Most people go looking for a church dashboard after they’ve seen one. A conference slide, another church’s staff meeting, a screenshot in a Facebook group: a single screen with the numbers on it, and the quiet sense that the people who have that screen know something you don’t.
The tool is the easy part of this. The hard part is deciding what goes on the screen, and almost nobody tells you that part, because it requires having an opinion.
So here’s mine. Thirteen metrics, why each one earns its place, which ones are worth watching weekly and which are just scoreboard, and then the honest version of how you build it out of Planning Center: start with the cross-product dashboards in Home, and use a spreadsheet only for the custom calculations the widget catalogue does not express.
One rule for what gets on the screen
A metric earns a tile if a change in it would change what somebody does next week.
That’s the whole filter. Not “is it interesting.” Not “would the board like to see it.” If attendance drops four percent and nothing about anyone’s Monday changes, attendance isn’t a dashboard metric at your church, it’s trivia. Meanwhile a first-time guest number that nobody looks at until the elders meet in March isn’t a dashboard metric either. It’s a report, and reports are fine, they just don’t need a tile.
A dashboard is a standing set of questions you’ve decided to keep asking. Everything else is a one-off, and one-offs belong in Planning Center’s per-product reports, where they’re already well handled.
The thirteen
Attendance
1. Weekend attendance, thirteen-week rolling average. The mean of the last thirteen weekend totals, a full quarter, plotted weekly.
Not the single-week number. Weekend attendance has enormous week-to-week noise: weather, school holidays, a long weekend, one big kids’ event, and a single week can produce a false alarm. A four-week window smooths some of that but still follows short school-term rhythms. Thirteen weeks gives you a quarter-sized view at the cost of responding more slowly to a genuine change. Whether that trade is right depends on how quickly your team acts; keep the raw weekly number underneath so the average can always be sanity-checked against it.
2. Attendance frequency mix. Of your active households, what share attended three or more times in the last month, one to two times, and zero?
This is one of the most useful ways to read beneath a total. A church can hold its total attendance perfectly flat while the underlying behaviour changes completely: fewer people coming more often, or more people coming less often. Those are opposite situations and the headline number reports them identically. The mix tells you which one you’re in and can expose a change that the total hides. More on getting this out of Check-Ins in the attendance tracking guide.
Giving
3. Giving versus the same period last year, like for like. Year-to-date total against the identical span of the prior year, January 1 to today, against January 1 to the same date last year.
Never against a full prior year, and never against budget alone. Budget tells you how you’re doing against a decision somebody made in October. Prior-year like-for-like tells you what’s actually happening to generosity.
4. Giving households. How many distinct households gave in the last 90 days, and how that compares to the prior 90.
Total dollars can hold up while your donor base is quietly hollowing out, because two or three large households can carry a lot of ground. The count of households giving is the health metric; the dollar total is the outcome. Watch both, and worry about the count first. Its slower cousin, who gave last year and hasn’t this year, is a standing list worth keeping rather than an annual panic.
5. Recurring giving as a share of total. The percentage of the last 90 days’ giving that came through scheduled, repeating gifts.
Recurring share tells you how much giving is already scheduled rather than initiated one gift at a time. It is useful operating context, not a guarantee of future revenue or donor retention. The giving reports guide covers where these numbers actually live in Planning Center Giving.
Guests
6. First-time guests per week. New households recording a first check-in or a first connect card, thirteen-week rolling average again, and here the smoothing matters more, not less, because the weekly counts are small enough that a single family of five can look like a trend.
This measures whether anybody new is being captured by the systems included in your definition. Decide whether a returning family who lapsed for two years counts as new, write the rule down, and never quietly change it.
One capture rule decides whether the next metric is possible at all. A check-in created through the one-time-guest flow leaves no person record behind, and Home’s First time visitors widget excludes one-time visitors, so a return rate only exists at churches that create a person record for every guest they check in.
7. Guest return rate. Of the first-time guests from a given month, what percentage came back a second time within 30 days?
If you only get one guest metric, I would take this one over the count. The count measures capture at the front door; the return rate measures a defined next attendance outcome.
I’d be careful with published benchmarks here. A figure is only comparable when it states how it defined a first-time guest, the return window and the data source. Your own consistently defined trailing twelve months is often a better operational comparison than an external average built differently.
Connection
8. Connected share. The percentage of active adults who are in at least one of: a group, a serving team, or a giving record in the last 90 days.
This is deliberately generous, any one of the three counts, because the review population is people with none of those recorded connections. That may mean a real pastoral gap or simply incomplete data, so the metric produces names to verify rather than conclusions. Building it properly requires a custom definition across four Planning Center products, which is the subject of measuring engagement.
9. Six-month retention of new people. Of the households whose first contact was six months ago, what share is still active today?
A useful measure of assimilation, and one many churches do not track because it requires preserving a fixed arrival cohort. It moves the conversation beyond how many people entered the guest pipeline to what happened to them afterwards.
Serving
10. Active volunteers. Distinct people who filled at least one serving slot in the last 30 days, plotted weekly.
11. Serving concentration. The share of all filled slots covered by your top twenty percent of volunteers.
The second one adds context the volunteer count cannot. Many churches pull disproportionately from a small core; the question is whether that concentration is getting worse. A rising concentration does not prove burnout or predict a resignation, but it is a reason for a team leader to inspect workload, recruiting and whether the wider roster is actually being scheduled.
Groups
12. Group participation rate. The percentage of active adults in at least one group, with the raw count of active groups beside it.
Two numbers on one tile because they move independently. Participation with a flat group count can indicate fuller existing groups; a rising group count without participation growth can indicate smaller or newly launched groups. Inspect the underlying groups before assigning a cause.
Milestones
13. Baptisms, salvations, and membership classes: year to date, with the names behind them. A cumulative count for the year, against the same point last year, and the tile opens to a list of who.
Every other metric here is about a population. This one is about people, one at a time, and each entry on it is the day something changed in somebody’s life. That is the reason a church has a dashboard at all, and a screen that reports twelve things about behaviour while saying nothing about the moments that behaviour was meant to lead to is measuring the machine instead of the point of it.
Two rules make the tile work. Count cumulatively, never week over week: year to date against last year to date, so the number only ever climbs and nobody is asked to interpret a two versus a four. And keep the names attached, because the value of this tile is not the integer. It’s that a pastor can click it in October and read thirty-one names, and it’s the one number on the dashboard nobody has to be persuaded to care about.
What does not belong
Be ruthless here. A dashboard with thirty tiles gets read the way a wall of thirty clocks gets read.
Total people in your database. It only goes up. It counts the family that moved to Arizona in 2019 and the visitor who filled in a card once. It is the most-displayed and least-informative number in church software, and treating it as a membership figure has caused more bad board conversations than any other metric I can think of.
Single-week anything, presented without a trailing average. You’ll spend the meeting explaining the weather.
Baptisms and salvations as a weekly trend line. The metric belongs, it’s number thirteen above, but as a running year-to-date count, not a sparkline. At weekly resolution the numbers are small enough that noise swamps signal, and you’ll spend the meeting explaining a two versus a four instead of reading the names.
Social followers, app installs, livestream view counts. Unless you can tie a view to a person who then did something, this is a media metric wearing a ministry costume.
A single engagement score per person, on the dashboard. Composite scores are useful for sorting a list. On a dashboard they’re actively harmful, because the average moved and you cannot tell which of the six inputs moved it. Show the inputs.
Anything nobody owns. Every tile should have a name attached: a person who’d be asked about it if it moved. Tiles without owners are the ones that go stale first and get believed longest.
Leading and lagging
Some church metrics react on different cadences, and the ordering of your dashboard should make that distinction visible without claiming a universal sequence.
| Leading | Lagging | |
|---|---|---|
| Attendance | Frequency mix | Weekend total |
| Giving | Giving households, recurring share | Total dollars |
| Guests | Return rate | First-time count |
| Serving | Serving concentration | Active volunteers |
| Connection | Connected share | Six-month retention |
Milestones sit outside this table on purpose. They aren’t a signal to act on; they’re the outcome everything in the leading column is supposed to produce, which is why the tile is a scoreboard and not a trend.
Giving can lag other changes because a recurring gift may continue until somebody actively changes it. Treat that as a hypothesis to check against attendance and serving, not as a fixed sequence or a six-month rule.
Serving, attendance and giving do not change in a universal order. The value of the leading column is that each measure can prompt a timely human check; the dashboard should show which signal moved rather than turning that movement into a diagnosis.
Put the lagging metrics on the bottom row. They’re the ones the board asks about, and they belong on the dashboard. They just shouldn’t be what your eye lands on first.
Building it from Planning Center: the honest version
Start in Home → Dashboards. Planning Center Home can place widgets from nine products on personal or shared dashboards: Calendar, Check-Ins, Giving, Groups, Home itself (notepad, quick links and tasks), People, Publishing, Registrations and Services. That covers more of this blueprint than it used to: attendance and headcounts, donors and total donated, group attendance and memberships, registrations, scheduled volunteers, workflow activity and new profiles all have native surfaces. Build those first. One permission caveat worth knowing before you promise a screen to anyone: the Groups attendance and memberships widgets are Groups-administrators-only, and a shared dashboard renders on the creator’s permissions, so what your staff see depends on who built it.
Home is not an arbitrary calculation canvas. Frequency mix, guest return cohorts, serving concentration, connected share and custom retention definitions still need records combined or reshaped. For those remaining metrics, a spreadsheet is a workable manual layer. The following is the workbook architecture, not a copy-and-paste template: export columns and row grain vary by report, so document and test each input before trusting the formulas.
One workbook, nine tabs.
Weeks is the spine, and you build it first. Column A: every Sunday for the last 104 weeks, oldest at the top. Decide what the comparison means before you build column B. Subtracting 364 days gives you a weekday-aligned 52-week comparison, Sunday against Sunday, but it is not the same calendar date. If the metric needs the prior calendar date or prior-year-to-date boundary, calculate that explicitly instead. Label the choice in the workbook. Every other tab keys off column A.
in_attendance, in_giving, in_people, in_serving, in_groups are your five paste targets. One Planning Center export each, pasted whole, never edited by hand. If a column is wrong, fix it in the export and re-paste. The moment somebody starts patching cells in a raw tab, the workbook stops being reproducible and starts being folklore.
Milestones is the one tab that isn’t an export: one row per person per milestone, with name, person ID, type and date. If you run baptisms through a workflow or a form, export that. A workflow’s CSV export has been a documented download since February 2026, carrying the person, the date the card was added, total days in the workflow, the current step, card status and the assignee, capped at the 10,000 most recent cards. Most churches are keeping this in someone’s head or a Google Doc, and moving it into a dated, ID-keyed tab is the whole build for metric thirteen.
Metrics is one row per week and one column per metric, and contains nothing but formulas. SUMIFS and COUNTIFS against the raw tabs, keyed on the week date from Weeks. Every date-window formula needs both a lower and an upper bound so future records cannot leak into it. A thirteen-week average should explicitly guard incomplete windows, for example =IF(COUNT(B2:B14)=13,AVERAGE(B2:B14),"") for the first complete range, rather than relying on Excel to leave the first twelve rows blank automatically.
Dashboard is thirteen tiles: a three-by-four grid of metrics, with the milestone tile full width beneath it, where it reads as the point of the other twelve rather than one more number among them. Each metric tile: the current value large, two small comparisons beneath it (versus the thirteen-week average, versus the same week last year), and a 52-week sparkline. Leading indicators in the top row, lagging in the bottom. Freeze the header. Type no numbers here, ever. Every cell references Metrics.
The weekly ritual is five exports and a paste, and if nothing has changed it takes half an hour. Check-Ins for attendance and guests, Giving for donations in the period, People for your active roster, Services for filled slots, Groups for membership and group meetings. That is the weekly top-up, not the initial two-year fill, which is a different job entirely and is the first thing in the wall below. The export guide covers what each product actually gives you and the gotchas per file.
Two things to get right on day one. Match on a verified stable ID, never on name. Name matching breaks on Mike versus Michael, marriages and households that give under different spouses. People list exports settle half of this for you: since February 2026 the Person ID is always included in a list export rather than being a column you remember to add. Do not assume the other four exports use the same identifier or include it by default: inspect the real headers, document the key, and use the API or a tested mapping table where an export does not expose a common ID. Giving’s donor numbers are optional envelope numbers, not Person IDs, and email is not a safe replacement because it can be shared, missing or changed. And write down your campus rule, because campus is stamped at a different grain in each Planning Center product: People holds it on the person record, Giving stamps it on each donation, Groups on the group. Check-Ins organises by events and locations, and the events themselves carry a campus; the campus-authority problem is real but smaller than it was. Pick one authority for a household’s assigned campus while preserving the campus attached to each historical activity.
The wall
Here’s what the spreadsheet can’t do, in the order you’ll hit it.
The 104-week spine is a history dependency, not a Sunday morning export. Check-Ins headcount reports cap at 53 past sessions and its custom reports are organised by person, so a two-year per-check-in attendance history does not come out of a single Check-Ins export. It comes from the API, or from exports you have been accumulating week by week and combining by hand. Plan the first fill before you promise the sparkline.
It becomes a point-in-time answer after you build it. That is not wrong, but it means every weekly dashboard promise includes a weekly refresh and reconciliation job. Assign that ownership before presenting the workbook as standing infrastructure.
Some metrics need a cohort or snapshot you may not have. Dated Check-Ins, Giving, Services and Groups activity can reconstruct many historical measures. Planning Center Home also preserves metric trends, and its List results widget can graph List totals over a selected timeframe, though the list has to be at least 24 hours old before the widget shows anything and it refreshes on a 24-hour cycle, so the tile can lag the live list. What current-state exports cannot recreate is the exact membership of an arbitrary List or mutable roster on every past date. Six-month retention needs a fixed arrival cohort; historical connected share needs clear dated evidence for every component. Inventory what can be rebuilt from dated records and start snapshots only for the pieces that cannot.
Custom cross-product measures are the expensive part. People Lists can combine conditions from several products to produce a person set, and Home can place their widgets side by side. What those surfaces do not provide is an arbitrary row-level join or a custom formula such as connected share with your exact denominator. That is where stable identifiers, compatible grain and the paste step still land on you.
There is a newer route past part of this wall. Planning Center ships an official AI connector, an MCP endpoint that answers live cross-product questions across People, Services, Groups, Registrations and Check-Ins, scoped to the asking user’s own permissions. For a one-off question, even a cross-product one, it is the fastest answer available and costs nothing to maintain. What it doesn’t give you is Giving, or a copy: it reads live and keeps nothing, so there is no history and nothing to reconcile against next quarter. The cohort memory and the Giving joins this workbook exists for are still the actual gap.
Planning Center’s dashboards are broad but not arbitrary. They live at the account level, Home → Dashboards, and can be personal or shared. The current catalogue spans the suite rather than four metric types, and many widgets support a timeframe, filters and a chart. Year-over-year is firmer ground than it used to be: Planning Center’s May 2026 changelog puts a year-over-year comparison in every metrics chart type. Use the controls available in your account rather than promising a fixed maximum window, or a year-over-year mode on the widgets that aren’t metrics charts.
Hold that against the thirteen. Home can cover several source metrics directly and put them on one cross-product screen. The gaps are the definitions that require custom arithmetic or a preserved cohort: a specific thirteen-week calculation, attendance-frequency buckets, guest return rate, serving concentration, connected share and six-month retention. Build the native widgets first, then decide whether the remaining questions are important enough to earn a maintained workbook.
The version that doesn’t go stale
This is where we build, so treat what follows accordingly.
Parable reads every Planning Center product on a schedule into one place where they share a person key, which removes the paste step and the matching step, and, more usefully, means the history accumulates whether or not anyone remembered to save a snapshot. The cohort metrics have data behind them from the day you connect.
Each tile begins as a question you type. “Weekend attendance by campus, thirteen-week rolling average, last two years.” “What percentage of active adults are in a group, a team, or gave in the last 90 days?” “Of first-time guests in March, how many came back within 30 days?” Validate the answer and definition before pinning it as a tile; the work is in agreeing on thirteen definitions, not merely creating thirteen visuals.
Dashboards can be shared with staff, and with the people who should see a number without needing day-to-day access: the board, a campus pastor, an elder team. And there’s SQL and warehouse access underneath for anyone on staff who’d rather write the query themselves.
Start with four
Whichever way you build it, don’t start with thirteen.
Pick four: weekend attendance as a rolling average, giving households, guest return rate, and serving concentration. Under the table above that is three leading measures and one lagging measure, deliberately weighted toward questions somebody can act on. Run those for a quarter until somebody’s Monday actually changes because of what a tile said.
Then add the milestone tile. It needs a reliable dated source and names behind the count, but no cross-product join when the milestones already live in one form or workflow.
Then add the rest. A small dashboard that named owners use is worth more than a larger one whose definitions and refresh process nobody owns.