How HelloFresh Turned Weekly Consumer Conversations into a Global Intelligence Engine

“We’ve moved beyond even the 24-hour research cycle to 24 seconds.”

— Jo Lindenberg, Director of Global Consumer Insights, HelloFresh

From the Succeet Frankfurt presentation, The New Blueprint for Global Empathy: Challenging the Research Status Quo with AI and Virtual Personas

Watch the full session here.

About HelloFresh

HelloFresh is the world’s leading meal kit company, delivering nearly a billion meals annually across 18 countries. With a mission to change the way people eat forever, HelloFresh operates at the intersection of food, technology, and consumer behavior where understanding what customers actually want, in their own kitchens and in their own words, is not a quarterly initiative. It’s a business imperative.

The challenge: scaling the “why” without losing the human

HelloFresh’s consumer insights team tracks brand health and consumer sentiment across global markets continuously. The data tells them what is shifting subscription behavior, meal preferences, purchase patterns. What it can’t tell them is why. That answer has always required qual.

The problem is that traditional qualitative research and the pace of a global CPG company make for an uneasy partnership. Jo Lindenberg, Director of Global Consumer Insights, put it plainly: her team had one chance to get qual right per cycle one study, a narrow window, limited reach and by the time findings landed in a report, the business had often already moved on.

This is the trap that project-based research sets for every insights team operating at scale. The study closes, the deck gets shared, and the knowledge quietly retires. The next time a stakeholder asks a similar question and they will the team either digs through folders hoping to find the old work, or commissions a new study from scratch. Either way, the organization pays twice for the same understanding. Research that should compound instead resets.

Two specific pressures made this untenable for HelloFresh:

1. Scaling the “why” fast enough to matter. When tracking data shows a trend shifting in multiple markets simultaneously, waiting weeks for exploratory qual to catch up means decisions get made on instinct. Jo’s team needed to close that loop not occasionally, but constantly. Every week, she told us, they see something shifting. Every week, they need to know why.
2. Getting stakeholders close enough to act. Insights teams know the feeling: a finding lands in a deck, receives polite acknowledgment, and quietly fails to change anything. The gap between knowing and doing is rarely about the quality of the insight. It’s about distance. Stakeholders who haven’t heard a customer’s voice, seen their face, or watched them navigate a product in their own kitchen are working with abstraction. Jo needed to close that distance at scale and fast enough to be useful.

And then there was ethnography. In-home ethnography has always been the gold standard for consumer closeness watching how customers actually live, cook, and make decisions in their own environment. It is also, as Jo described, “really costly, really hard to set up,” limited to a handful of respondents across two or three countries, and dependent on video crews and months of scheduling. The insight value is undeniable. The operational weight is immense.

Something had to change.

The approach: AI moderation where it earns its place, human research where it matters most

The move from human-moderated ethnography to AI moderation wasn’t made lightly. Jo describes herself as a long-standing skeptic of AI “I think there still is a lot of faff around it” and her team approached the pilot with genuine uncertainty about whether the quality would hold.

It did. A member of Jo’s team who ran the UK pilot was, in her words, “consistently blown away by the quality of the AI moderation.” Stakeholders watching the sessions had the same reaction they couldn’t believe what they were seeing. Consumers were engaged, detailed, and candid in ways that surprised everyone.

Two things explain it. First, the cultural moment: COVID normalized screens, and ChatGPT normalized talking to AI. The behavioral barrier that would have made AI-moderated research feel strange five years ago has largely dissolved. Second, what Jo called the “stranger on the bus” phenomenon without a human moderator to perform for, respondents simply open up. “There’s not a person that I have to prove something to or be shiny for,” she said. “It’s just a bot. I’m happy to show it my messy kitchen.”

What makes the quality hold at scale isn’t just the format. Discuss AI Agents are purpose-built for research trained to probe, follow threads, and adapt based on what a respondent says, in the respondent’s own language, on their own schedule. That’s a different capability than general-purpose AI repurposed for interviews, and it’s what lets HelloFresh run genuine 45-minute in-depth conversations rather than shallow surveys dressed up as something more.

HelloFresh now runs in-home ethnography every single week. The scale comparison with where the program started traditional in-person ethnography tells the story directly:

Human Ethnography	Before Discuss	With Discuss today
Format	Human-moderated, in-person	AI-moderated, self-paced
Cadence	Occasional, project-based	Every week
Respondents per study	6	18
Countries/languages	3	10
Internal observers per session	2–4	30–40

Respondents move through sessions at their own pace answering questions from their car, pausing mid-conversation, picking back up in their kitchen. The resulting footage is genuinely UGC-style: the customer’s point of view, their environment, their words. As Jo put it, “They’re the ones holding the camera.” That shift alone changes the depth of what you learn.

Instant, one-click translation means the team writes discussion guides once in English and deploys them across all ten markets without friction. When a stakeholder asks a new question mid-cycle, Jo’s team updates the guide, the translation updates automatically, and the answer arrives in the following week’s session. The loop from question to answered is no longer measured in months.

Jo is equally deliberate about where human moderation still belongs. Shop alongs, digital user journeys, message and concept testing where a skilled moderator’s judgment in the moment changes the quality of what gets uncovered those stay human. The distinction isn’t philosophical. It’s practical. Discuss makes both possible on the same platform, which means the choice stays where it belongs: with the researcher.

The results: from data points to stories at scale

Every Friday, Jo’s team pulls the week’s sessions down to a 20-minute highlight reel and plays it at the start of a cross-functional meeting. Thirty to forty people from across the HelloFresh business watch it together. Then they discuss it.

“Sometimes our stakeholders don’t want to leave the meeting,” Jo said. “They’re still talking, asking questions, thinking about what they can do with this.”

That’s a different relationship to consumer insight than a quarterly readout produces. The empathy is direct. The stories are fresh. And because the cadence is weekly, the organization’s understanding of its customers compounds rather than resets.

Alongside the weekly sessions, Discuss’s AI-powered analysis tools sit across all of this accumulated data summarizing themes, surfacing relevant quotes, and letting anyone interrogate the dataset with their own questions without commissioning a new project.

Jo describes what this changes for her team: the analysis becomes interactive rather than static. A stakeholder who wants the data cut a different way doesn’t create a new workload. They ask a new question, the analysis shifts, and they can click through to watch a specific customer make the point in their own words. That combination of mass evidence and individual human voice is what turns insight into action.

The outcomes Jo’s team points to:

Stories, not just data points. Every team in the business now has their own customer stories to carry and tell not filtered through an insights report, but drawn directly from what they watched and heard. “In years past, there was that one customer interview that everyone kind of passed around,” Jo said. “Now we’re flooding people with stories.”

Continuous consumer closeness. Rather than relying on occasional deep-dive projects, the team now has a standing weekly window into how customers across ten countries are actually living and thinking. The question isn’t whether they’ll hear from customers this week. They will.

Stakeholder empathy at scale. When teams across the business hear directly from customers see their faces, hear their hesitations, watch them navigate a kitchen the distance between insight and decision collapses. Jo’s team doesn’t just deliver findings. They deliver the conditions for other people to care.

Time and cost savings. Replacing expensive in-person ethnographic crews with AI-moderated, self-paced research at global scale while dramatically increasing the frequency and reach of that research represents a fundamental reallocation of research budget toward output rather than overhead.

What’s next: research that never stops learning

HelloFresh’s ambition doesn’t stop at weekly ethnography. Jo’s team is building toward something more consequential: a continuously compounding intelligence layer where every study makes the next one smarter and where the knowledge never retires.

Using Discuss Virtual Personas, the team is exploring AI-powered versions of their core customer segments built from real interview transcripts, quantitative survey data, and behavioral research, not synthetic guesses or open-internet training data. The distinction matters. A persona grounded in HelloFresh’s own years of consumer conversations will tell you something true about HelloFresh’s customers. A persona built from generic AI training data will tell you something plausible about consumers in general. Those are not the same thing, and for a brand making product and campaign decisions, the gap between them is the gap between confidence and risk.

With Virtual Personas, a brand manager with a question about messaging doesn’t wait for the next study cycle. They chat with a persona built from actual customer conversations and get directional signals in seconds. Jo described where this takes them: “We’ve moved beyond the 24-hour research turnaround. Now I’m chatting with the data exactly when I want, with exactly the questions I have.”

That’s what it means for research to never sleep. Not just faster projects, but a living system one where curiosity doesn’t have a queue, where any question gets an answer rooted in real human understanding, and where the organization’s knowledge of its customers grows more valuable with every study rather than resetting when the project closes.

The HelloFresh journey maps directly onto three shifts Discuss believes the entire industry needs to make:

1. Projects → memory. Research shouldn’t die when the study ends. Every interview, every session, every data point should feed a system that retains it, connects it to what came before, and makes it queryable by anyone in the organization who needs it.
2. Linear → always-on. The question a product manager has on a Wednesday afternoon shouldn’t wait for the next research cycle. The infrastructure should be capable of answering it — now, from what already exists.
3. Reports → intelligence. A static deck is a snapshot. What organizations actually need is a living asset: one that can be interrogated, updated, and shared across teams without losing fidelity to the original human voices it came from.

Every study should make the next one smarter. At HelloFresh, that’s no longer a principle. It’s the operating model.

Ready to see what this looks like for your organization? Talk to us or download the Forrester Wave™: Experience Research Platforms, Q1 2026 — the independent evaluation that named Discuss a Leader in the category.