Table of Contents
Data-Driven apps are the future of web design and development; there’s no doubt about that. But what happens when you add Quantum UX to the mix? Data-driven design is powerful on its own. Adding Quantum UX is like infusing it with rocket fuel!
If you paid attention to our experiment last week on User Behavior Prediction, you’ll remember we used the upcoming elections in Argentina to perform a small experiment using AI and Quantum UX. Several reasons drove this choice, but one of the most significant was the fact that in the provincial elections in the previous months and weeks, all political and public opinion consultancies were wildly off in their predictions, failing to explain the staggering discrepancies. None were exceptions.
Given that, although our main offices are in Argentina, we usually don’t work for companies in our country, and we have no bias towards any politician or interested party, it seemed interesting to investigate the subject. Additionally, if everyone was wrong by truly astonishing and unprecedented percentages, what was the worst that could happen? In the worst case, we would need to adjust our experiments and gain new insights to learn from.
Data-Driven Design: Why Don’t We Trust It?
I have often written about the surprise I feel when companies prefer to believe in their instincts rather than real data. But in this case, I was the one who didn’t believe in the data that I should have trusted, based on systems that I had designed myself.
In my defense, I must say that the data was so unusual that I thought it was a statistical aberration, as can be read in the previous article that motivates this one. And although I still find it hard to believe, the fact is that the raw data my team had gathered (by far the largest number of cases any consultancy had ever dealt with, with over 11,000 data points!) were almost perfect. And the adjustment provided by our systems was too.
In fact, the biggest “mistake” we made (I made) was in the qualitative analysis, for which I take responsibility: the data were so “unreal” that I adjusted the parameters. Terrible scientific practice, I admit. But I apologize, and again: it was so unreal that it seemed impossible.
The Final Results as We Predicted Them
So, the results we predicted were as follows:
As previously explained, Argentina’s unique electoral system considers this phase as merely a first step where voters select the candidates for each party. Therefore, the candidate with the most votes doesn’t necessarily win; instead, the sum of all votes gathered by all candidates within each coalition is taken into account. In other words, the winner is not chosen for any position but wins the nomination for president by their party (this applies in cases where the party has more than one candidate).
Consequently, the final results that I predicted, once adjustments were considered and votes for each political candidate were summed up for each coalition, were as follows:
These results are more logical than what our previous data had indicated. Even so, with all the adjustments, the “La Libertad Avanza” party reached third place in a virtual tie among three parties. This is quite unusual, as no one gave more than a 20% chance to this party, and because Argentina has an antagonistic two-party logic, similar to the Democrats vs Republicans in the US or Tories vs Labour in the UK.
To understand the problem: the figures we were handling were extraordinary, even when I (mea culpa) adjusted the algorithm to make them more “realistic.”
Well, the candidate who was expected to finish third surprised everyone and won. Not only that, he won by only 1% more than what we had predicted at the beginning before doubting our own data-driven system.
Here are the data, as they were published in the newspaper “La Nación.” The black line at the top represents the actual result obtained by the candidate, the bars below what consultancies predicted.
#1 Javier Milei
#2 Sergio Massa
#3 Patricia Bullrich
#4 Horacio Rodríguez Larreta
#5 Juan Grabois
There’s more, but I won’t continue to keep this brief. You can read all the details in the article that inspired this one.
The images provided above represent individual candidates. The differences between reality and our predictions are quite minimal, falling within a range of 1-2%, with one exception. However, when considering the results by party, the discrepancies become more significant because those small percentages accumulate. Despite this, we were by far the most accurate in predicting the PASO election results. Considering the time constraints and limited resources, our performance was not too shabby, don’t you think?
Data-Driven Design to the Rescue
Data-Driven Design (not to be confused with UI design or web design, although both can and should include data-based design) is the ultimate tool for effective user behavior prediction. And what is better than knowing what your users want and how to act accordingly?
We can improve your business!
Let us help you with the best solutions for your business.
It only takes one step, you're one click away from getting guaranteed results!I want to improve my business NOW!