Is 100% accuracy the goal?

Why not 100%?

We score 85–93% against real interviews. The obvious question: why not 100%?

The better question: would you even want it?

What lives in the gap?

Every validated Echo study shows three things side by side:

• Matched: themes that come up in both the synthetic and the real interviews
• Missed: themes real people raised that Echo didn't
• Added: themes Echo raised that real people didn't

The matched part is the 85–93%. It confirms your direction. The interesting conversation is about the other two.

Why does Echo miss things?

Echo can miss themes, like any method. Some reasons:

• Lived context. A real person mentions the rainy Tuesday their delivery came late. That kind of detail lives in experience, not in data.
• Culture and emotion. Local habits, inside jokes, a strong feeling about a brand that nobody ever wrote down.
• Plain human irrationality. People do things that make no sense on paper. That's part of what makes them human.

Missed themes are a reminder of why real customers stay the ground truth.

Why would Echo say things humans don't?

This is where it gets interesting. There are several reasons an AI respondent might raise a theme that real respondents didn't:

• Social desirability. In front of an interviewer, people don't always say what they think. They avoid sounding cheap, lazy or uninformed.
• Tacit needs. Some things people feel but can't put into words, until someone else does.
• Small samples. Twelve real interviews can't cover everything a broader population would say. A theme missing from twelve people may well exist in twelve thousand.
• New connections. A model can link ideas that customers haven't linked yet.

Not every added theme is a hidden truth. Some are simply wrong. But some are signals.

Is the gap a bug or an opportunity?

Both, and that's the point.

An added theme isn't a fact. It's a hypothesis worth testing: a possible unmet need, a new angle on a proposition, a lead for innovation. Teams that treat the gap as a list of questions for their next round of real interviews often find their most interesting insights there.

Innovation rarely lives in what customers already say. It lives in what nobody has said yet.

Is 100% even achievable?

Not in any meaningful way. Ask two groups of real customers the same questions and they won't agree 100% with each other either. That's why we score Echo against the natural disagreement between real people, not against perfection (see Measuring the echo).

A synthetic study that matched one human group perfectly wouldn't be more accurate. It would just be a copy of one group, with that group's blind spots included.

Is 100% desired?

Here's where we land. We want Echo to be close enough to trust the direction, and open enough to surprise you.

We'll keep pushing the accuracy up. Every engagement teaches us something. But we'll also keep showing you the gap, openly, in every validated study. Because that's often where the next idea is.

The short version

85–93% tells you you're on the right track. The rest might tell you where to go next.