Case study: See how combining English and French on one Instagram account crashed our Reels views from 180k to 500, and how Quebec businesses should adapt.

For businesses in Quebec, managing social media often turns into a balancing act: on one side, the massive English-speaking market of North America, and on the other, local legislation and local clients. When we at Million Sales Lab faced the need to create content in two languages, we simply didn't have the resources to build two full-fledged profiles. We made what seemed like a logical decision at the time: merge everything into a single account.
Spoiler alert: this turned out to be the absolute worst decision for free organic promotion.
Let's be blunt: Instagram algorithms hate mixed signals. The Reels recommendation engine promotes a video based on how the very first, earliest group of viewers reacts to it.
When you publish a video in English with French subtitles, the algorithm faces an impossible task. Instead of combining the two audiences (English-speaking + French-speaking), the system starts looking for their intersection. Instagram tries to find those rare individuals who consume content in both languages simultaneously, and who also happen to be interested in your specific niche. As a result, your potential reach shrinks to a statistical margin of error.
What the numbers say: Before switching to the bilingual format, our 3-second Retention Rate held steady at 68%, and we easily reached a pool of 120,000 non-followers. But the moment we added French text over an English video, retention plummeted to 29% within the first 7 days. People simply didn't understand what language was being spoken to them, so they swiped past the video. Ultimately, the overall Completion Rate dropped from 42% to a dismal 11%.
Our mistake and attempt to fix it: At first, we thought the issue was formatting. We tried using Instagram's Auto-Captions and built-in translation stickers. It only got worse: the screen turned into visual clutter, and the audio desynchronized with the text. Viewers swiped away even faster. We gave up, removed the text from the video, and hid the French translation in the post description. Did it help? No. The algorithm had already flagged our content as "confusing" and stopped pushing it into the recommendation feeds.
The full weight of algorithmic punishment is perfectly visible in the numbers for Ton Atelier — a DTF (Direct-to-Film) transfer manufacturer. We were shooting various types of content: professional, entertaining, sometimes humorous. But what worked best was the mesmerizing manufacturing process.
The shift from a purely English strategy to a hybrid format caused median views to plummet from 70,000–180,000 to under 500 views per Reel in just 30 days. We completely dropped out of recommendations for "cold" traffic.
The cost of the experiment: While the profile was run in English, organic reach generated those 180,000 free views and consistently brought in about 14 warm B2B requests for test prints every week. We didn't spend a dime on this. Fast forward 45 days after launching the bilingual format, and views were stuck in the 480–520 range, with the influx of free leads dropping to zero.
The stress test that proved it all: We couldn't believe things were this bad. Maybe we just started making boring videos? To test this, we brought out the "heavy artillery" — we shot a shocking stretch-and-boil stress test on a DTF film. In the past, this kind of hook guaranteed explosive reach.
We uploaded it in the bilingual format and... got the exact same 500 views.
This became our main lesson: the algorithmic trust rating for the profile (Account Quality) had been downgraded. The 500 people watching us were likely just our most loyal fans, regular clients, or relatives. This is a "dead zone" from which it is impossible to extract commercial value.
Any business owner in Quebec knows this acronym: OQLF (Office québécois de la langue française). The requirements of Law 96 regarding the mandatory use of the French language in commercial communications force companies to make compromises. It is this fear of fines that drives business owners to merge two languages into a single profile to save time and resources. But attempting to solve a legal issue within a single account kills your only free customer acquisition channel.
The cost of fear and false economy: Initially, we at Million Sales Lab insisted on creating two separate accounts. However, due to a lack of resources and the client's fear of receiving complaints from the language inspection agency regarding the main profile, the decision was made to merge everything.
The logic was understandable: fines for small businesses under Law 96 can range from 3,000 to 30,000 CAD. It seems like saving about 1,200 CAD a month on hiring a separate Francophone community manager is a great idea. But in practice, this "savings" cost us the loss of organic reach, whose equivalent value in Meta Ads was over 4,500 CAD per month.
What real-world practice shows: When we began to dive deep into the OQLF's enforcement practices, it became clear: the inspection agency strictly monitors official websites, contracts, and customer support. When it comes to social media, you absolutely do not need to break your global English-language profile. It is entirely sufficient to have a separate, localized Francophone account (or sub-account) to fully comply with the requirements of the Charte de la langue française.
Manufacturing process videos are a distinct genre. In the DTF printing niche, clips showing the film peeling process or harsh stress tests perform exceptionally well. This content works on a basic instinctual level: it's shocking, mesmerizing, and relaxing (ASMR). It requires instant visual recognition.
When you duplicate text in two languages, you overload the visual field. The viewer's brain doesn't have time to enjoy the process—it's too busy trying to read a jumble of letters. The visual hook is broken.
How to kill a viral meme: In the purely English version, our hot peel videos easily gathered 2,400 shares for every 180,000 views. It was a universal visual language, understood anywhere in the world. When we added bilingual subtitles and disclaimers, the number of shares crashed to just 6 per 500 views. The video stopped being viral and turned into a boring instructional clip.
The AI failure: Trying to salvage the situation, we decided to experiment with AI dubbing. We tried splitting the audio channels and overlaying voiceovers in both languages. The result? A wave of negative comments and a sharp drop in watch time due to the unnatural, "robotic" sound. Our solution: We completely abandoned these experiments, returning to the original machine sounds and manufacturing ASMR, without the unnecessary text overload.
If your business is located in Quebec but you want to sell across all of North America, forget about hybrid profiles. The only viable model is a separate account architecture: a global profile (in English) for free organic scaling, and a local one (in French) for hyper-local market targeting and legal compliance.
Recovery numbers: As soon as we returned to the separate channel architecture, the English profile began to revive. Within 60 days, average organic reach recovered to 45,000 views per Reel. With the Francophone profile, we took a smarter approach: instead of torturing organic reach, we launched targeted advertising. With a budget of just 500 CAD per month via Meta Ads Manager, we secured a stable Cost Per Lead (CPL) for high-quality B2B leads at 18.50 CAD.
Hidden pitfalls of splitting: When we created the second (Francophone) account, an unexpected problem arose. Instagram's security algorithms almost flagged us as spammers for duplicating content, since the visuals were similar. How we bypassed this: We linked everything through the Meta Business Portfolio for legitimate cross-management. But more importantly, we stopped simply reposting the exact same files. We started adapting visuals (using Canva/Figma templates) specifically for Quebec insights, altering video metadata, and properly configuring analytics (GTM and Conversions API) to clearly track where every dollar comes from.
Content