Case study · Italy Food PRN · May 2024 – Dec 2025
Italy Food PRN: publishing five times more, with the same two hands
A regional food publisher with an audience to build from zero. The constraint was never ideas, it was throughput: how many good videos could leave the door each week. So the job became rebuilding the production line.
Context
Italy Food PRN covers food experiences across Emilia-Romagna for more than 15 local partners, on four platforms at once. Every restaurant visit produces raw footage; the value only appears when that footage is published, in the right format, on every channel, quickly enough to stay current.
When I took the project over, the account was publishing two to five reels a month. The editorial plan was not the bottleneck. Everything between a finished shoot and a scheduled post was: subtitles, cover, caption, hashtags, then repeating it by hand for each platform.
What I did
I rebuilt the publishing step into a fixed sequence and timed it until one reel took seven minutes end to end, which is what makes 20 to 25 pieces a week possible without a team.
- Set the rails once. A Canva project for covers, a fixed subtitle style, a ChatGPT project holding the tone of voice, the recurring hashtags and the hook rules, and every channel connected in the scheduler.
- Subtitles first, on mobile. The raw file goes from Drive into an AI subtitle tool; this single step carries most of the speed.
- Export and keep the transcript. The transcript is the input for the next step, so it never has to be typed.
- Caption from the transcript. The transcript goes into the ChatGPT project, which returns a caption already in the right voice. It always gets reread and humanised, and every correction is fed back as an example.
- Cover and schedule. Five title options from the same transcript, one cover in Canva, then a single upload that schedules Instagram, TikTok, YouTube and Facebook together.
Tools: Google Drive, Captions, ChatGPT, Canva, Publer. Write-up of the full workflow: Publishing 20 reels a week without losing your mind.
What happened to output
Show data table
| Month | Reels |
|---|---|
| 2024-01 | 2 |
| 2024-02 | 3 |
| 2024-03 | 5 |
| 2024-04 | 13 |
| 2024-05 | 17 |
| 2024-06 | 21 |
| 2024-07 | 21 |
| 2024-08 | 15 |
| 2024-09 | 14 |
| 2024-10 | 16 |
| 2024-11 | 18 |
| 2024-12 | 17 |
| 2025-01 | 10 |
| 2025-02 | 12 |
| 2025-03 | 14 |
| 2025-04 | 16 |
| 2025-05 | 18 |
| 2025-06 | 21 |
| 2025-07 | 20 |
| 2025-08 | 14 |
| 2025-09 | 11 |
| 2025-10 | 17 |
| 2025-11 | 18 |
| 2025-12 | 22 |
The three months before the new workflow averaged 3.3 reels a month. The twenty-one months after averaged 16.4, and never dropped below ten, including August, Christmas and every other month that usually kills a publishing calendar. Same person, same shooting days.
The reels that travelled
Across those 20 months the 332 reels collected 31,693,506 views on Instagram alone. The five biggest account for about 19% of that, which is a much flatter distribution than a creator programme: here the reach comes from volume and consistency rather than from a handful of hits.
Four channels, three audiences
Everything published went out on Instagram, TikTok, Facebook and YouTube from the same upload. Three years in, the audiences look like this.
Show data table
| Channel | Followers |
|---|---|
| 92,400 | |
| TikTok | 76,100 |
| 23,343 | |
| YouTube | 2,830 |
Instagram and TikTok carry the project: about 168,000 followers between them, with 1.8 million likes on TikTok alone. Facebook holds a real but smaller audience that skews older and local, which is exactly the one regional partners care about. YouTube received the same 369 videos and never took off: single shorts land in the thousands of views, against tens or hundreds of thousands on Instagram.
Cross-posting from one upload costs almost nothing, so keeping a channel that underperforms is defensible — as long as it stays a by-product. It stops being defensible the moment someone starts cutting, titling and thumbnailing for it. That was the rule: same export everywhere, effort only where the audience already is.
What the audience was for
An audience of this kind pays for itself in two ways: the local partners the project was built to serve, and brand campaigns that rent the format. The clearest example is a Deliveroo shoot in Parma in October 2024, produced end to end: a travel-and-food piece where the talent orders from a local restaurant, receives the delivery in front of the Pilotta, tastes the dish, then goes into the kitchen to watch the tortelli being made and packs them for the next order.
Two things made it work as a piece of client work rather than an ad dropped into an editorial feed. The product appears as part of the story, not as a banner: the order, the rider, the bag, all inside a sequence a viewer would watch anyway. And the format is the same one the account publishes every week, so the sponsored piece does not feel like a foreign object on the profile.
One thing most people get wrong
Teams treat publishing as the boring end of the process and protect the creative start instead. But the publishing step is where the calendar actually dies: when a post takes forty minutes, the third video of the day never goes out. Time that step, take it apart, and the editorial plan you already had starts working.
What didn't work
- Volume does not buy reach on its own. August 2024 published 15 reels and collected fewer views than months with the same output. Monthly totals swung by four times regardless of how much went out, because reach depends on the subject and the hook, not on the count.
- The AI-written caption is a draft, never the final text. Left unedited it reads generic and flattens the account's voice; the rule became that every caption gets rewritten by hand before scheduling.
- No link between views and partner value. Reach was measured, but what those views did for the 15+ local partners, or for a brand campaign like the Deliveroo one, was never tracked beyond the view count. That is the number I would instrument first if I did it again.
Figures read from the public @ifpemilia profile in September 2026: every reel published in the period, with its view count and its publication date.