Data-driven design is powerful. Add Quantum UX and it becomes predictive, so predictive that we doubted our own numbers, and they were right.
As a neutral experiment, we combined data-driven design with Quantum UX and AI to forecast Argentina’s 2023 PASO primary, using over 11,000 data points, by far the largest set any consultancy used. Our per-candidate predictions landed within 1 to 2% of the real results, and we correctly called the outcome that every major pollster missed, including a third-placed candidate surprising everyone. The honest lesson: the raw data was so unusual that even its author distrusted it, and adjusting it ‘to look realistic’ was the only real mistake.
Key takeaways
- Data-driven design plus Quantum UX can predict user (and voter) behavior with high accuracy.
- Over 11,000 data points made this the largest dataset any consultancy used for the event.
- Per-candidate predictions fell within 1 to 2% of the actual results.
- The one real error was human: adjusting the raw data because it looked 'too unreal'.
- The moral: trust validated data over instinct, even when the data is surprising.
Why we distrust our own data
The recurring surprise in this field is that companies (and researchers) prefer instinct over real data. Here the author admits doing exactly that: the raw numbers his own user-behavior-prediction systems produced looked like a statistical aberration, so he adjusted the parameters, which is poor scientific practice. The unadjusted data turned out to be almost perfectly right.
The predictions vs reality
Argentina’s primary is a first round where coalitions total their candidates’ votes, so party-level errors compound the small per-candidate gaps. Even so, this forecast was by far the most accurate published, correctly calling a third-placed contender that no one gave more than a 20% chance, in a system with strong two-party logic.
The point: data-driven design predicts behavior
Data-driven design (distinct from UI or web design, though both should use data) is the strongest tool for predicting behavior. Married to Quantum UX, it lets you anticipate needs and adapt interfaces with unusual accuracy, which raises satisfaction and retention. If a small team could forecast an election this closely, imagine what the same rigor does for a product.
Combining data-driven design with the Quantum UX framework and AI to predict and adapt to user behavior, using large datasets and analytics.
Per-candidate predictions were within 1 to 2% of the actual results, the most accurate of any published forecast, and it correctly called the outcome pollsters missed.
More than 11,000 data points, by far the largest set any consultancy used for that event.
The only real error was human: the raw data looked so unusual that it was adjusted to seem more realistic, which turned out to be less accurate than the original data.
No. It was a neutral technical experiment; the studio states it has no political bias and does not usually work for companies in its own country.
Want decisions driven by data that predicts behavior? Talk to Dorve.
Start a project