What data do you need to build accurate digital twins?

"What do we need to give you?"

It's the second question every team asks, right after how accurate is it? And the answer is less than most people expect.

Echo builds digital twins from whatever you have. Sometimes that's a one-page description of who you want to learn from. Sometimes it's a folder of past studies. Sometimes it's dozens of real interviews with your own customers. Each one builds a different kind of audience. We call these levels of audience fidelity: how closely an audience is anchored in real voices.

Level 1: a brief

What you bring: a description of the audience you want to learn from. Who they are, what they do, what matters to them.

What Echo does: turns that description into a population of distinct personas, each with its own background, personality and point of view. Echo fills in the gaps from public knowledge about similar people.

What it unlocks: speed. You can test an instinct the same day, before betting anything on it.

Less can be more. In one pilot, the audience description came out of a single roundtable between the marketing, sales and insights teams. Together they wrote down only what they knew to be objectively true about their customers: no wishful thinking, no assumptions. The result was a simple one-pager. Measured against real interviews, the twins built from it reached an accuracy of 83%. A sharp, honest brief beats a thick pile of loosely related documents.

Know the trade-off: a Level 1 audience is only as specific as the brief behind it. The quality of the description matters more than its length.

Level 2: your existing research

What you bring: the material you already have. Past studies, segmentation decks, survey reports, market research, customer personas.

What Echo does: extracts the audience from those documents: who they are, how they behave, how common each behaviour is, where the segments differ. Your twins are primed with what you already know about your market.

What it unlocks: research that keeps working. The studies gathering dust in a shared drive become a living audience you can question again, from a new angle, whenever a new question comes up.

Level 3: your customers' own voices

What you bring: real data from your audience, down to the individual respondent. Interview transcripts, open survey answers, customer feedback: respondent-level data, not just the summary slide.

What Echo does: grounds every persona in real language. Echo discovers the sub-groups in your data (the groups that emerge from what people actually say, not the ones you assumed up front) and builds twins that draw on voices like their own. When a twin answers a question, it answers from the concerns, words and contradictions your real customers expressed.

What it unlocks: the highest confidence. Answers you can defend in the boardroom, segments that come from evidence, and twins that sound like the people you actually sell to.

Know the trade-off: it takes real respondent data to start. If you don't have it for a new audience, we can source suitable research data for you, or run a small study with real people to build the foundation.

Which level do you need?

It depends on the decision, not on how much data you have lying around.

You don't need a big data project to begin. A sharp question and an honest audience description are often enough to start learning. When the stakes go up, you move up a level.

What's next: a customer panel that never goes stale

Today, every Echo audience is a snapshot of what you know at a given moment. We're working towards the next level: live audiences.

Picture a panel of your customers that updates itself as your market moves, fed by the signals you already collect. Always current, always in the room. Before a pricing change, a campaign or a product decision, you check it against your customers as they are today, not as they were in last year's study.

That's where we're heading.

Your data stays yours

Whatever level you work at, your data is used only to build your audiences and run your studies. It's never used to train AI models, and never shared with anyone else. Read more on our Trust page.