Nobody has published data like this before, because nobody else has the scripts. Churches submit their written announcements to us every week for production, and since 2017 that has added up to more than 23,000 real announcement scripts. In October 2026 we scored every one of them, full text, against the same 100-point rubric our free audit and the Next Steps Engine use. Nothing here is a vibe. Every number on this page comes from the dataset you can download. The written report, with what the best scripts do differently, is The State of Church Announcements 2026.
How the average church scores
The tight cluster is the headline. The standard deviation is 10 points. Eight in ten scripts fall between 33 and 58. This is not a talent problem spread across thousands of individual churches. It is one default approach, repeated everywhere: lead with logistics, list the events, end with three different places to sign up.
Table: scripts by score band
| Score band | Scripts | Share |
|---|---|---|
| 0-9 | 12 | 0.1% |
| 10-19 | 44 | 0.2% |
| 20-29 | 950 | 4% |
| 30-39 | 5,621 | 23.9% |
| 40-49 | 9,212 | 39.2% |
| 50-59 | 5,892 | 25.1% |
| 60-69 | 1,440 | 6.1% |
| 70-79 | 281 | 1.2% |
| 80-89 | 55 | 0.2% |
| 90-99 | 11 | 0% |
Table: scripts by rubric level
| Level | Score | Scripts | Share |
|---|---|---|---|
| Exceptional | 80-100 | 66 | 0.3% |
| Strong | 70-79 | 281 | 1.2% |
| Effective | 60-69 | 1,440 | 6.1% |
| Growing | 50-59 | 5,892 | 25.1% |
| Foundation | 40-49 | 9,212 | 39.2% |
| Getting started | 0-39 | 6,627 | 28.2% |
A score of 52 beats three quarters of churches. A score of 60 beats 92.4%. A score of 70 beats 98.5%. Levels are on the primary scorer's scale. A stricter second reading of every script at 60 or above confirmed 70 of 1,787; see Method.
The six most common failures
The rubric scores ten things. Churches do three of them reasonably well and six of them badly, in a remarkably consistent way. Ranked by how much of the available score goes unclaimed:
Table: sub-category averages and bands
| Sub-category | Average | Max | Captured | In lowest band | In top band |
|---|---|---|---|---|---|
| Story presence | 3.21 | 15 | 21.4% | 70.6% | 0.4% |
| Emotional transport | 3.46 | 10 | 34.6% | 47.4% | 0.8% |
| Content balance | 5.08 | 15 | 33.9% | 93.8% | 1.1% |
| The hook | 1.91 | 5 | 38.3% | 29.4% | 1.8% |
| Brand, voice, authenticity | 2.87 | 5 | 57.3% | 3.3% | 19.7% |
| Pacing and energy | 2.47 | 5 | 49.4% | 4% | 5.2% |
| CTA singularity | 4.8 | 10 | 48% | 24% | 14.3% |
| CTA clarity | 9.52 | 15 | 63.4% | 26.5% | 26.3% |
| Guest accessibility | 5.48 | 10 | 54.8% | 11.5% | 28.3% |
| Simplicity and focus | 6.66 | 10 | 66.6% | 1% | 59.2% |
Lowest and top bands are the rubric's own: for example story presence 0 to 3 is "no stories, pure logistics" and 12 to 15 is "compelling story tied to the announcement."
Failure 1 · Story presence · 78.6% of points unclaimed
No story, just logistics
70.6% of scripts scored 3 or less out of 15 for story presence: no specific person, moment or outcome a viewer could remember. Only 0.4% told a story the rubric calls compelling.
The fix: One 10-to-20-second testimony per week. A name, a face, what changed.
Failure 2 · Emotional transport · 65.4% of points unclaimed
Nothing to feel
47.4% of scripts scored 3 or less out of 10 for emotional transport. The viewer learns something and feels nothing.
The fix: Lead with what the viewer gains. Replace "don't miss it" with "we would love to see you there."
Failure 3 · Content balance · 66.1% of points unclaimed
Information first, last and only
93.8% of scripts were mostly or purely informational (7 or less out of 15). The target mix is 40% celebration, 40% inspiration, 20% information. Most scripts invert it.
The fix: Report what happened last time before you announce what happens next time.
Failure 4 · The hook · 61.7% of points unclaimed
An opening that earns nothing
The average hook scored 1.91 out of 5. 29.4% opened with "good morning" or straight logistics, and only 1.8% opened with a line that makes you lean in.
The fix: Cut the greeting. Start with the person the first announcement is for.
Failure 5 · Brand, voice, authenticity · 42.7% of points unclaimed
Could be any church
The average brand and voice score was 2.87 out of 5. Just 19.7% of scripts were unmistakably their own church in voice, name and next step.
The fix: Say your church's name, your URL and the way you actually talk, every week.
Failure 6 · CTA singularity · 52% of points unclaimed
Three next steps, so none
24% of scripts had multiple competing calls to action, and only 14.3% pointed everything at one unmistakable next step. The average script carries 3.6 announcements, usually with three different destinations.
The fix: One URL, one text keyword or one table in the lobby. Everything points there.
What churches already do well
- Simplicity and focus: 59.2% of scripts scored 7 or more out of 10. Most churches write clearly.
- Guest accessibility: 28.3% scored 7 or more out of 10. A first-time guest could follow most of what was said.
- CTA clarity: the average was 9.52 out of 15, the highest share of any category. A next step is almost always present. The problem is that it competes with two others.
Put together, the data says the foundation is there. Most churches write clearly, most are understandable to a guest, and almost all of them ask for something. What is missing is not competence. It is story, feeling, and the discipline to ask for one thing.
Length versus engagement: fewer announcements score higher
The average script carries 3.6 announcements. Scripts with two announcements score highest, at 48.6 on average, and every announcement added after that costs about 2 points. By five announcements the average is 42.7, and the chance of an Effective score (60 or more) falls from 12.7% to 3.5%, roughly 4 times less likely.
Table: score by number of announcements
| Announcements | Scripts | Average | Median | Scored 60+ | Story avg (of 15) | One next step avg (of 10) | Simplicity avg (of 10) |
|---|---|---|---|---|---|---|---|
| 1 | 499 | 44.7 | 46 | 13.8% | 3.58 | 5.41 | 6.34 |
| 2 | 2,512 | 48.6 | 49 | 12.7% | 3.74 | 5.49 | 6.81 |
| 3 | 8,624 | 47.1 | 47 | 9.5% | 3.33 | 5.11 | 6.83 |
| 4 | 6,237 | 44.6 | 44 | 6% | 3.06 | 4.58 | 6.61 |
| 5 | 5,458 | 42.7 | 42 | 3.5% | 2.89 | 4.22 | 6.42 |
| 6+ | 178 | 41.7 | 40 | 5.1% | 3.25 | 3.75 | 5.8 |
A further 10 scripts contained no distinct announcement the scorer could identify (average 31.1). They are excluded from this table and included in the overall figures.
Part of this is by design: the rubric rewards a single next step, so a longer list of asks is penalised. But the table shows it is not only that. Story presence, simplicity and content balance all fall as the announcement count rises. More items means less room for the one story that would have made any of them land. The practical ceiling is three announcements per video, and two is better. Everything else goes in the bulletin, the app, or next week.
Words per script
Announcement count is one measure of length. Word count is the other. The average script runs 405 words (median 394), about 120 words per announcement. Half of all scripts fall between 325 and 470 words. Across all 23,518 scripts with a word count, the correlation between length and score is 0.053: almost no relationship, with longer scripts scoring slightly higher. From 100 words up, every length band averages between 44.8 (100-199 words) and 47.5 (600 or more words), a spread of 2.7 points. The only length that reliably fails is almost no script at all: the 48 scripts under 100 words average 24.3.
Table: score by words per script
| Words | Scripts | Share | Average | Median | Scored 60+ | Announcements avg | Story avg (of 15) | One next step avg (of 10) |
|---|---|---|---|---|---|---|---|---|
| Under 100 | 48 | 0.2% | 24.3 | 18 | 2.1% | 1.5 | 1.44 | 3.92 |
| 100‑199 | 403 | 1.7% | 44.8 | 44 | 9.4% | 1.9 | 3.06 | 5.48 |
| 200‑249 | 1,096 | 4.7% | 45.4 | 45 | 7.3% | 2.5 | 3.17 | 5.14 |
| 250‑299 | 2,533 | 10.8% | 45.1 | 45 | 5.8% | 2.9 | 3.07 | 4.96 |
| 300‑349 | 3,764 | 16% | 45.5 | 45 | 6.9% | 3.3 | 3.11 | 5.01 |
| 350‑399 | 4,348 | 18.5% | 45.5 | 45 | 7.5% | 3.6 | 3.2 | 4.87 |
| 400‑449 | 3,973 | 16.9% | 45.6 | 45 | 8.6% | 3.8 | 3.26 | 4.73 |
| 450‑499 | 3,295 | 14% | 44.9 | 44 | 7.1% | 4 | 3.16 | 4.53 |
| 500‑599 | 2,775 | 11.8% | 45.5 | 45 | 7.1% | 4.2 | 3.23 | 4.51 |
| 600+ | 1,283 | 5.5% | 47.5 | 47 | 12.6% | 4.6 | 3.85 | 4.58 |
Word counts are the spoken words of the script as submitted: the introduction, the text of each announcement, and the conclusion. Every scored script carries a word count.
Scripts of 600 or more words score highest, at 47.5 on average, with 12.6% reaching Effective. Scripts of 100-199 words score lowest, at 44.8. Length by itself does not make a script better or worse. What the words are spent on does.
Is it getting better?
For eight years, no. From 2017 to 2024 the yearly average never left the band between 44.2 and 45.9, and story presence sat between 3 and 3.3 of 15 the whole time. Then it moved. Scripts submitted in 2025 average 50.1, and 2026 so far 49.5, with story presence at 4.1 of 15.
It is not a change in who submits. 27 churches sent at least 10 scripts in both 2022-2024 and 2025-2026, through the same intake and scored the same way. Their average went from 45.3 to 49.8, and 23 of the 27 improved. The same churches started writing differently.
Table: score by year
| Year | Scripts | Average | Scored 60+ | Story avg (of 15) | Content balance avg (of 15) | Hook avg (of 5) |
|---|---|---|---|---|---|---|
| 2017 | 5,398 | 44.9 | 7.3% | 3.3 | 5.04 | 1.89 |
| 2018 | 4,942 | 44.8 | 6% | 3.13 | 4.98 | 1.89 |
| 2019 | 4,143 | 45.3 | 6.2% | 3.14 | 5.06 | 1.9 |
| 2020 | 1,880 | 44.4 | 5% | 3.01 | 4.84 | 1.84 |
| 2021 | 1,368 | 44.2 | 4.5% | 3 | 4.75 | 1.83 |
| 2022 | 1,227 | 44.9 | 5.3% | 3.09 | 4.83 | 1.85 |
| 2023 | 738 | 44.6 | 5.8% | 3.04 | 4.95 | 1.82 |
| 2024 | 1,352 | 45.9 | 6.7% | 3.05 | 5.07 | 1.92 |
| 2025 | 1,468 | 50.1 | 19.1% | 3.49 | 5.82 | 2.19 |
| 2026 | 1,000 | 49.5 | 20.9% | 4.09 | 6.12 | 2.24 |
Scripts submitted through the PVA client dashboard since March 2026, where every script is scored and Next Steps Engine subscribers get a rewrite before filming, average 49.2, and 20.5% of them reach 60, against 7.1% of scripts that arrived by email form. Month by month the dashboard average has risen from 44.2 in March 2026 to 54 in August 2026. Scripts that get a score every week are getting better. It is the best argument we have for scoring your own.
Table: client dashboard scripts by month
| Month | Scripts | Average | Scored 60+ |
|---|---|---|---|
| 2026-03 | 122 | 44.2 | 10.7% |
| 2026-04 | 119 | 47.8 | 12.6% |
| 2026-05 | 115 | 46.8 | 18.3% |
| 2026-06 | 135 | 52.7 | 26.7% |
| 2026-07 | 108 | 49.4 | 24.1% |
| 2026-08 | 97 | 54 | 32% |
| 2026-09 | 104 | 50.3 | 21.2% |
The gap between churches
The average hides a wide spread. Among the 252 churches with 20 or more scored scripts, the typical church averages 45.7. The bottom tenth average under 38.3 and the top tenth over 53.4: a gap of 21.9 points between the top and bottom deciles, every week, with the same rubric. 57 churches average 50 or more, six average 60 or more, and the best church in the dataset averages 65.3. Effective is not a ceiling. Some churches live there.
Table: spread of per-church averages
| Per-church average | Value |
|---|---|
| Lowest church | 28.6 |
| Bottom decile (mean of lowest 10%) | 35.6 |
| 10th percentile | 38.3 |
| 25th percentile | 42.3 |
| Median church | 45.7 |
| 75th percentile | 49.6 |
| 90th percentile | 53.4 |
| Top decile (mean of highest 10%) | 57.5 |
| Highest church | 65.3 |
Per-church averages for churches with 20 or more scored scripts. No church is named. The 252 churches account for 22,277 of the 23,518 scripts.
What the best scripts have in common. We had a second, stricter model re-read every script the first scorer put at 60 or above. It agreed on 70 of 1,787. Those 70 are the best announcement scripts in the dataset, confirmed twice, and they look alike: story presence 10.6 of 15 against 3.2 for the average script, content balance 11.5 of 15 against 5.1, emotional transport 7.1 of 10 against 3.5. They carry 3.2 announcements and run about 463 words, so they are not shorter or thinner than the rest. They spend the words on a person and an outcome, then ask for one thing. 52 of the 70 were written in 2025 or 2026. See it written out in our sample scripts and explained in the writing guide.
A 60 beats 92.4% of scripts. It is not a high bar. It is a different default.
Method
- Source
- Every announcement script churches submitted to Pro Video Announcements for production, from 2017 through September 2026: by email order form through February 2026 and through the PVA client dashboard from March 2026. Each script is one weekly video order, typically 3.6 announcements. Churches are primarily in the United States, single-site and multisite, across denominations.
- Sample
- 23,533 scripts went to the scorer after cleaning. Before scoring, submissions that are not announcement scripts (title graphic and sermon bumper orders, mail with no order form), test orders, exact duplicate scripts and fragments under 20 words were removed. Rows whose answer could not be parsed are unscored and excluded from every figure. In all, 453 submissions were removed before scoring and 15 after it; 23,518 scripts carry a score. This is the full population of submissions, not a sample, so there is no sampling margin of error.
- Rubric
- Church Announcements Effectiveness Grading System v2: 100 points across 6 categories and 10 sub-categories. Story and emotional transport (25), next step clarity and singularity (25), simplicity and guest accessibility (20), content balance (15), hook and pacing (10), brand and voice (5). It is the same rubric behind the free audit and the Next Steps Engine.
- Scoring
- Each script scored once by Claude Haiku 4.5 through the Anthropic Batch API: the full spoken script, the complete rubric, and a one-sentence written reason required before each of the ten scores. The total is the sum of the ten sub-scores. No human adjustment to individual scores. Scoring was completed on October 3, 2026. Sub-category averages use the 23,518 scripts whose ten sub-scores were all recorded. An earlier scoring of the same corpus in February 2026 read each script cut at 3,000 characters and without written reasons; it averaged 48.5. This edition replaces it.
- Second reading
- Claude Opus 5.5 re-scored the 1,787 scripts the primary scorer put at 60 or above, same rubric, same prompt, same text, and confirmed 70 at 60 or above (0.3% of all scripts). It is a harsher reader: its average on those scripts was 46 against the primary scorer's 65.7, and it agreed more often the higher the primary score: 0.5% of the 60-64 band, 2.2% of the 65-69 band, 9% of the 70-79 band, 45.5% of the 80+ band. Scripts the primary scorer put below 60 were not re-read, so the second reader's count is a floor on its own scale, not a correction of the primary scale. Every level and share on this page is on the primary scorer's scale; the two readers order scripts and categories alike, which is why the ordering findings carry the page. To measure the two scales against each other across the whole range, the second reader also scored a random sample of 500 scripts: it averaged 36.3 against the primary scorer's 45.4, a gap of 9.1 points, and its average rises from 30.4 in the primary scorer's lowest decile to 43.2 in the highest (rank correlation 0.59). It put one of the 500 at 60 or above against the primary scorer's 34. Both readers rank story presence as the weakest category (21.3% and 15.1% of its points).
- Limits
- Scores measure the written script, not the finished video or what happened in the room. A model scorer is consistent but not infallible, and two models put the same script at different absolute levels; we treat individual scores as estimates, population figures and rankings as reliable, and the share at any one cut-off as the least certain number on the page. Churches that use a production service may write differently from churches that do not, so treat this as the state of church announcement scripts as submitted for video, not a census of every church.
- Privacy
- Aggregates only. No church, campus, person or script text is published.
- Word count
- Words per script are the whitespace-separated words of the spoken script as submitted: the introduction, the text of each announcement, and the conclusion. The order form's own fields are not counted. All 23,518 scored scripts carry a word count. Words per announcement divides by the number of announcement sections in the submitted script.
- Dataset
- state-of-church-announcements-2026.json: every figure on this page, with distributions. Licensed CC BY 4.0. Cite as: Pro Video Announcements, State of Church Announcements 2026, https://provideoannouncements.com/state-of-church-announcements.
- Next edition
- Refreshed yearly at the same URL. The 2027 edition will add scripts submitted from October 2026 onward and extend the year-by-year series.
Where does your church land?
Send last Sunday's announcements to the free audit and get them scored on the same rubric. Your report arrives by email in about an hour during business hours, or by 9 the next business morning.
Score My Announcements Free Start My Two Free Sundays