318 simulated customers beat 4,426 survey respondents

Triumph sells lingerie to millions of women across Europe, in six distinct customer segments from Casualist to Fashionista. Every range, price point and channel decision depends on one question: how will these customers respond?

The classic answer is a survey. But people don’t report their own shopping accurately. They say they buy more often than they do, shop in store more than they do, and guess prices from memory. Decisions built on those answers start from a skewed picture.

Triumph needed a customer view it could trust before deciding, and one it could go back to again and again, not a one-off study that’s outdated by the time it lands.

Elaiia built an AI customer simulation of Triumph’s customer base: 318 simulated customers across the six segments in Germany, France and Italy. We then asked Elaiia’s simulated customers about six purchase behaviours and checked every answer against what real customers did in Triumph’s order data. A consumer survey, matched segment by segment, ran alongside as the conventional reference.

The six behaviours: online vs. store share, bras vs. briefs share of spend, price per bra and per brief, and purchase frequency for each.

Every number was computed in R and re-derived independently in Python: 958 checks, 0 inconsistencies. Each verdict is backed by 2,000 bootstrap resamples, and a result only counts as “clear” when the 95% interval excludes a tie.

Elaiia was closer to real customers on 5 of 6 behaviours, 4 of them clearly. The survey was closer on one.

The biggest gap is habit. Survey respondents said they buy briefs 5.65 times a year; real customers buy 2.29. Elaiia said 2.87. On channel, the survey overstated online by 6.8 points, Elaiia by 2.4.

The one miss is bra price: Elaiia overshot by €12.95, the survey undershot by €7.68. Both were asked their maximum willingness to pay, which customers’ actual paid prices don’t always reach.

It holds segment by segment. Across all 36 segment × behaviour comparisons, Elaiia was closer in 24, 13 of them clearly. Elaiia won 5 of 6 behaviours for Stylists and Aspiring Stylists, and predicted brief purchase frequency better in all six segments. Including every country (565 simulated customers vs. 8,297 real ones), the picture stays the same: 5 of 6.

Triumph now owns a customer population that behaves like its real customers, not like survey respondents describing themselves. And it doesn’t expire after one study.

• Decide on reality, not recall. Range, pricing and channel questions can be tested against simulated customers that track actual buying behaviour.

• Every segment, on demand. All six segments are there to ask, including small ones like Fashionistas, where real samples are hard to come by.

• It compounds. The simulated customers stay, remember prior studies and get sharper with each simulation Triumph runs.

This is what Elaiia is built for: a persistent customer simulation layer that shows how your customers are likely to respond, before you decide.