# Best Time to Post on Facebook: 19:00 UTC (+44.6%) (https://www.socialcrawl.dev/blog/best-time-to-post-on-facebook)
> The best time to post on Facebook is 19:00 UTC: +44.6% vs each Page's own average. 180 posts, 15 Pages, pulled 2026-08-31. Saturday led at +55.4% (median 1.12).
The best time to post on Facebook in this sample is **19:00 UTC**: **+44.6%** vs each Page's own average (mean `norm_er` 1.45, median **1.11**, n=9 posts, 7 Pages). We measured 180 posts across 15 public Facebook Pages on 2026-08-31 via `GET /v1/facebook/profile/posts`. That is 15:00 New York (EDT), 12:00 Los Angeles (PDT), 20:00 London (BST), 21:00 Paris (CEST). Formula: `norm_er = (likes + comments) / page_mean`; 1.00 = that Page's own average. Shares were null on 180/180, so shares are not in the rate.
Facebook was absent from [the 30 Jun 2026 TikTok/Instagram/YouTube/Reddit study](/blog/best-time-to-post-social-media-2026); this post fills that gap. Do not copy that study's 20:00 UTC / +52% onto Facebook.
[Sprout's live Facebook page](https://sproutsocial.com/insights/best-times-to-post-on-facebook/) says Tue/Wed **12–8 p.m. local** (volume, six-network 2B/307k, window closed 27 Feb 2026). [Buffer's live page](https://buffer.com/resources/best-time-to-post-on-facebook/) says **Thursday 9 a.m.** and calls **12–5 p.m. the trough** (14M Facebook posts, no date window). Neither ships rate + date window + per-cell n + a formula anyone can rerun.
## What is the best time to post on Facebook?
**19:00 UTC.** Mean `norm_er` **1.45**, median **1.11**, **+44.6%** vs each Page's own average, n=9 posts from 7 Pages. It is the only hour with n ≥ 8 where both the mean and the median beat 1.0.
Hours below the n ≥ 8 floor stay out of this table. Overnight and early-morning UTC slots are almost empty because these Pages rarely post then — a selection effect, not a "bad hour" finding. Thin buckets live in the limits section.
| Hour (UTC) | Posts | Mean norm_er | Median | vs. Average | Pages |
| --- | --- | --- | --- | --- | --- |
| **19:00** | **9** | **1.45** | **1.11** | **+44.6%** | **7** |
| 22:00 | 13 | 1.30 | 0.89 | +30.4% | 7 |
| 16:00 | 27 | 1.19 | 0.47 | +19.4% | 10 |
| 15:00 | 19 | 1.08 | 0.68 | +8.3% | 10 |
| 18:00 | 16 | 1.06 | 0.85 | +5.9% | 10 |
| 14:00 | 21 | 0.98 | 0.81 | -1.9% | 10 |
| 13:00 | 22 | 0.93 | 0.55 | -6.7% | 12 |
| 17:00 | 18 | 0.84 | 0.58 | -16.3% | 9 |
Grand mean `norm_er` is 1.0000 by construction. Grand median is **0.6364**. Grand mean raw (likes + comments) is 18,925.8. That median-vs-mean gap is the whole reason this table reports both: a handful of viral posts pull the mean up, so a "peak" that only exists in the mean is usually two outliers.
**Do not treat 16:00 UTC as a second peak.** It has the largest n (27) and a +19.4% mean, but the median is **0.47**. Two posts do the damage: Canva at 10.15× on 2026-08-13, MrBeast at 8.75× on 2026-08-23. Ask when to post on Facebook, take the hour with the most posts and a shiny mean, and 16:00 UTC is the false peak that falls out.
The honest runner-up is **22:00 UTC** (+30.4% mean, median 0.89, n=13, 7 Pages). It takes the n ≥ 8 lead only in the thinnest sensitivity cut (2026 posts and drop the capture day), where 19:00 falls below n=8. Treat the headline as **19:00 UTC**, with **22:00 UTC** as the runner-up — not as a 14-million-post law.
19:00 UTC holds as the top eligible hour in **4 of 5** sensitivity cuts (full table in the method section). n=9 is above the floor we set before the pull, not a power calculation. 180 posts is not Buffer's 14 million.
Every timestamp is UTC; the local clocks for the headline hour are 3pm New York / 12pm Los Angeles / 8pm London / 9pm Paris. What time to post on Facebook for a Page whose audience lives in one city is still that city's clock, not a blended, "universally applicable" one.
19:00 UTC is the only eligible hour with mean and median above the Page's
own average. Converted: 3pm New York / 12pm LA / 8pm London / 9pm Paris.
16:00 UTC's +19.4% mean is two outliers (median 0.47). 22:00 UTC is the
runner-up.
## What is the best day to post on Facebook?
**Saturday.** Mean `norm_er` **1.55**, median **1.12**, **+55.4%**, n=13 posts from 8 Pages. Saturday holds as the top eligible day in **every** sensitivity cut. The best days to post on Facebook in this sample are Saturday then Sunday — Sunday is the only other day with both mean and median above 1.0 (+32.9% / median 1.01, n=20).
All 180 posts, bucketed by UTC date:
| Day | Posts | Mean norm_er | Median | vs. Average |
| --- | --- | --- | --- | --- |
| Monday | 52 | 0.86 | 0.43 | -13.7% |
| Tuesday | 24 | 0.57 | 0.51 | -42.9% |
| Wednesday | 29 | 1.05 | 0.70 | +5.0% |
| Thursday | 19 | 1.15 | 0.68 | +15.5% |
| Friday | 23 | 0.97 | 0.88 | -3.3% |
| **Saturday** | **13** | **1.55** | **1.12** | **+55.4%** |
| Sunday | 20 | 1.33 | 1.01 | +32.9% |
**Do not headline Monday as bad.** Capture day was Monday 2026-08-31. 30 of 52 Monday posts published that day (mean `norm_er` **0.59**); the other 22 Mondays sit at **1.23**. Drop 2026-08-31 and Monday flips to **+18%**. The −13.7% row is recency bias: likes and comments on same-day posts have not finished arriving.
The worst eligible day on the raw table is **Tuesday −42.9%** (n=24, median 0.51). That one is not a capture-day artifact.
Weekday × hour is almost empty. Only two cells hit n ≥ 8, and both sit below average: Monday 17:00 UTC at −77.0% (n=8, capture-day contaminated) and Wednesday 13:00 UTC at −9.0% (n=8). There is no honest "best hour on Saturday" or "best hour on Monday" in this grid. Saturday as a *day* is usable. A Saturday hour is not.
### What time works on Tuesday?
Tuesday is the weakest eligible day (−42.9%, n=24, median 0.51). Capture day was Monday, so this is not recency. No Tuesday × hour cell reached n ≥ 8. Treat Tuesday as weak in this sample; do not invent a Tuesday clock time.
### What time works on Thursday?
Thursday's mean is +15.5% (n=19) but the median is 0.68 — mean above 1, median not. No Thursday hour hit n ≥ 8. [Buffer's live page](https://buffer.com/resources/best-time-to-post-on-facebook/) currently names Thursday 9 a.m. as the single best slot; this sample cannot confirm or deny that hour.
### What time works on Friday?
Friday scored −3.3% (n=23, median 0.88). Buffer's 26 January 2024 thread called Friday the best Facebook day, with engagement from [9 a.m. through 4 p.m.](https://x.com/buffer/status/1750881757996499285). This 2026-08-31 sample does not. No Friday hour reached n ≥ 8.
## When is the worst time to post on Facebook?
Among hours with enough posts to score, the worst time to post on Facebook is **17:00 UTC: −16.3%** (mean 0.84, median 0.58, n=18, 9 Pages). Next-worst eligible: 13:00 UTC −6.7% (n=22), then 14:00 UTC −1.9% (n=21). Converted from the 19:00 clocks, 17:00 UTC is 13:00 New York / 10:00 Los Angeles / 18:00 London / 19:00 Paris.
Empty overnight buckets are not "worst." Hours with n < 8 are selection effects: these Pages almost never publish then. Monday is not the worst day; Tuesday is.
The vendor pages that rank for this query cannot agree, and they are not measuring the same thing:
| Source | Headline window | Metric | Sample / window | Timezone |
| --- | --- | --- | --- | --- |
| This study | Worst eligible hour **17:00 UTC (−16.3%)**; afternoons are mixed (16:00 mean is outliers) | per-page `norm_er` (likes+comments) | 180 Facebook posts / 15 Pages / 2026-08-31 | UTC, conversions given |
| [Sprout](https://sproutsocial.com/insights/best-times-to-post-on-facebook/) | Tue/Wed **12–8 p.m. local** is the *best* block; weekends worst; FAQ worst **4–7 a.m.** | engagement **volume** | ~2B engagements / 307k profiles **across 6 networks**; Facebook-only N never broken out; 27 Nov 2025–27 Feb 2026 | "local time" |
| [Buffer](https://buffer.com/resources/best-time-to-post-on-facebook/) | Afternoons **12–5 p.m. are the worst** on most days; best slot Thu 9 a.m.; best day Wednesday | "engagement rates" (formula unstated) | 14M Facebook posts sent via Buffer; **no date window** | "universally applicable" black box |
| [Hootsuite](https://x.com/hootsuite/status/1899207634755990011) | "early morning" local | unspecified | Facebook-specific claim via Mar 2025 X replies | local TZ, not EST |
Sprout's winning afternoon block is Buffer's trough. That is not a rounding error. Sprout counts raw engagement **volume** across six networks and never breaks out Facebook. Buffer talks about "engagement rates" without publishing the formula, using posts **sent via Buffer**, with timezone handled by what the page itself calls "mathematical magic." Hootsuite's Facebook claim is local early morning, not EST. Averaging those clocks produces a number nobody measured.
[Ignite Visibility](https://ignitevisibility.com/best-times-post-facebook/) is the only ranker with a "why studies disagree" heading, and that heading never compares the live studies. The table above is the comparison. Do not import TikTok's 24-hour distribution window onto Facebook Feed; that quote is TikTok-only.
## What should a Facebook posting schedule look like?
Here is a Facebook posting schedule you can actually run — from this sample only. Facebook posting times below are the 19:00 UTC conversion plus the day table — not a per-weekday hour grid, because that grid does not exist at n ≥ 8.
- **Post:** 19:00 UTC (15:00 EDT / 12:00 PDT / 20:00 BST / 21:00 CEST). Runner-up 22:00 UTC.
- **Day to prefer:** Saturday; Sunday second (both mean and median > 1).
- **Day to treat as weak in this sample:** Tuesday (−42.9%).
- **Hour to avoid among eligible cells:** 17:00 UTC (13:00 EDT / 10:00 PDT / 18:00 BST / 19:00 CEST).
- **Do not:** skip Monday on the strength of the −13.7% row; treat 16:00 UTC as a peak; claim a Wednesday or Monday hour.
## When is the best time to post on Facebook for business?
**19:00 UTC** — this sample is 15 public brand Pages (Nike, Starbucks, NASA, Netflix, Shopify, Buffer, Hootsuite, and the rest of the list in the method), not personal profiles. The windows describe when *these business Pages* over-performed their own average. Mega-page bias is real: MrBeast's page mean raw is 187,670 vs Buffer's 4. Per-page normalization stops follower count from drowning the hour; it does not create a local-SMB sample.
LinkedIn Company Pages vs personal profiles behave differently enough to need [the LinkedIn posting-time analysis](/blog/best-time-to-post-on-linkedin). This Facebook pull did not measure cadence, so there is no 1–2×/day vs 3–5×/week claim here.
## How did we measure this — and can you reproduce it?
**Question.** What is the best hour (UTC) and weekday to post on Facebook, measured as per-page-normalized engagement?
**Formula** (written before the pull):
1. `raw_engagement = likes + comments + shares` (null = 0 only if at least one field is present; all-three-null dropped).
2. `page_mean = mean(raw_engagement)` per Page.
3. `norm_er = raw_engagement / page_mean`.
4. Bucket by `published_at` → hour UTC and weekday.
5. Headline = highest mean `norm_er` among buckets with **n ≥ 8**; median reported beside the mean.
That is the same per-account normalization used in the [2026 engagement-rate benchmarks](/blog/social-media-engagement-rate-benchmarks-2026). Shares were null on **180/180** kept posts, so raw engagement is likes + comments on every row. Do not quote this as a shares-inclusive rate.
**Reels.** `/v1/facebook/profile/reels` was not the primary metric (rounded views, null likes/comments/shares). Reels that appeared on the Page feed via `/profile/posts` were kept when they carried likes/comments: **65 reels / 115 non-reel**. Mean `norm_er` reels = 0.78; non-reel = 1.13. That is a composition note, not a timing split. No Reels-only hour.
**Sample.** 15 primary Pages × 4 cursor pages × 3 posts/page = 180. Facebook serves this feed in slices of 3 (`pagination.page_size = 3` on every successful call). Every primary returned 12 kept posts. No backups. Failed Pages: none. We did not sample personal profiles.
Pages: MrBeast, Nike, National Geographic, BBC News, NASA, Netflix, Red Bull, Buffer, Hootsuite, Canva, Duolingo, Starbucks, Spotify, Adobe, Shopify. Pulled 2026-08-31.
**Endpoint.** [`GET /v1/facebook/profile/posts`](/docs/facebook) on host `https://www.socialcrawl.dev`, header `x-api-key`. Credit check: `GET /v1/credits/balance`.
```bash
curl "https://www.socialcrawl.dev/v1/facebook/profile/posts?url=https://www.facebook.com/nike" \
-H "x-api-key: YOUR_API_KEY"
# next slice
curl "https://www.socialcrawl.dev/v1/facebook/profile/posts?url=https://www.facebook.com/nike&cursor=NEXT_CURSOR" \
-H "x-api-key: YOUR_API_KEY"
```
**Spend.** 60 credits (1 probe + 59 harvest; 1 of 60 harvest calls was a cache hit — MrBeast page 1). Harvest-window balance 153,237 → 153,178. Endpoint cost 1 credit/call; cache hits billed 0.
Any developer with a SocialCrawl key reruns the same 4-slice walk, takes `published_at` + `engagement.likes|comments|shares` from the unified envelope, drops all-null engagement, divides by Page mean, and buckets by UTC hour/weekday. 180 posts / 15 Pages cost **60 credits** on 2026-08-31. The SocialCrawl API is the instrument; `norm_er` is ours.
Sensitivity (same 180 posts; page means recomputed inside each subset — not a second harvest):
| Cut | Posts | Pages | Best eligible hour | Best eligible day |
| --- | --- | --- | --- | --- |
| All (headline) | 180 | 15 | 19:00 UTC, +44.6% (n=9, median 1.11) | Saturday +55.4% (n=13) |
| 2026 only | 130 | 11 | 19:00 UTC, +54.8% (n=8, median 1.59) | Saturday +84.1% (n=9) |
| Drop capture day | 150 | 14 | 19:00 UTC, +39.8% (n=8, median 1.20) | Saturday +24.0% (n=13) |
| 2026 and drop capture day | 100 | 10 | 22:00 UTC, +29.3% (n=8, median 0.82) | Saturday +38.6% (n=9) |
| Drop Buffer + Hootsuite | 156 | 13 | 19:00 UTC, +39.9% (n=8, median 1.07) | Saturday +55.4% (n=13) |
## Where these numbers hit their limits
1. **Capture-day recency (Monday 2026-08-31).** 30 of 52 Monday posts are same-day; mean `norm_er` 0.59 vs 1.23 for the other 22. The −13.7% Monday row is not a rule. Drop today and Monday flips to +18%.
2. **All timestamps are UTC.** The study cannot split timezone effects. 19:00 UTC = 15:00 EDT / 12:00 PDT / 20:00 BST / 21:00 CEST.
3. **Facebook serves 3 posts per page.** This sample is 4 slices × 15 Pages = 180, not a randomized year of content.
4. **The feed is not a clean "latest 12."** Year mix: 2026=130, 2025=11, 2024=12, 2021=10, 2017=10, plus a handful of 2018–2023. Netflix's entire 12-post window is 2017–2021-era; Shopify's 12 are all pre-2025; Spotify contributes 9 pre-2025. A 2017 Netflix post has had years to accumulate likes. The 2026-only cut still puts 19:00 UTC on top.
5. **Mega-page bias + already-optimized schedulers.** Per-page `norm_er` stops raw follower count drowning the hour. It does not create a mid-tier local-business sample. MrBeast's page mean raw is 187,670 vs Buffer's 4.
6. **Selection bias.** We measured when these Pages happen to post. Empty overnight hours are not proven bad hours.
7. **Shares null 180/180.** The rate is likes + comments.
8. **Feed mix includes Reels (65/180).** No format-split headline. `/profile/reels` was not used as a primary list.
9. **Thin buckets.** Hours with n < 8 are not findings: 00:00 (n=1), 01:00 (n=2), 07:00 (n=1), 09:00 (n=3), 10:00 (n=5), 11:00 (n=2), 12:00 (n=5), 20:00 (n=5), 21:00 (n=6), 23:00 (n=5). 19:00 itself is n=9. n ≥ 8 is a floor, not a power calculation. This is a dated 180-post snapshot, not Buffer's 14 million or Sprout's 2 billion. The advantage is live-API reproducibility.
10. **No per-weekday best hour.** Sibling-keyword demand exists for Wednesday, Saturday, Monday, and the rest. The grid does not support it.
## Frequently asked questions
### What is the best time to post on Facebook on Monday?
Capture day was Monday 2026-08-31. 30 of 52 Monday posts went live that day (mean `norm_er` 0.59) vs 1.23 for the other 22. The raw Monday row is −13.7%; drop the capture day and Monday is **+18%**. The only Monday × hour cell at n ≥ 8 is Monday 17:00 UTC (−77.0%, n=8), also capture-day contaminated. There is no Monday hour finding. Do not treat "avoid Monday" as a rule.
### What is the best time to post on Facebook on Saturday?
Saturday is the headline **day**: +55.4% mean, median 1.12, n=13, 8 Pages. It holds in every sensitivity cut. No Saturday × hour cell reached n ≥ 8, so there is no honest Saturday *hour*. Prefer Saturday as a day; do not invent 10 a.m.
### What is the best time to post on Facebook on Wednesday?
Wednesday's mean is +5.0% (n=29) but the median is 0.70. The only Wednesday × hour cell at n ≥ 8 is Wednesday 13:00 UTC (−9.0%, n=8) — below average, not a peak. Buffer's live page calls Wednesday the best *day*; this sample does not. No Wednesday hour finding.
### What is the best time to post on Facebook on Sunday?
Sunday scored +32.9% mean / median 1.01 / n=20 — the only other day besides Saturday with both mean and median above 1.0. Sprout's Sunday H3 ("worst day") is volume on a six-network sample; this rate on Facebook-only Pages disagrees. No Sunday hour at n ≥ 8.
### What is the best time to post on Facebook today?
There is no vendor "today" clock. If "today" is Saturday, the day finding applies; if Tuesday, this sample is weak. The hour to try is still 19:00 UTC in the four converted local clocks (15:00 EDT / 12:00 PDT / 20:00 BST / 21:00 CEST). Recency: posts published the day you measure will look worse because likes and comments have not finished arriving.
### What is the best time to post on Facebook for likes?
The study metric is `norm_er` on **likes + comments**, not likes-only. Shares were null on 180/180 so they never enter the rate. We did not rank hours by likes alone. Do not treat 19:00 UTC as a likes-max window; it is a per-page normalized likes+comments window.
Reproduce the pull in the [Explorer](/explorer) against your own Page list and see your data before writing a single line. Endpoint docs for `GET /v1/facebook/profile/posts` live at [/docs/facebook](/docs/facebook).