Using data science to customize the user experience

Sigalei

UX UI Designer

Dec 2020 - Dec 2021

Why customize the search results?

It can be an important ally in scenarios with high complexity and volume of information where the user has too many options and needs to make a decision.
‍
From a business perspective, it can help to increase engagement, prevent errors, give more agility and fluidity to the users, and consequently, scale your product, increasing its revenue.

The problem

Different data formats:
- More than 300 hundred government websites with its own formats.
- Non-government platforms such as news, social media, etc.

Time:
‍Our users were used to spending hours manually monitoring their stakeholders (anyone that could impact their business) or paying an expensive consultancy agency to do the job.

Control:
‍You can only monitor someone who has already been identified. Therefore, there were stakeholders that became relevant for some reason, which were not being monitored.

Ways to customize the search results?

There are many inputs that we could use to build this. Such as:
- Social relationships between users
- Social media activity
- Transnational relationships between users (date, amount, frequency)
- Phone contacts
- Geolocation and more.

The algorithm behind the scenes

Through a discovery process, we were able to build an artificial intelligence that we called MaIS, which searches and monitors specific words or people, prioritizing the result accordingly to the user’s profile.
‍
As a designer, I needed to understand very well how the algorithm worked to know what minimum user input was required to:
- generate relevant information;
- guide how to configure this monitoring;
- what is the best way for the user to visualize this data, and more.
‍
Then I called the Data scientist so we could build a diagram explaining the behaviour behind it and it remained as a legacy for other teams as well.

How we customize the search results?

Every search is a decision-making process. There is a user's intent when searching, the perception if the displayed items are satisfactory and the decision to interact with the result.

Tags with relevance %
‍Our users were used to spending hours manually monitoring their stakeholders (anyone that could impact their business) or paying an expensive consultancy agency to do the job.

Before: the user had to analyze all the variables (reach of the information, likes, shares, source, public relevance, ambiguous words, etc) to decide what was relevant to their business.

Stakeholder identifier
‍We added a section with mentioned stakeholders with a redirect link to their accounts.

Before: the user had to search on the browser of each website for the stakeholder name, most of the time missing out the opportunity to identify new stakeholders related to the topic.

Smart keyword search
‍Possibility to search for multiple and combined keywords, plus options to prioritize or exclude other keywords found on the search.

Before: the user had to think about what they would look for, missing the opportunity to find other trends related to the same topic, leading to the loss of strategic opportunities.

Extra features
‍We also added over 3000 fonts, charts, word clouds, filters, etc.
So our users could save all the settings, thus customizing what to monitor and allowing the AI to learn from it.

Results

Provided the user with data-based decision-making:
- Understands complex information
- Identifies patterns and opportunities
- Prevents risks

User’s psychological needs met:
‍Effective professional performance, control, and safety.

Business metrics:
- Increased annual sales by around half a million, including upsells and pre-launch sales.
- 45% increase in NPS (Net Promoter Score).
- R$ 200 (reais) pay raise for all company employees.

Values of: December/2021.

Using data science to customize the user experience

More about it