FACT survey has congregations estimating that 43% of regularly participating adults volunteer regularly. Leadership and volunteering are different denominators. Pick one, define it in writing, and never let it drift between board meetings.

Giving participation rate

Giving-age adults with at least one gift in twelve months, over giving-age adults. Match donor keys against that list, not every donor in the export. The clearest read on whether generosity is broad or borne by a few.

The closest thing to a benchmark comes from the Lake Institute’s 2019 National Study of Congregations’ Economic Practices, and the scope matters: among congregations that ask participants to make a pledge, on average 53% of regularly participating adults made a pledge and 87% of regular participants made at least one gift in the past year. That 87% is conditional on congregations running a pledge campaign, not a figure for all churches, and dropping the condition would make it false.

Two Planning Center facts shape how you compute this. Giving reports per donor, not per household, and linking two people as joint donors is a manual action covering two people whose combined history surfaces in statements rather than in the By donor report. And donor lists do not sum gifts: a rule for people who gave over $100 returns single gifts over $100, not households who gave $30 four times. Church giving analytics covers the full set, and finding lapsed givers is the worked example of the retention half.

Direction: what every number has to be compared against

Same period last year

A month is mostly weather, school holidays and one long weekend. Compare this month with the same month last year, and keep a rolling 13-week average beside it so a single Easter doesn’t read as growth. The Check-Ins headcount report holds up to 53 past sessions, a year of Sundays plus one. That is a deliberate ceiling: one report can hold this week against the same week last year. Church attendance trends is about reading that line without fooling yourself, and how to track church attendance covers what to do if the underlying data isn’t there yet.

Volunteer availability

Volunteers rarely resign. They decline a shift, then decline another, then block out three weeks. Services records both halves. The Statuses of People Scheduled report gives a breakdown of who is confirmed, unconfirmed and declined across a date range, the Blocked Out People report shows who declined because of a blockout, and the Timecard report totals the time each person served. Those reports come out as PDF or web page, not CSV, and each describes the current state rather than the change, so the quarter-against-quarter comparison exists only if you export on a fixed date each quarter and keep the files.

Serving is usually the first thing to move, months before attendance or giving. That’s the argument in volunteers don’t quit, they stop being available, extended into a full metric set in how to measure volunteer engagement.

How do you compute an attendance-rhythm number this week?

The one metric worth building by hand first, because everything else on this page is easier once you’ve done it.

1. Pick two equal windows. Thirteen weeks against the thirteen before. Long enough to survive a family holiday, short enough that the answer is still actionable.
2. Export person-level check-ins. A custom report from the Reports tab on your weekend event, as CSV. The event chart’s export is counts by date and will not work here.
3. Check the file’s shape before you go further. You need one row per person per date. If the export has collapsed to one row per person, the dates are gone. Run the report per session and stack the files.
4. Keep the Person ID column. Planning Center exports include a Person ID unique to each profile; match on it, never on name.
5. Add a week-number column from the check-in date, so one person checked into two sessions on the same Sunday counts once. This dedupes a person, not a family. Each child has their own Person ID and their own rows, and the parent who dropped them off may have none.
6. Add a household key, then reduce to household-and-week pairs. Assign the key by hand (surname and address, since Household ID doesn’t export), and collapse the rows so a family is present that Sunday if any member is. Do this before anything is counted: two children who each made five different weeks are a household that made nine, and a floor applied to the children would drop a family that was actually regular.
7. Pivot by household with two measures: distinct weeks in the earlier window, distinct weeks in the recent one.
8. Convert both to rates, weeks attended over weeks available, and subtract.
9. Filter to households that were regulars first. A family that came twice in the earlier window cannot meaningfully drop. Set a floor around six of thirteen, then sort by the size of the fall.
10. Check the top twenty against serving and giving before anyone picks up the phone.

Budget half a day the first time. If you’d rather keep it live, exporting Planning Center into Google Sheets covers that route.

There is one genuinely useful way to make a Planning Center number trend on its own. Build a People list for the thing you care about, since lists take dated conditions across products, and Planning Center documents auto-refresh on a nightly, weekly or monthly schedule. Then add the List results widget to a Home dashboard, which shows list result totals over a timeframe. Planning Center introduced it in June 2025 precisely so you can “visualize list activity over time” and compare with the previous year. The list itself still answers as of now; the widget is what remembers. Lists must be at least 24 hours old before the widget will show anything, so set it up a day before you need it.

That is the ceiling of what’s built in, and it’s higher than most churches realise. Past it, the recurring cost is the problem rather than the difficulty: the pivot table you built in March doesn’t refresh, and a drop-off list from six weeks ago is a list of conversations you’ve already missed. That’s the job Parable does. Every Planning Center product is read nightly into one place where they share a person key, so these nine sit on a church metrics dashboard that is current on Monday morning, with SQL and warehouse access underneath for whoever would rather write the query.

Which of these has a national benchmark?

Almost none of them, and you should know which is which before you put a number in front of a board.

Real benchmarks exist for the share of adults serving (NCS 2018-19: median 20% in a leadership role from the average attendee’s perspective, flat since 1998; FACT 2025: 43% volunteering, from leaders’ estimates), small group participation (FACT 2025: 34%, same caveat), and giving participation among pledge-campaign congregations (NSCEP: 87%).

No credible benchmark exists for first-time guest return rate. The figures that circulate, “fewer than 15% of visitors return” and ladders of percentages by visit number, trace back to consultancies and church-software vendors citing each other, with no underlying study anywhere. There’s a structural reason: measuring it nationally would mean linking one individual’s first visit to a later visit across congregations, and no national instrument does that. National surveys ask congregations about themselves, or ask individuals without knowing which church they mean.

The same goes for household retention and time-to-first-serve. Giving concentration has no current benchmark either. What exists is a 1990s literature, useful as a shape to check yourself against rather than a target. If someone quotes you a number for any of them, ask for the study.

This is less of a loss than it sounds. The comparison that changes decisions is your own church twelve months ago, not a national median you share a denomination with by accident. Build twelve months of your own history for two or three of these and you have a baseline nobody can argue with, and one that is actually about your people.

The most current national picture makes the point better than any benchmark would. Hartford’s 2026 EPIC report found median in-person attendance rising for the first time in twenty-five years, and immediately cautioned that “nearly half of congregations continue to experience attendance decline”, with co-investigator Allison Norton summarising it as “not a revival — it’s a recalibration.”

Half of churches are in the good half of that sentence and half are not, and no national median will tell you which one you’re in. Nine numbers, computed on your own people, will.

Start with two: attendance frequency by household, and the share of adults serving. Both are computable this week from data you already have, both take about half a day, and between them they catch most of what a headcount hides. Add the rest when someone asks a question those two can’t answer. And if you want the version a board meets around rather than the version you rebuild each month, the seven metrics a board should see monthly is the shorter list.