The Room Changes the Story
Roxanne publishes across three platforms. The data shows that they are not three versions of the same audience — and that her most valuable asset may not be virality at all.
She is nervous. So she becomes Pikachu.
In one of the last posts captured in this study, Roxanne looks directly into the camera and describes a familiar moment: an interview, a client presentation, a deadline, the sudden acceleration of the heart, the tightening of the shoulders. Her solution is deliberately absurd. When pressure arrives, she imagines herself as Pikachu charging electricity before a challenge.
The post is useful here for a reason that has little to do with its early metrics. It reveals the mechanism that appears repeatedly across nine months of content. Roxanne encounters something difficult or unfamiliar, finds a way to interpret it, and returns with an explanation that another person can use.
We should not treat the physiological explanation in the video as a finding of this report; it is Roxanne's own framing. What matters analytically is the storytelling structure: pressure becomes metaphor, metaphor becomes a tool, and private experience becomes transferable knowledge.
That same structure appears in Canada, in international work, in business English, in immigration, in pregnancy and in cultural difference. The subjects change. The function keeps returning.
The biggest audience is not where the story moves
At first glance, YouTube should be the center of Roxanne's system. It has 27,900 subscribers, compared with 15,545 followers on Instagram and 8,629 on Facebook. But follower counts describe stored audience potential. They do not describe what happened between January 1 and September 20.
Audience size
Follower/subscriber counts reported in the supplied platform export.
During the period, the dataset records 113 publications and 80,654 interactions. Instagram generated 76,159 of them. Facebook generated 4,104. YouTube generated 391.
of all recorded interactions came from Instagram. It produced this share with 58 of the 113 posts and without holding Roxanne's largest nominal audience.
This is the first major finding: Instagram is not merely another channel in Roxanne's portfolio. During the observed period, it is the active attention engine.
Publishing was relatively balanced. Response was not.
Instagram received only moderately more publishing activity than Facebook, yet it captured almost nineteen times Facebook's total interaction. That matters because it rules out the simplest explanation: this is not merely a story of “she posts more there.”
Where interaction accumulated
The asymmetry is primarily in response. Facebook is receiving Roxanne. Instagram is amplifying her.
Hold the creator still. Hold the message still. Change the room.
Cross-platform comparisons are usually messy because creators publish different things in different places. Here, the data gives us a cleaner test. We identified 25 Facebook–Instagram pairs with the exact same normalized text posted on the same day. This is not a randomized experiment, but it controls for far more than a generic platform comparison.
Median Instagram-to-Facebook interaction ratio across the 25 exact-text, same-day pairs.
That number is more strategically useful than a single viral post. It tells us that the Instagram advantage recurs even when the message itself is held constant.
| Example | What it shows | ||
|---|---|---|---|
| Corporate English #27 | 6,133 | 37 | Same idea, radically different response |
| Canadian workplace culture shock | 5,902 | 277 | Experience + interpretation travels strongly on IG |
| Corporate English #18 | 405 | 2 | Series recurrence does not transfer equally |
| Corporate English #25 | 337 | 11 | Useful professional language remains IG-native |
Cross-posting makes content available. It does not make content native. The strategic problem on Facebook is therefore not “more Roxanne.” It is discovering what an idea must become to behave like Facebook content.
Then one ice-cream post bent the entire year
On February 6, a playful Instagram post referencing Ben & Jerry's and Canadian life generated 39,822 interactions. No other post came close.
interactions from one Instagram post. That single publication generated 52.3% of all Instagram interaction recorded in the period.
This is where a dashboard can tell the truth and still mislead. Instagram's mean is 1,313 interactions per post. Its median is 307.5. The distance between those two figures is the fingerprint of an extreme distribution.
Average
Useful for total output, but heavily pulled upward by the February event.
Median
A better description of the middle of ordinary Instagram performance.
Roxanne therefore has at least two kinds of performance: a baseline and an event layer. Any future reporting that shows only the average will overstate what a typical post does.
February looks unbeatable — until the anomaly is removed
Instagram interactions by month
*September through the 20th only.
February's 41,746 interactions look like a dominant month. But 39,822 came from the Ben & Jerry's post. Remove that event and February falls to 1,924 interactions. May, March and even the first twenty days of September suddenly become more strategically interesting.
Five posts carried almost three quarters of Instagram
The five highest-performing Instagram publications generated approximately 74.4% of all Instagram interaction. Yet they do not share a single topic.
| Date | Editorial territory | Interactions |
|---|---|---|
| Feb 6 | Canada / humor / Ben & Jerry's | 39,822 |
| Sep 1 | Corporate English #27 / changing work vocabulary | 6,133 |
| May 23 | Canadian workplace culture shock | 5,902 |
| Mar 14 | Pregnancy / life transition | 2,646 |
| Mar 1 | Canada / urban legend / humor | 2,122 |
Humor works. Utility works. Cultural interpretation works. Personal stakes work. The common denominator is not topic. The stronger hypothesis is that high-performing posts give the audience something to recognize, use, discover or feel.
A recurring series may matter more than a viral miracle
Across the dataset, one editorial product keeps returning: #30秒真實外商英文 — 30 Seconds of Real Corporate English. The series is not simply a hashtag. It creates an expectation: another episode will come, and it will translate something Roxanne has actually encountered inside an international workplace.
On Facebook, 23 identified series posts generated 1,006 interactions, with a median of 12. The same franchise exists on both platforms. It does not have the same audience behavior.
The strategic value of the series is recurrence. A 39,822-interaction event is extraordinary. A product that can be published every week, recognized by name and connected to an email call-to-action is infrastructure.
Episode 27 may tell us more than the year's biggest post
On September 1, episode 27 connected new workplace vocabulary to the after-effects of the pandemic, AI, changing work habits and the anxiety of keeping up. It generated 6,133 Instagram interactions and 37 on Facebook.
The content is important because it does more than teach vocabulary. It locates language inside a changing workplace. Roxanne's recurring introduction reinforces the point: she shares business English she has actually learned while working in Canada's technology industry.
She is not simply saying, “I teach English.” Her stronger proposition is: “I encountered this language while doing the work.”
That is experiential authority. It is harder to copy than a vocabulary list and more defensible than generic instruction.
Canada is the setting. Adaptation may be the story.
The dominant hashtags look like a category map: #加拿大 (Canada) appears 94 times; #海外工作 (overseas work) 87; #英文 (English) 57; #商用英文 (business English) 51; #海外生活 (overseas life) 48; #加拿大生活 (life in Canada) 24.
But frequency is not the same as function. Canada supplies context and difference. Work supplies recurring problems. English supplies a practical tool. Roxanne supplies continuity.
Her May 23 workplace-culture post is a good example. She recounts entering a Canadian workplace and realizing that she had carried habits from previous work environments into the new one. The post is personal, cultural and professional at the same time. Instagram responded with 5,902 interactions.
Across the year, Roxanne increasingly moves among all three positions.
Some posts are consumed. Others are kept.
On May 17, Roxanne reconstructed a first-month-in-Canada checklist: bank account, Social Insurance Number, phone service and other practical steps. It generated 1,389 interactions on Instagram and 285 on Facebook — unusually strong for Facebook relative to its median.
This kind of post has a different editorial job. It is structured to be useful after the first viewing. The supplied export does not give us reliable save counts, so we cannot claim that people saved it at a particular rate. But it gives us a clear hypothesis to test: structured utility content may have longer shelf life than ordinary feed content.
The account looks like video. The aggregate data points toward structure.
The source export's format summary reports a substantially higher aggregate interaction rate for carousel content than for video/Reels/Shorts: approximately 5.76% versus 0.09%.
Aggregate interaction rate by format
Different content types and viral outliers may confound the comparison. This is an association, not proof that carousel format causes performance.
The correct response is not “stop video.” The stronger reading is that structured, reference-oriented content deserves to be treated as a primary format rather than an accessory. The next analysis should separate format from topic, platform and content job.
Distribution without equivalent response
Facebook generated 4,104 interactions across 50 posts, an average of 82 and a median of 24.5. Yet its strongest posts show that the platform is not inert.
The introduction of the Corporate English series produced 607 interactions. A long post about surviving an all-English work environment produced 393. The pregnancy announcement produced 359. A post about private health insurance as an employment benefit produced 345. Roxanne's first Canadian full-time-job story produced 318.
There is a pattern worth testing: Facebook appears more responsive when the premise is immediately legible as a story, a consequential life event or a practical resource.
The largest audience is still an unfinished question
Only five YouTube publications appear in the period. Together they generated 391 interactions. One video — asking whether they regretted choosing Canada for immigration and sharing their real situation in 2026 — generated 289, or roughly 74% of YouTube interaction in the dataset.
Five observations are not enough to diagnose a 27,900-subscriber channel. The defensible conclusion is narrower: YouTube is Roxanne's largest nominal audience and her least-tested publishing environment during this period.
The current data shows low publishing activity, not the underlying responsiveness of the subscriber base under a sustained YouTube-native strategy.
The person does not dilute the expert. She may authenticate her.
On March 14, Roxanne announced her pregnancy in a reflective post about a life that repeatedly refused to follow the original plan: leaving Taiwan, being sent to Shanghai, entering technology despite disliking computers, moving to Canada despite the cold, and now preparing for a child.
The post generated 2,646 interactions on Instagram and 359 on Facebook. It is one of the strongest pieces on both platforms.
The surface topic is personal. The deeper narrative is familiar: Roxanne enters a life she did not fully plan, adapts and finds language for what happened. That is the same narrative mechanism operating in her professional content.
Attention without needing an enemy
The sentiment export classifies only part of the publishing universe, so its counts must not be treated as a complete census. Among the posts it does classify, positive and neutral dominate and only one is labeled negative. Facebook's classified comment data also contains substantially more positive than negative sentiment.
The useful conclusion is not that “everyone likes Roxanne.” The files cannot support that claim. The useful conclusion is that the strongest observed attention does not require a recurring conflict engine.
The February viral post is playful. The May workplace post is reflective. The September breakout is useful. The pregnancy post is intimate. The urban-legend post is humorous.
Roxanne's attention system currently appears more compatible with recognition, usefulness, surprise and personal stakes than with sustained outrage.
Three Roxannes, one system
The person
Canada, family, humor, uncertainty, everyday life.
Job: connection.
The translator
Cultural differences, workplace behavior, immigration reality, adaptation.
Job: interpretation.
The useful expert
Business English, checklists, interviews, job search, benefits, workplace phrases.
Job: repeat utility.
The owned relationship
The recurring series already invites users to leave an email to receive practical business-English notes.
Job: move beyond rented attention.
This gives the account a potential funnel without forcing every post to sell something. Personality can attract. Interpretation can establish relevance. Utility can create return behavior. Owned contact can preserve the relationship beyond the algorithm.
The next question is not “how much engagement?”
Platform analytics are good at telling a creator what happened inside each platform. They are less good at explaining what the three platforms mean together. The next measurement system should classify every post by editorial job and then follow the chain beyond reaction counts.
Reach, views, non-follower exposure.
Interactions, comments, shares, saves.
Profile visits, link taps, repeat viewers.
Follows/subscriptions attributable to content windows.
Email sign-ups and subsequent opens/clicks where available.
A viral joke and a business-English carousel can both be successful while doing completely different jobs. The next intelligence layer should judge each against the job it was designed to do.
The missing follower curve matters
The historical follower export does not provide a usable daily movement series. That prevents two potentially decisive analyses: whether the February viral event produced durable follower growth, and whether the September Corporate English breakout converted attention into relevant audience.
Those are not the same outcome. A post can create tens of thousands of interactions and almost no durable audience. A smaller post can recruit highly relevant followers or email subscribers.
We cannot infer reach, unique people, revenue, conversion, algorithmic causation or follower acquisition from the supplied interaction totals. Instagram comment sentiment is also not sufficiently complete for a full emotional map.
Turn publishing into an experiment
Do not simply publish more. Instrument the next period.
| Question | Test | Measure |
|---|---|---|
| Is Facebook weak, or merely under-adapted? | Native FB versions vs. straight cross-posts | Median interaction, comments, shares, profile actions |
| Can YouTube reactivate? | Consistent YouTube-native publishing around 2–3 proven territories | Views, returning viewers, subscriber conversion |
| Is the series building an owned audience? | Unique CTA/link for Corporate English | Email sign-ups per episode |
| Do carousels win because of format or utility? | Same topic expressed as carousel and short video | Saves, shares, completion, interactions |
| What survives virality? | Track 7/14/30-day follower and profile movement after peaks | Durable audience gain |
Do not organize by platform. Organize by job.
Discover
Humor, Canadian observations, cultural surprise, recognizable moments.
Purpose: bring new people in.
Connect
Personal life, uncertainty, family, transitions.
Purpose: create identification.
Explain
Canada, workplace culture, immigration reality, professional differences.
Purpose: become the interpreter.
Use
Business English, checklists, interviews, benefits, job search.
Purpose: become useful enough to return to.
Then add a fifth layer outside the feed: Own. Newsletter, notes or another opt-in resource can turn temporary platform attention into a durable relationship.
Not every important number is large
The first number proves that extraordinary attention is possible. The second keeps the analysis honest. The third points toward something repeatable.
If Roxanne builds only around 39,822, she is chasing an event. If she understands how to raise 307.5 and convert the recurring 198 into an owned audience, she is building a system.
The room changes the story. The protagonist does not.
Roxanne entered this analysis looking like a creator with three social channels. The data shows something more asymmetric. YouTube holds the largest nominal audience. Facebook carries a large share of the publishing load. Instagram produces almost all of the observed response.
But the platform story is only the surface.
Underneath it, a more durable editorial identity is taking shape. Roxanne repeatedly enters unfamiliar systems — a country, a workplace, a phrase, a cultural rule, a life transition — and comes back with an explanation. Sometimes it is funny. Sometimes practical. Sometimes intimate. Increasingly, it is useful.
The February 6 post proved she can capture extraordinary attention. The recurring Corporate English series suggests she may also be building authority. The May workplace-culture post shows that the strongest bridge may be the space between lived experience and useful interpretation.
That changes the strategic question.
It is no longer simply: Which platform is performing best?
It becomes: How does Roxanne turn attention into a repeatable editorial product — and a repeatable editorial product into a relationship she owns?
On September 20, she told her audience to imagine becoming Pikachu when pressure arrives: the same physical signal, interpreted differently.
The metaphor accidentally fits the account itself.
The energy was already there. The next step is deciding what to do with it.
How this reading was built
The analysis covers January 1 through September 20, 2026 and uses the four supplied Fanpage Karma exports: general metrics, content, sentiment and follower history. The observable content universe contains 113 publications across Facebook, Instagram and YouTube.
Interactions are treated as the supplied combined measure of reactions/likes, comments and shares. Means and medians are shown together where outliers materially affect interpretation. The cross-platform natural comparison uses 25 exact normalized-text Facebook–Instagram pairs posted on the same day; broader near-match comparisons were not used for the headline ratio in order to preserve a stricter matching rule.
Sentiment is interpreted only for the subset classified by the source export. The follower-history series is not usable for daily gain/loss analysis. Format relationships are associations and are not presented as causal effects. The September 20 Pikachu post was captured on the final day of the period and therefore is used as qualitative evidence of voice and editorial mechanism, not as mature performance evidence.
Traditional Chinese source material is translated or summarized into English where analytically useful. Translations are functional rather than certified literary translations.
Evidence before certainty. Contradiction matters. Recurrence is a signal, not proof. Correlation is not causation. Limits are part of the answer.