What is a consumer simulation?
A useful consumer simulation is impossible without a live, standing model of a consumer audience. The hard part of building one is not the agent that predicts, but the infrastructure underneath that knows what consumers want in real time. Most attempts at simulation fail because they are built on static, generic data that is six to twelve months old, which cannot answer a question about now. At Hugo, we build the infrastructure first: a proprietary map of who consumers are, what they think, and what they will buy next. A consumer simulation is an agent we deploy on top of it to test a commercial decision before it goes live. This piece explains how they work, why the data layer is everything, and what is required to build a simulation that is actually useful.
Published 24 August 2026. This is an explanation of a capability, built from our own research and development. Every figure comes from a real run, and we state the method's limitations plainly.
How do consumer simulations work?
A consumer simulation works by building a model of an audience and then exposing that model to a new stimulus, like an ad campaign or a product feature. The core components are the audience model and the predictive agent. First, an audience must be mapped: who they are, what they think, what they need, and what they buy. This map cannot be a generic persona. It must be a detailed, dynamic model of the real people who make up a market. Second, an agent is built to query this map, asking how this specific group of people would respond to a specific action. The quality of the simulation's output is entirely dependent on the quality of the audience model underneath it. A simulation built on a generic or outdated understanding of consumers will only ever produce a generic or outdated answer.
Why are most consumer simulations not useful?
Most consumer simulations are not useful because they are built on the wrong foundation. They rely on static, generic training data that is often six to twelve months old. Consumer needs, however, are not static. They change daily, influenced by culture, trends, and millions of individual conversations. A simulation that uses a snapshot of the past cannot accurately predict a decision about the future. It is like trying to navigate a city with last year's map. The result is a fluent but hollow answer, one that feels plausible but has no connection to what real customers want right now. The problem is not the simulation agent itself, but the missing layer beneath it: a live, standing infrastructure that holds a real-time model of a consumer audience.
What data does a good consumer simulation need?
A useful consumer simulation must be built on a standing infrastructure that maps a specific consumer audience, continuously. This is the hard part, and it is the part we own. This infrastructure must contain four layers of understanding. First, who the consumers are, resolved to real people and their relationships, not abstract segments. Second, what they think about a category, a brand, and its competitors, in their own words. Third, what they need, including the unmet needs they express through workarounds and complaints. Fourth, what they want to buy next, down to the product and SKU level. This infrastructure cannot be a one-time study; it must be always on, capturing the shifts in consumer demand as they happen. Without this live foundation, any simulation is just a guess.
Can ChatGPT run a consumer simulation?
No. Asking a general assistant like ChatGPT or Claude to run a consumer simulation demonstrates the core problem. The model will give you a well-written, confident answer based on its vast but generic training data. That data is a snapshot of the web from months or years ago. It does not know your specific audience, what they were discussing this morning, or which product they have been trying and failing to find. The assistant's limitation is not in its reasoning ability, but in the data it can access. It lacks the live, audience-specific infrastructure required to give a commercially useful answer. The result is an answer that is right in general but wrong in particular, which is useless for a brand making a specific decision about what to launch or how to market it next week.
What is the difference between a simulation and a survey?
The difference is in how the data is sourced. A survey is an elicited method: it works by asking people questions directly. A researcher writes a questionnaire, recruits a panel of respondents, and collects their answers. This is a powerful technique for testing a concept that does not exist yet or for understanding a private decision. A consumer simulation, when built correctly, runs on observed signal. It reads the millions of public conversations, images, and videos people are already sharing to understand what they want, without a researcher in the room. It does not ask what people might do; it models what they are already doing and thinking to predict what they will do next. A survey answers the question you thought to ask. A simulation, built on a standing audience model, can surface the opportunity you had not seen yet.
What is the difference between a simulation and a focus group?
A focus group and a consumer simulation both aim to understand customer reaction, but they operate in fundamentally different ways. A focus group is an elicited method. A brand recruits 8 to 10 people into a room to discuss a new product or ad campaign. It is invaluable for getting deep, qualitative feedback on a concept from a handful of people. However, it is slow, expensive, and subject to group dynamics and moderator influence. The findings from one group may not represent the broader market. A consumer simulation runs on observed, unprompted data from thousands of people, like the 876 accounts in our New York map. It predicts how an entire audience will react based on what they already say and do in public. A focus group gives you a deep read on a few people's opinions, after you have booked the room. A simulation gives you a directional forecast of a whole audience's behavior, before you commit to the idea.
How do you build the audience for a simulation?
An audience for a simulation cannot be an off-the-shelf segment. It must be a bespoke map of the real people a brand wants to reach. Building this is the foundational step. We do this by mapping the public web, starting from known customers or communities and expanding outwards. For our research into what consumers actually wear, we used a method from survey science called chain referral to reach real, representative people that a simple search would miss. For the New York map alone, this process involved analyzing 876 accounts and 4,464 of their posts. This is not about keywords; it is about identifying the nodes and connections that form an audience. This map of people, their relationships, and their expressed needs becomes the substrate. The simulation is an agent built to act on top of it.
What does a real audience map look like?
A real audience map is not a persona document. It is a live graph of people, the things they own, and the ideas they share. Our work mapping what people wear provides a concrete example. In our New York map, we identified 832 distinct garments. More importantly, we joined them to the people who wear them. The map shows that a specific pair of white sneakers sits on seven different accounts. The same people also wear a particular crew neck t-shirt, a style of jeans, and specific leggings. The network extends from there, connecting to neighbours who carry a certain woven leather bag found on four other accounts. This is the texture of a real audience. It is a dense network of preferences and co-occurrences. As we wrote in our study, "The map is the substrate. Agents are what we build on it." A simulation queries this rich, detailed graph, not a flat list of demographics.
How does a simulation test a marketing campaign?
To test a marketing campaign, a simulation agent exposes the campaign's message and creative to the underlying audience model. Imagine a brand wants to launch a campaign for white sneakers. Our research mapping what people wear in New York provides a concrete example of the data a simulation needs. The map shows a specific model of white sneakers appearing on 7 different accounts, and it also shows that the people wearing them also own a particular crew neck t-shirt and style of jeans. A simulation would test the proposed campaign against this real context. Does the ad creative show the sneakers with items these people actually wear? Does the copy use language that reflects their real conversations about style and comfort? The simulation's output is not a score, but a specific recommendation grounded in the map: "The proposed tagline resonates with the audience's stated desire for versatility. However, the creative shows the product styled in a way that is disconnected from how this audience actually dresses. Revise the imagery to reflect the co-occurrence patterns we see in the data, pairing the sneakers with the crew tee and jeans already present in their wardrobe."
How does a simulation test a new product idea?
Testing a new product idea with a simulation starts with the unmet need. Because the underlying infrastructure continuously maps what an audience wants, it can identify gaps in the market before a brand does. A simulation agent can take a proposed product concept and check it against this database of needs. Does this new product solve a problem people are actively complaining about? Does it offer a feature they are trying to hack together from other products? For example, our mapping of consumer conversation might show a recurring complaint about the pockets in women's running shorts. A simulation could then take a new short design with a novel pocket system and predict its reception. It would base this prediction on the volume and intensity of the existing need, providing a clear signal of whether the product will land with its intended audience.
What questions can a consumer simulation answer?
A consumer simulation, built on the right infrastructure, can answer a range of critical business questions that are typically left to instinct. The applications span product, marketing, and brand strategy.
- Product: Which of these three new features will customers value most? Should we launch this product in blue or green? What is the single biggest flaw in our current top-selling product?
- Marketing: Which of these campaign messages will resonate most strongly with our target audience? Which influencer has the most credibility on this specific topic for this specific group? What is the best way to talk about our sustainability efforts without sounding generic?
- Branding: How will our customers react if we change our logo? Does our brand's tone of voice align with how our customers actually speak? What is the biggest threat to our brand perception right now?
What can a consumer simulation not do?
A consumer simulation has clear limits, and it is important to state them. It is not a crystal ball. It can only model and predict based on the public, observable data it is built on. First, it cannot predict true "black swan" events or the emergence of a need that has no precedent in existing conversation. It finds the next move, not the move after next. Second, its understanding is limited to what people choose to share in public. It works for consumer categories where people talk, but not for private B2B decisions or unexpressed thoughts. Third, it can tell you what and why, but not always how many. The signal skews toward the delighted and the annoyed, so frequency is directional, not a statistically representative measure of incidence. For a defensible market share claim, a brand still needs a traditional panel survey.
How accurate are consumer simulations?
The accuracy of a consumer simulation is not a single percentage. Its accuracy depends on the question being asked and the quality of the underlying data. When predicting the resonance of a marketing message, it can be highly accurate because it is matching the message's language against a vast corpus of the audience's own words. When forecasting the sales of a completely new product category, its accuracy is lower because it is extrapolating further from the available data. The most honest answer is that a simulation provides directional confidence. It tells you which of your available options is *most likely* to succeed. It is designed to reduce uncertainty and kill bad ideas early, not to provide a guaranteed sales figure. Its value lies in making decisions better, not in making them perfectly predictable.
Why is it hard to find real information on consumer simulations?
Finding credible information on this topic is difficult because most of what is published is marketing rather than a report of real work. Search the questions a buyer would actually ask and the results define the term, list some benefits, and never show a run: no sample, no dates, no method, nothing a reader could check or repeat. That is a low bar and almost nothing clears it, which leaves a brand comparing two vendors with little to go on beyond which one describes itself more confidently. It is why we publish our methods and our numbers, including the runs that did not work. The work is the part that can be verified.
How are Hugo's consumer simulations different?
Hugo's approach to consumer simulations is different because we start with the hardest part. We do not offer a simulation agent as a standalone tool. Instead, we first build the underlying infrastructure: a live, standing model of a brand's specific consumer audience. This is a capability we have shipped. The simulations themselves are the next layer, and they are on our roadmap. An agent built on our infrastructure will be grounded in a real-time understanding of a real audience. The goal is not to hand over a report. The agent we are building is designed to deliver a decision, in Slack, into the thread where the team is already working. This closes the loop between insight and action. Describing this as roadmap is important: it is the honest state of this layer, and a capability we are actively building, not something available today.
How do I know if I need a consumer simulation?
The question is not whether you need a specific tool, but whether the most important decisions your business makes are backed by data. Every consumer brand decides what to put in the range, what to discontinue, how to phrase a marketing claim, and which trends to act on. For most brands, most of the time, these decisions are made on instinct, experience, and the opinion of the highest-paid person in the room. This is not because of carelessness, but because the alternative was a six-week, five-figure research study that was too slow to be relevant for a decision that had to be made on a Tuesday. The right question is whether the decisions that drive your revenue are based on what your customers actually want, or on what your team believed to be true at the time. A consumer simulation is for any brand that wants to move from the second position to the first.