One pipeline, two languages, 727,000 views

The hypothesis

Does a short-form educational format transfer between languages, or does each audience need its own? I built a pipeline to find out: one concept goes in, and two localised videos come out, an English one for CleverFacts and a Polish one for Ogarnik. Over ten days in March 2026 it published 74 videos on each channel. Six months later I exported the full analytics history of both and analysed it. This page reports what the data supports, and says where it stops.

What I built

• An automated production pipeline: the videos were generated through APIs, not edited by hand, so one concept becomes a vertical video with a voice.
• A localisation stage: each concept is adapted for the Polish audience, then dubbed expressively, so the delivery survives the change of language.
• A measurement stage: an export of each channel’s full history through the YouTube APIs, an analysis script that reduces it to aggregates, and the charts on this page, drawn from that file at build time.

I did not build or train the generation models. My work is the pipeline around them, the two channels it produced and the analysis that measures them.

Technical stack

Decisions that mattered

Adapt, not translate literally

Each Polish version is an adaptation of the concept for the audience it is going to, with its own title and phrasing.

Matched pairs, measured per video

Each concept yields an English and a Polish video, and the export holds a daily series for every one, so the channels can be compared concept by concept.

Publish aggregates only

The raw exports stay private. The analysis script writes medians, shares and per-concept view counts to one file, and the page reads nothing else.

Evidence, from the analytics export

YouTube Analytics for both channels, 11 March 2026 to 19 September 2026, exported on 19 September 2026. Views count plays, not people, and these figures measure audience reach.

727,455

views across 148 videosCleverFacts 486,941 and Ogarnik 240,514, 74 videos each; about 3,608 hours of watch time.

4,318 and 1,302

median views per video, English and PolishRanges 207 to 24,644 and 37 to 49,750.

0.46

rank correlation across 73 matched pairsEnglish views against Polish views for the same concept. It falls to 0.34 when the first day's batch is left out.

2.4 and 5.4

subscribers gained per 1,000 views1,186 gained on CleverFacts and 1,288 on Ogarnik.

The public counters on the channel pages count differently; this page uses the export throughout so that every number can be recomputed.

What the analytics show

The analysis covers the 148 short-form videos published between 11 March 2026 and 20 March 2026. The CleverFacts channel dates from 2017 and carries 3 unrelated uploads, excluded here; they account for 71 views.

Nearly all of the views arrived in the first three weeks: 99% in English and 97% in Polish by 31 March.

CleverFacts, EnglishOgarnik, Polish

Cumulative views as a table

Date	CleverFacts	Ogarnik
11 March 2026	18	0
14 March 2026	43,441	5,088
17 March 2026	215,831	55,942
20 March 2026	399,906	173,119
23 March 2026	457,660	217,509
26 March 2026	470,858	228,220
29 March 2026	477,521	232,787
1 April 2026	480,662	234,732
4 April 2026	482,932	236,165
7 April 2026	483,891	236,674
10 April 2026	484,320	236,924
13 April 2026	484,482	237,119
22 April 2026	484,919	237,457
13 May 2026	485,321	238,111
3 June 2026	485,551	238,613
24 June 2026	485,756	239,107
15 July 2026	486,210	239,562
5 August 2026	486,422	239,987
26 August 2026	486,683	240,225
16 September 2026	486,948	240,517
19 September 2026	486,948	240,517

Both channels peaked within a day of each other, CleverFacts on 19 March 2026 with 77,066 views and Ogarnik on 20 March 2026 with 66,431. Publishing stopped on 20 March and the feed moved on within days, leaving a tail of about a dozen views a day by September.

A short is over within a week: the median video had 95% of its lifetime views by day seven in English and 90% in Polish.

CleverFacts, EnglishOgarnik, Polish

Launch curve as a table

Day	CleverFacts median	Middle half	Ogarnik median	Middle half
0	36.4%	23% to 53%	28.7%	8% to 49%
1	65.4%	46% to 74%	54.8%	37% to 76%
2	76.6%	63% to 87%	72.5%	54% to 83%
3	85.1%	74% to 91%	79.5%	65% to 88%
5	91.6%	85% to 95%	86.0%	76% to 93%
7	95.0%	92% to 97%	89.9%	85% to 96%
10	97.1%	95% to 99%	93.8%	89% to 98%
14	98.5%	97% to 99%	95.6%	91% to 98%
21	99.3%	98% to 100%	97.3%	93% to 99%
28	99.5%	99% to 100%	97.5%	94% to 99%
56	99.8%	99% to 100%	98.4%	96% to 99%
120	99.9%	100% to 100%	99.8%	99% to 100%

The same shape holds video by video: for most, the busiest day was the first or the second. A short that has not found an audience within a week is very unlikely to find one later.

The Shorts feed delivered 77% of English views and 88% of Polish ones; subscribers and search barely registered.

CleverFacts, EnglishOgarnik, Polish

Traffic sources as a table

Category	CleverFacts views	Share	Ogarnik views	Share
Shorts feed	375,994	77%	210,914	88%
Subscriptions and notifications	56,652	12%	9,613	4.0%
Channel page	32,421	6.7%	8,313	3.5%
YouTube search	15,759	3.2%	9,096	3.8%
Everything else	6,190	1.3%	2,580	1.1%

Subscribers, search and the channel page supplied under a quarter of English views and about a tenth of Polish ones. A view count here is mostly a record of how far the feed chose to carry a video.

What worked in English was only a loose guide to what worked in Polish: a rank correlation of 0.46 across 73 matched pairs.

All 73 pairs as a table

Concept	English views	Polish views
House things that can hurt a cat	8,265	49,750
House things dangerous for dogs	6,171	35,211
Signs never to ignore in a dog	751	3,178
Vegetables that grow back from scraps	2,097	8,474
Never leave these foods out	1,783	4,661
Animals more dangerous than they look	397	986
Never pick these mushrooms	2,866	5,673
Never store these foods together	2,172	4,018
Never dry these on a radiator	989	1,635
Bad conditions for walking a dog	20,423	31,992
Never flush these	5,974	9,065
Never store these under the sink	5,389	8,026
Never refreeze these foods	868	1,116
House smells that mean trouble	4,368	4,981
Removing stains	1,595	1,526
Foods dangerous for cats	2,371	2,258
Fruit mistakes, part 2	1,623	1,241
Your car is dirtier than a toilet seat	4,856	3,656
Things that attract bugs	2,567	1,782
How not to kill houseplants	14,513	8,687
Plants for bees	1,556	884
Foods that feed gut bacteria	3,549	1,525
Fruits with the most pesticides	4,712	1,990
Never mix these together	3,242	1,328
Tap water	1,138	446
Homemade drinks	4,172	1,505
Laundry	788	268
Kitchen items that hold bacteria	3,991	1,278
Types of tea	3,861	1,232
Foods with more fibre than you think	4,464	1,418
Never store these in the bathroom	4,513	1,404
Cleaning tricks that work	18,717	5,646
Things that start mould	4,752	1,326
Foods never to reheat twice	597	163
Foods good for your gut	12,925	3,128
Plants that repel pests	787	190
Plants for butterflies	554	129
Sea animals	207	47
Vase flowers	10,420	1,934
Never clean these with bleach	664	116
Things that expire fast	8,547	1,422
Kitchen things to replace more often	12,046	1,973
Plants never to grow together	8,193	1,237
Things that attract mice	13,432	1,920
Never put these in the dishwasher	674	96
Never put these in the microwave	10,303	1,454
Garden mistakes that attract slugs	8,906	1,252
Foods dangerous for dogs	24,644	3,166
Foods you wash wrong	14,384	1,643
Cleaning products that need gloves	16,428	1,750
Things that ruin air quality	1,291	135
Herbs that belong in the same pot	17,093	1,754
Houseplant mistakes	1,238	119
Never put these in the fridge	15,114	1,424
Never pour these down the sink	19,944	1,812
Skincare mistakes	1,681	112
Never put these in the air fryer	7,623	500
Plants that belong together, part 2	2,252	147
Garden products dangerous for pets	5,694	332
Vehicles and pollution	1,397	77
Plants that belong together	21,254	1,169
Plants that can kill a dog	6,736	362
Dangers in a baby's room	6,721	321
Cat mistakes	1,465	69
Mould you may not notice	2,026	95
Your dog wants this	8,271	363
Leftovers that can make you sick	19,922	648
Fruit mistakes	4,268	127
Your bed is dirtier than you think	10,999	306
Your cleaning tools are filthy	17,755	449
Jungle cats	1,695	37
Plants dangerous for pets	13,363	281
Woodland creatures	1,864	37

This is the question the experiment was built for. English won 59 of the 73 pairs, and the median pair had 4.3 times more views in English, which mostly reflects the size of the two audiences. The more useful reading is relative: of the concepts that beat their own channel’s median, 23 did so in both languages, 13 only in English and 13 only in Polish. 21 pairs differ by ten times or more.

One caveat: the nine Polish videos published on 12 March 2026 had a median of 69 views, against 1,465 in English the same day, and they lift the correlation; without them it is 0.34 across 65 pairs.

The Polish channel lived on a few hits: 4 videos made half its views, where the English channel needed 14.

CleverFacts, EnglishOgarnik, Polish

Concentration of views as a table

Top videos	CleverFacts share	Ogarnik share
1	5.1%	20.7%
3	13.6%	48.6%
5	21.8%	56.0%
10	39.3%	69.6%
20	64.2%	82.1%
37	85.8%	93.2%
74	100.0%	100.0%

The Polish channel’s three pet-safety videos account for 49% of its views and 73% of the subscribers YouTube attributes to a video, which is why Ogarnik converts views to subscribers at about twice the English rate.

English videos were watched further through: the median video was watched for 80% of its length, against 67% in Polish.

Average percentage viewed as a table

Average percentage viewed	CleverFacts	Ogarnik
Median video	79.9%	67.1%
Lower quartile	74.9%	62.2%
Upper quartile	85.6%	77.1%
Lowest and highest	52.8% to 183.1%	48.2% to 154.3%
Weighted by views	81.9%	76.9%
Median view duration	31.5 s	26.5 s

Average percentage viewed, the only retention measure in this export, was higher in English on the median video; weighted by views the gap narrows (81.9% against 76.9%), because the Polish hits held attention well. Within a channel its link to views is weak.

Vertical video is not only a phone format: about one view in six was on a television, in both languages.

CleverFacts, EnglishOgarnik, Polish

Devices as a table

Category	CleverFacts views	Share	Ogarnik views	Share
Phone	290,947	61%	170,680	71%
Television	81,672	17%	38,132	16%
Tablet	80,528	17%	18,997	8.0%
Computer	27,295	5.7%	11,157	4.7%

Who watched, and where

Age and gender are YouTube’s estimates for signed-in viewers and cover the whole channel.

Limitations and status

Nothing new has gone out since March 2026. The analysis is observational: the two markets differ in size, the platform’s distribution decides most of what any single video receives, and a concept was published once per language, so no pair is a controlled comparison. A correlation of 0.46 across 73 hand-matched pairs, with an interval of roughly 0.25 to 0.63, is evidence of a loose relationship and nothing finer.

Views count plays, not people. Per-video retention curves are not in this export, the demographic reports are channel-level, and the API has no hour-of-day dimension, so nothing here speaks to publishing time. The production code is not public, and this page describes that side of the pipeline by layer and interface only.

What I learned

The channels

Both channels are public: CleverFacts on YouTube (opens in a new tab) and Ogarnik on YouTube (opens in a new tab). Thumbnails here are self-hosted and the player loads only when clicked.