Unilever Food Solutions runs a global network of professional chefs — chefluencers — who create content on behalf of the brand on Instagram. The programme, built on a "by chefs, for chefs" philosophy, transformed UFS's own in-house chefs into digital voices rather than hiring external influencers. The logic was sound: in the food service industry, chefs trust other chefs far more than they trust brands.
I started this research with a question that the reach numbers alone could not answer: when a chefluencer posts a video and impressions look strong, what is that actually telling you about the brand's position with the people who matter commercially?
UFS is not a consumer brand. It sells to kitchens — hotels, restaurants, catering operations, institutional food service. So a chef with a large following who mostly reaches home cooks and food enthusiasts is telling a very different commercial story than a chef with a smaller but more professionally skewed audience. Follower count does not tell you which one is which, and neither does raw engagement rate.
WHY THIS IS A HARDER PROBLEM THAN IT LOOKS
The UFS chefluencer programme has publicly documented results that are genuinely strong by any B2B standard. A pilot phase produced 98% average follower growth per chef and a 44% uplift in global reach for the UFS chef community, with engagement rates averaging 4.20% across channels — above typical B2B Instagram benchmarks, where food and drink content on the platform tends to sit well below that figure.
But strong pilot numbers create their own analytical challenge. As a programme scales across a large network of chefs and markets, the aggregate metrics that looked impressive at the pilot stage start to mask a lot of variation underneath. Some of that variation is about content quality or posting cadence. Some of it is about audience composition — which is where the more commercially important questions live.
Instagram's algorithm does not distinguish between a food service professional and a home cook who found a chef's page through a trending Reel. From a platform perspective, both are valid audience members. From UFS's commercial perspective, they are not interchangeable.
WHAT THE RESEARCH INVOLVED
Rethinking what the metrics were measuring
Rather than looking at individual chef performance in isolation, the research examined audience composition and engagement quality patterns across the programme. The aim was to identify signals that would say something structural about how the programme was developing — not just who grew and who did not.
That meant cross-referencing follower growth velocity against engagement rate trends over time, flagging cases where reach was scaling while interaction depth was declining, and looking at which content categories drove stronger audience quality signals relative to their volume. It also meant being willing to say where the data was inconclusive. That tends to be an undervalued output in reporting contexts, but it matters: acting on weak signals is its own kind of analytical error.
Translating data into a strategic narrative
A significant part of this work was about presentation — not in a superficial sense, but in the way that analytical findings get shaped for the people who need to act on them. Raw platform data, even from a sophisticated social listening infrastructure, does not arrive pre-organized around the questions a commercial or brand team is actually asking. Structuring analysis around strategic questions rather than metric categories, and making the directionality of findings clear without overstating confidence, is work that sits on top of the data rather than inside it.
The difference between a report that gets filed and a report that shapes planning decisions usually comes down to that layer of translation.
Thinking about measurement at scale
One of the structural challenges of a global programme is that a single benchmarking framework tends to flatten meaningful variation. UFS operates across more than 75 countries. Audience behavior on Instagram in the Netherlands looks structurally different from Southeast Asia or the Middle East — different platform maturity, different chef culture, different levels of institutional trust in social media as a B2B channel. A benchmark that makes sense as a global coordination tool can simultaneously misrepresent performance in a significant number of individual markets.
Good CMI work holds the framework and the variation at the same time, rather than treating one as the exception to the other.
THE TENSION THE DATA DOES NOT RESOLVE ON ITS OWN
My MA thesis at Utrecht University, which ran alongside this research, examined the chefluencer programme from a different angle: how Instagram's algorithmic design shapes the content chefs produce, and what happens to brand alignment when visibility becomes the primary incentive in the relationship.
That framing surfaces a tension worth naming here. A chef whose content strategy is heavily optimized for Instagram's reach algorithm — short-form video, trending audio, high posting frequency — may generate strong impression numbers while gradually drifting away from the brand positioning UFS actually needs. The programme was designed to create authentic chef-to-chef connection. The platform's incentive structure does not always point in the same direction. Reach and relevance are not the same thing, and a social listening tool will surface the first much more readily than the second.
This is not a critique of the programme — it is a structural feature of any large-scale influencer strategy that runs on a platform with its own algorithmic agenda. But it is a question that a CMI practitioner should hold alongside the performance data, because the decisions downstream of the analysis depend on understanding what the metrics are and are not actually capturing.
WHAT I TOOK FROM IT
The core skill this research tested was the translation layer: between what social listening data produces at scale and the strategic language a commercial team uses to make decisions. That translation is not just about making charts readable. It requires a working model of what question the stakeholder is trying to answer, and a willingness to reshape the analysis around that question rather than the other way around.
The other thing it reinforced is that aggregate performance numbers in a global programme are always a starting point, not an answer. The variation underneath the average is usually where the most commercially useful signal lives.
REFERENCES
The Drum (2025). How Unilever Food Solutions' Chefluencer program boosted chef influence. The Drum Awards Case Study. thedrum.com/awards-case-study/how-unilever-food-solutions-chefluencer-program-boosted-chef-influence
Unilever (2024). Menu trends to spark inspiration and sales in professional kitchens. Unilever News. unilever.com/news/news-search/2024/menu-trends-to-spark-inspiration-and-sales-in-professional-kitchens/
Sprout Social (2024). Influencer Marketing Benchmarks Report 2024. sproutsocial.com/insights/influencer-marketing-statistics/
Influencer Marketing Hub (2026). Influencer Marketing Benchmark Report 2026. influencermarketinghub.com/influencer-marketing-benchmark-report/
Meilania, M. (2025). Cook, Capture, Repeat: Performing Digital Labour under Platform Capitalism — Unilever's Chefluencers on Instagram. MA Thesis, Utrecht University.
---
Disclaimer: The analyses in these case studies reflect my personal perspective. All data, campaign results, and figures cited are drawn from publicly available sources, listed in the references above. I have no insider knowledge of these campaigns beyond what has been published.