Telegram RegisterThe public register of Telegram

Open data · reference

Telegram engagement-rate baselines by channel size

How many views a Telegram channel’s posts actually get, relative to its subscriber count, across every size band — measured on public channels, not surveyed, not self-reported. Nobody publishes this in English, so we do.

ERR is average views per post over a rolling window, divided by subscribers, times 100. It is TGStat’s definition and the one used on every channel page here, so a figure taken from this table is comparable with a figure taken from anywhere else that uses it.

Edition
2026-08-12
recomputed every 6 hours
Cohorts
83
7 size · 66 size × language · 10 decile
Channels
281,733
eligible under the gates below
Window
30d
rolling, ending at compute time

Download the whole edition. CSV · JSON — every column on this page plus p05, p95, p99, mean, min and max, which are in the files and not in the tables below. The JSON carries the definitions and caveats in its meta block, so the file stays interpretable away from this page. Free to use with attribution to tgregister.com.

Read this before reading the table

ERR declines with size, structurally. In this edition the median falls from 18.3% in the 1,000–3,162 band to 2.81% in the 1,000,000+ band — a factor of 6.5, across channels doing nothing differently except being larger. ERR falls as a channel grows, and it falls structurally: a larger audience is a more passive one, and the denominator grows faster than the numerator. Comparing a small channel’s ERR against a large one’s therefore measures audience size, not quality — which is the whole reason this dataset is cohorted, and the whole reason a single global “good ERR” figure would be worthless.

A low ERR has many innocent causes, including audiences that read in the Telegram app without opening the channel, and channels whose subscriber growth long predates their current output. These baselines describe a distribution. They do not label any channel.

By subscriber band

Fixed half-decade bands — each spans a factor of 3.16, from 1,000 to 1,000,000+. Fixed rather than population-derived so that every bucket covers the same multiple and so that successive editions stay comparable with each other.

Subscriber band — ERR percentiles, in percent
CohortChannelsSubscribersMedian subsp01p10p25p50p75p90p99Median viewsMedian posts
1000-3162109,1201,000–3,1621,7130.246%3.97%8.85%18.3%34.3%56.9%154.9%31417
3162-1000087,9343,163–9,9995,4700.1%2.49%5.98%12.8%24.9%42.8%109.1%70121
10000-3162355,14010,000–31,61915,8400.082%1.60%4.19%9.77%20.1%36.0%90.1%1,58028
31623-10000021,43531,623–99,98148,6350.07%0.976%2.98%7.43%16.3%29.5%70.2%3,76034
100000-3162286,562100,004–316,057143,8680.066%0.73%2.20%6.07%14.3%27.4%64.9%9,13038
316228-10000001,134316,245–999,955440,7920.069%0.563%1.80%4.63%12.6%26.1%60.4%22,00043
1000000+4081,000,322–15,640,0391,451,0740.028%0.172%0.956%2.81%8.20%16.9%49.3%47,500165

Cohort is the band’s nominal edges; Subscribers is the range its members actually occupy. Median viewsis the median of the members’ own mean views per post, shown to the precision Telegram published it at. p05, p95, p99, mean, min and max are in the CSV and JSON.

By subscriber band and language

The same bands split by the modal language of the channel’s posts, published only where the cell holds at least 300 channels and that language accounts for at least 60% of the channel’s window. Language is not a footnote here — the spread between languages inside a single band is wider than the spread between two adjacent bands.

Subscriber band × language — ERR percentiles, in percent
CohortLanguageChannelsSubscribersMedian subsp10p25p50p75p90Median viewsMedian posts
1000-3162Arabic7,6061,000–3,1621,7282.59%5.46%11.6%22.8%40.7%20320
1000-3162Belarusian7581,000–3,1571,7889.16%19.8%36.3%60.9%93.1%62119
1000-3162Bengali4401,003–3,1301,7322.11%4.08%9.56%19.2%37.5%15019
1000-3162German8101,001–3,1581,6748.07%13.7%25.4%46.5%73.1%45020
1000-3162English8,6941,000–3,1621,7311.62%5.02%12.7%27.5%51.4%22919
1000-3162Persian15,8131,000–3,1621,7273.91%8.33%17.2%32.3%53.6%29420
1000-3162French3811,002–3,1561,7084.52%10.1%18.9%34.2%54.3%32419
1000-3162Hindi4641,001–3,1491,8664.61%9.73%19.5%36.9%62.6%36320
1000-3162Indonesian8481,001–3,1581,7421.87%3.77%8.03%15.3%29.9%14120
1000-3162Italian8061,001–3,1551,7145.81%12.0%21.6%35.2%53.7%36220
1000-3162Korean3861,004–3,1271,8005.46%9.52%19.4%37.5%63.0%34920
1000-3162Portuguese7771,010–3,1591,7921.90%4.48%9.39%17.6%32.1%16920
1000-3162Russian54,1641,000–3,1621,6945.33%10.5%20.1%35.6%57.1%34315
1000-3162Spanish1,1091,000–3,1581,7363.84%7.89%16.0%30.8%47.5%27820
1000-3162Turkish1,1451,001–3,1621,7445.29%10.3%24.2%46.7%78.6%42219
1000-3162Ukrainian3,1711,001–3,1621,76210.9%18.4%29.6%47.0%70.4%52717
1000-3162Chinese2,1761,001–3,1601,7061.22%5.51%16.6%38.7%69.4%29217
3162-10000Arabic6,5983,163–9,9995,4031.67%4.00%9.03%17.5%31.4%48822
3162-10000Belarusian7383,173–9,9895,6915.05%12.4%25.0%45.1%68.4%1,36025
3162-10000Bengali3763,163–9,9875,7971.16%3.48%7.32%13.8%27.4%42922
3162-10000German6003,164–9,9895,2966.60%13.0%21.4%35.9%55.9%1,20023
3162-10000English8,3363,163–9,9995,5941.18%3.55%9.27%20.0%36.6%51823
3162-10000Persian12,1553,163–9,9995,3622.57%5.73%12.3%23.7%40.5%66524
3162-10000French3893,165–9,9605,6861.77%5.53%12.2%23.0%37.1%66821
3162-10000Hindi7453,175–9,9695,8673.60%7.14%14.3%25.3%41.1%83222
3162-10000Indonesian8683,169–9,9875,6921.05%2.53%5.60%11.4%20.1%30424
3162-10000Italian5743,172–9,8675,4264.25%9.29%17.2%27.1%42.1%90724
3162-10000Korean4243,166–9,9925,5324.58%8.81%16.4%30.4%55.0%89926
3162-10000Portuguese7293,174–9,9995,5131.52%3.20%6.57%11.9%21.4%36926
3162-10000Russian40,4833,163–9,9995,4643.46%7.15%14.2%26.5%43.8%77120
3162-10000Spanish1,0363,165–9,9985,5183.32%6.67%12.2%21.9%37.3%67324
3162-10000Turkish1,2243,163–9,9935,6863.27%7.36%16.4%33.2%55.2%91423
3162-10000Ukrainian2,6783,163–9,9995,3807.34%13.7%23.2%38.9%59.6%1,28022
3162-10000Vietnamese3343,169–9,9775,8021.10%2.72%7.41%16.4%28.0%41223
3162-10000Chinese2,3553,164–9,9905,7850.77%2.67%8.16%19.2%38.6%46620
10000-31623Arabic4,32410,000–31,61916,0001.01%2.98%7.20%14.5%25.3%1,16028
10000-31623Belarusian57110,001–31,53815,8854.28%8.36%17.4%32.1%63.1%2,91040
10000-31623Bengali30610,038–31,37416,1720.988%2.52%5.74%11.3%21.3%90632
10000-31623German33410,013–31,30915,6116.18%10.2%18.2%29.7%47.5%2,85034
10000-31623English6,34410,002–31,60916,0620.726%2.73%7.33%15.7%28.3%1,22031
10000-31623Persian7,49010,001–31,61115,8121.70%3.83%8.86%17.6%29.9%1,42033
10000-31623French30710,013–31,37216,8271.48%3.65%8.63%16.0%27.9%1,47028
10000-31623Hindi77910,021–31,60017,0271.66%4.63%10.00%19.3%34.8%1,67029
10000-31623Indonesian63010,001–31,61117,0140.491%1.72%4.69%9.01%15.8%74232
10000-31623Italian35910,005–31,58615,9482.20%4.94%11.6%21.1%31.6%1,90037
10000-31623Portuguese51910,049–31,53316,4280.828%2.16%4.85%9.84%19.2%77338
10000-31623Russian23,17110,000–31,61615,7192.54%5.51%11.7%23.5%42.3%1,88024
10000-31623Spanish66510,024–31,44515,4402.43%5.19%11.1%18.6%32.3%1,71031
10000-31623Turkish79110,015–31,52315,9891.67%4.34%10.2%20.8%38.0%1,68034
10000-31623Ukrainian1,75410,014–31,61515,3366.39%11.2%19.4%31.2%49.0%3,06032
10000-31623Vietnamese33110,002–31,56716,7230.763%1.47%3.17%7.19%13.9%53131
10000-31623Chinese1,87110,003–31,61115,6510.566%1.93%4.82%11.5%24.9%80725
31623-100000Arabic1,84131,637–99,84349,0310.504%1.83%5.31%12.5%22.9%2,68035
31623-100000Belarusian31531,643–99,77753,8242.88%6.90%13.7%26.2%43.1%8,09053
31623-100000English2,90631,629–99,98148,2340.437%1.91%5.54%12.2%23.6%2,74036
31623-100000Persian3,06731,627–99,86949,4011.01%2.41%5.96%12.2%22.9%2,97044
31623-100000Hindi41231,629–99,56949,0201.02%3.19%7.07%14.6%24.9%3,58032
31623-100000Russian7,90431,623–99,96148,3502.39%4.93%10.4%21.0%37.1%5,19030
31623-100000Turkish33131,861–99,78350,3591.85%4.94%9.94%19.6%32.6%5,30039
31623-100000Ukrainian55831,629–99,32547,3245.61%9.83%16.3%26.4%42.2%8,09051
31623-100000Chinese81031,647–99,92646,7400.305%1.14%3.03%7.37%14.8%1,55030
100000-316228Arabic525100,020–314,917146,6680.622%1.66%4.02%11.3%21.7%6,11034
100000-316228English1,008100,025–316,023141,1680.27%1.23%4.15%9.59%18.0%5,99036
100000-316228Persian1,072100,131–315,252148,3880.929%2.11%4.72%9.87%18.6%7,24051
100000-316228Russian2,317100,026–316,057145,3422.03%4.37%10.3%21.5%35.4%15,40034
316228-1000000Russian475316,877–999,955450,8221.31%2.76%7.36%19.0%37.7%35,60040

Cohort is the band’s nominal edges; Subscribers is the range its members actually occupy. Median viewsis the median of the members’ own mean views per post, shown to the precision Telegram published it at. p05, p95, p99, mean, min and max are in the CSV and JSON.

By population decile

Population deciles of the same eligible set. DESCRIPTIVE ONLY — published because it shows how the corpus is distributed, and because the width of the top decile is precisely why the size cohorts above are fixed bands and not deciles.

Population decile — ERR percentiles, in percent
CohortChannelsSubscribersMedian subsp10p25p50p75p90Median viewsMedian posts
Decile 128,1741,000–1,3111,1424.32%10.0%20.6%38.3%63.0%23617
Decile 228,1741,311–1,7441,5124.40%9.33%18.8%35.3%58.5%28317
Decile 328,1741,744–2,3612,0273.81%8.45%17.8%33.2%54.6%35918
Decile 428,1732,361–3,3012,7833.41%7.76%15.9%29.9%49.3%44518
Decile 528,1733,301–4,7303,9732.91%6.85%14.3%27.2%46.1%56620
Decile 628,1734,730–6,7155,6062.42%5.87%12.6%24.7%42.7%70924
Decile 728,1736,715–10,0258,1522.14%5.20%11.4%22.5%38.9%93026
Decile 828,17310,025–16,08712,3971.68%4.44%10.2%20.5%35.6%1,27027
Decile 928,17316,087–33,23121,9671.50%3.91%9.27%19.5%36.2%2,06029
Decile 1028,17333,231–15,640,03962,5950.86%2.65%6.92%15.5%28.6%4,89036

Cohort is the band’s nominal edges; Subscribers is the range its members actually occupy. Median viewsis the median of the members’ own mean views per post, shown to the precision Telegram published it at. p05, p95, p99, mean, min and max are in the CSV and JSON.

Definitions

ERRMean views per post over the window ÷ subscribers × 100. TGStat’s definition, and the same one this site computes on every channel page, so the figures are comparable rather than a house metric.
ViewsThe latest reading we hold for each post, taken from t.me. Telegram renders any counter at or above 1,000 to three significant figures, so those readings are rounded; the per-channel evidence carries the relative uncertainty that propagates from it, and a channel is only eligible when that uncertainty is under 2%.
SubscribersThe latest channel_snapshot reading for the channel — the exact figure from the profile page, never the rounded counter on the /s/ preview header.
err_pNNThe NNth percentile of ERR within the cohort, in percent.
subs_min / subs_maxThe subscriber range the cohort’s members actually occupy, which is narrower than the nominal band label.
avg_views_p50Median of the cohort members’ own mean views per post.
posts_p50Median number of posts published in the window.

Eligibility

A channel is counted in a cohort only when it clears every gate below. They exist so that a percentile is computed over channels we have genuinely measured rather than over channels we have glanced at — which is why the population above is much smaller than the register.

TypeBroadcast channels only, not deleted
Subscribersat least 1,000
Posts in windowat least 8
Posts carrying a view readingat least 8
Readings taken 24h+ after postingat least 8
Share of window posts with a readingat least 50%
Relative uncertainty on ERRat most 2%

What this dataset is not

It is not a ranking, and no cohort here is a target. A channel below its cohort’s tenth percentile is not thereby doing anything wrong — the list of innocent explanations is long and starts with an audience that reads inside the Telegram app without ever opening the channel. It is also not a census: it describes the 281,733 channels in this register that clear the gates above, not Telegram.

Views above 1,000 reach us rounded to three significant figures, because that is how Telegram renders them. That rounding is propagated into a relative uncertainty per channel, and a channel whose uncertainty exceeds 2% is excluded rather than included with a caveat. The percentiles are therefore computed on readings whose error is bounded and stated, not on readings assumed exact. Full methodology.

Citing this.Telegram channel ERR (views-to-subscribers) cohort baselines, edition 2026-08-12, tgregister.com”. Editions are dated and the files are regenerated in place, so cite the edition date — the numbers move as the register grows.